rangifer’s diary: pt. xcviii

Guan Yunchang

If you’re a reader my diary, then you’re already familiar with Taima, a.k.a. Boymoder, a.k.a. Tacgnol, a.k.a. Hanyou, a.k.a. Nyanners, a.k.a. Numidium, a.k.a. Gambolpuddy, a.k.a. Hitodama, a.k.a. midorin, a.k.a. Girlmoder, a.k.a. Inugami… [:

Well, I want to promote her new diary for her new character (IGN Yunchang): Guan Yunchang’s World Tour Diary! Partly because Taima made it and I love her, but also because if you like my diary, I think you’ll probably like hers too! If you like MapleStory diaries/journals that are high-effort, full of fun images, and document people doing weird shit in MapleStory, check out Guan Yunchang, because it’s really good stuff. Plus, she’s funnier than I am.

You can read the diary on Taima’s website, and you can also find her on Codeberg as well. <3

Une randonnée à Monaco

In pt. xcii of this diary, I finally had some good luck with Chaos Scrolling (CSing) Pink Gaia Capes (PGCs), after a number of wasted Chaos Scrolls that we’re not going to talk about… But I did document those, as well… The point is, I was able to make an 8⁢ WAtk, 4⁢ slot PGC. I’ve been using this as my main cape on almost all of my characters for a while now, & as a result, I am too scared to CS it again. If it loses WAtk, I just lost my damn cape!!

But okay, maybe someone is willing to trade me their ≥10⁢ WAtk cape that has 0⁢ slots in exchange for my 8⧸4^[1], if I “top up” with an appropriate number of Prestigious Coins (pCoins). There are still some issues, though. In particular, is someone willing to make that kind of trade? I mean, sure, they lose a WAtk here or there, but they get four slots? Eh…? Well, I have run into a few interested people, but that still leaves the question of exactly how many pCoins I should need to add on top to sweeten the deal. Sure, it obviously depends on exactly how much WAtk their cape has, but other than that, it’s not obvious to me at all how many pCoins makes for an appropriate offer. Any economists in the chat…?

Not having any damn idea how many pCoins makes for an appropriate offer is crippling here, because as I desperately search for interested traders, any success in my search — viz. any person whom I find who is interested in my trade offer — is going to come to me asking how many pCoins I’m gonna give them. Then what?!? I don’t fuccing know!!! If I offer too high, I’m ripping myself off; or perhaps worse, I won’t even be able to afford it at all without seriously liquidating a lot of shit really fast & maybe taking out a loan. If I offer even one (1) pCoin lower than what they’re hoping I’ll offer, I won’t get any response. People are generally allergic to offering at all, be it an initial offer or, as in this case, a counteroffer. I don’t really know why that is, since they obviously have a decent idea of how much they’re willing to accept when they do or do not accept my offer, but I guess it’s a psychological thing or whatever. I don’t know, I don’t talk to people.

But what is a “Chaos Scroll 60%”, really…?

Now, I’m not an economist, but I do know some things sometimes.^[2] For example, I know how a CS works. When you use a CS on an equipment item, the equip’s slot count decreases by 1. Furthermore, there’s a 60% probability of the CS “succeeding”, & otherwise ($\mathrm{100\%}-\mathrm{60\%}=\mathrm{40\%}$ probability) it “fails”. If it fails, then nothing else happens. If it succeeds, then each nonzero stat on the item is independently increased by a random amount. In particular, the $𝑖$^th stat on the item is increased by ${𝑋}_{𝑖}$, where ${𝑋}_{𝑖}\sim 𝒰\left\{\mathrm{-5},5\right\}$.^[7] Finally, each negative stat is set to zero.

Because stat increases are independent of one another, we can focus on just one stat, which is what we’re going to do anyway — we only care about WAtk here. If we do this, & we assume that the stat is at least 5 to begin with (equivalently, we ignore the final step, wherein negative stats are set to zero), it’s easy to see the distribution of how much the stat is increased by the use of a single CS. Assuming a success, the distribution is $𝒰\left\{\mathrm{-5},5\right\}$ as stated above, so that each outcome has a $\frac{1}{11}$ probability of occurring.

But this is assuming a success, which in reality only has a $\mathrm{60\%}=\frac{3}{5}$ probability; thus, each success outcome ultimately has a probability of $\frac{3}{5}\cdot \frac{1}{11}=\frac{3}{55}$. Because one of the success outcomes is actually that the stat increases by 0 (because 0 is an integer, and $\mathrm{-5}\le 0\le 5$), the overall probability of the stat not changing is $\mathrm{40\%}+\frac{3}{55}=\frac{2}{5}+\frac{3}{55}=\frac{5}{11}=45.\overline{45}\mathrm{\dots \%}$. The remaining $\frac{6}{11}$ are cases where the stat is changed: $\frac{3}{11}=27.\overline{27}\mathrm{\dots \%}$ probability of increasing, and $\frac{3}{11}$ probability of decreasing.

This is pretty dire, as it means that there’s an $\frac{8}{11}=72.\overline{72}\mathrm{\dots \%}$ probability that you don’t get what you want, assuming that what you want is for the stat to increase by more than zero. For reference, these are actually mildly worse odds — in terms of “probability of getting what you want” — than you’d have when using a 30% scroll. Yikes.

So, that’s cool & all, & it’s important that we actually know how a CS works. But we pretty much knew that anyway, right? And how does that help me decide how many pCoins to offer people? Unfortunately, I’m not really sure what the answer to the latter question is. Again, any economists reading this plz help me…

Bridging the gap

Then, one day, I attracted serious interest from someone with a PGC that had 10⁢ WAtk & 0⁢ slots. Actually getting interest from someone — after a nontrivial amount of time posting in the #free-market channel of the MapleLegends Discord™^® server — was what really kicked me in the ass & got me to take this problem seriously. The interested party in question didn’t make any offers, nor did they indicate/imply a general ballpark of how many pCoins I would have to contribute, so I was flying blind.

My initial instinct was just to bridge the gap, so to speak, between an 8⧸4 cape & a 10⧸0 cape. What I mean is something vaguely like this: if we have an 8⧸4, & we attempt to turn it into a 10⧸0 by CSing it, how many pCoins do we expect to spend? There’s kind of a lot to unpack here, though:

• First of all, “attempt to turn it into a 10⧸0” is overly specific. For example, if we end up with an 11⧸0, that’s not strictly a 10⧸0, but we also certainly wouldn’t say that we “failed to reach our goal”, right? Similarly, if we end up with a 10⧸1, that’s also not strictly a 10⧸0, but it’s strictly better than a 10⧸0, & moreover, we can obtain exactly a 10⧸0 by using any single scroll that isn’t dark, isn’t a CS, & isn’t a CSS (like cape DEX 60%, to give just one example).

  We can “fix” this by amending “attempt to turn it into a 10⧸0” to “attempt to turn it into a $𝑤$⧸$𝑠$ (where $𝑤\ge 10$)”. Notice that $𝑠$ is not explicitly constrained; logically, we already know that $𝑠\in \left\{0,1,2,3\right\}$, & we don’t really need to constrain any further than that.

  This “fix” results in a hypothetical scenario that tends to underestimate how much economic value we get out of the resulting cape, for two reasons:

  • Some (but not all) failed capes still have nontrivial economic value. For example, an 8⧸0 is still worth something — just obviously significantly less than the cape that we started with.
  • Some (but not all) successful capes are nontrivially more valuable than the 10⧸0 that is being offered to me. For example, an 11⧸0, or a 10⧸2.

  This is somewhat problematic, but is nonetheless balanced out by other simplifications (as we shall see), as well as by the simplicity that it confers upon the model.

• It also goes without saying that there’s absolutely no guarantee — no matter how rich we are — that we can turn the 8⧸4 that we’re given into a $𝑤$⧸$𝑠$ — where $𝑤\ge 10$. Being rich just means that we can afford to throw unlimited CSes at it, but if — to give just one example — every CS lowers the WAtk, then we’re definitely screwed anyway. Thus, we cannot quantify “how many pCoins […] we expect to spend”.

  We can “fix” this by changing the premise so that we have access to as many 8⧸4 capes as we need to get our first success, & no more. Every time that a cape gets to a point that we consider “failed”, we discard it & start with a fresh 8⧸4. As soon as we get even one successful cape, we immediately stop everything — even if the successful cape still has more than zero slots remaining. To be clear, we have exactly one cape at any given time, because failed capes are immediately discarded.

  This “fix” results in a hypothetical scenario that tends to overestimate how much net economic value we get out of the result. The reason is simple: we’re magically pulling 8⧸4 capes out of thin air, because the first one is the only one that’s actually provided in the original problem statement. The second, third, &c. capes are all ex nihilō, & we started with the somewhat obvious premise that an 8⧸4 is worth (i.e. is economically valued at) quite a bit — certainly a nontrivial amount. This is somewhat problematic, but is nonetheless balanced out by other simplifications (like the one above), as well as by the simplicity that it confers upon the model.

• Nothing about our process of scrolling actually explicitly involves pCoins, & yet we ask “how many pCoins do we expect to spend?”. What gives? Well, the actual source of expenditure here is clearly the use of CSes. So, if we can associate CSes with a price in pCoins, we’re good to go.

  Consider that there are effectively just two ways to obtain CSes in MapleLegends at the time of this writing:

  • Buying them directly from the Prestige Shop at 100⁢ pCoins + 10M⁢ mesos each,
  • or randomly generating them via Gachapon, limited-time event mechanics (summer event, Χmas event^[3], &c.), Scroll of Secrets, &c..

  If a player obtains a CS via the latter method, & wants to liquidate it to afford directly useful things — rather than use it to make a 0⁢ WAtk PGC or whatever — then it seems that they have only one choice: sell it to the Prestige Shop. CSes are untradeable, so they cannot be sold to other players. Or can they? Because selling to the Prestige Shop grants a single MCP5, which contains an average of 90⁢ pCoins, another player might outbid the Prestige Shop by offering, say, 93⁢ pCoins for the CS.

  But wait — I literally just said that CSes are untradeable? Hello?? Well, if the item that you want to CS is tradeable (e.g. a PGC), then you can trade it to the holder of the CS, have them use the CS on your item, pay them the 93⁢ pCoins, & get your (now probably ruined) equipment item back. It’s called CS service (or CS svc, for short). Indeed, this is how I made my 8⧸4 back in pt. xcii of this very diary.

  93⁢ pCoins seems to be about the going rate for CS svc these days, so we can equate 1⁢ CS with 93⁢ pCoins, & be on our way.

• What about CSSes?? What if we end up with a 9⧸0 or something, & still want to make it into a $𝑤$⧸0 (where $𝑤\ge 10$)???

  Stop. Just stop. Please… The theoretical model is already shite, & you’re just trying to rub it in my face… 😭

Sniff… Ahem. Anyway, erm, where was I? Oh yeah, we wanna actually use this hypothetical scenario to approximate, or at least give some intuition for, the answer to our original problem. So, uhm… how do we do that, exactly?

Una gita a Monaco (con un pitone)

Ever been to le Casino de Monte-Carlo? Me neither. Nevertheless, we’re going to use Monte Carlo methods to squeeze an answer out of this bad boy. What does that mean? It just means that I’m going to write a shitty simulation in Python, and run that. I know, it sounded more glorious when we used fancy terms for it.

I’m going to admit to you right now that I have poopy doo-doo brain, & so the first time that I tried this, I accidentally forgor about the fact that Chaos Scrolls can fail. So I was accidentally working with *“Chaos Scroll 100%”s — which do not exist — instead of actual CSes. But we’re past that now! 🤡

Right, so, we need that shitty Python script. We’re working with randomness to do that Monte Carlo shit, so we’re gonna need…

import random

Right, okay. And we’re gonna need to do a lot of trials to get a crisp idea of the probability distribution underlying our model, so we’ll need a loop for that…

for _ in range(ATTEMPTS):

Okay, sure. We can make ATTEMPTS some large constant integer, like a million, or ten million, or whatever.

Within each attempt, we want to keep track of the state of the cape that we’re attempting to scroll. The only state that we care about is its WAtk & its slots, so…

for _ in range(ATTEMPTS):
    watk, slots = 8, 4

Starting with 8⧸4, just like the original problem. All checks out.

Within each attempt, we want to keep slamming the cape with CSes until we reach the goal, right…? Oh, or until we run out of slots. Of course…

for _ in range(ATTEMPTS):
    watk, slots = 8, 4
    while slots > 0 and watk < 10:

Well, actually, maybe not. I mean, if the WAtk starts getting really low, then surely we’re not stupid enough to keep slamming it? The obvious case is if we reach 0⁢ WAtk; then it truly makes absolutely no sense to keep CSing, because there’s no WAtk left to increase! We can capture this behaviour by setting a bail threshold; if the WAtk gets to the bail threshold or lower, then we consider the cape to be “failed” & discard it:

for _ in range(ATTEMPTS):
    watk, slots = 8, 4
    while slots > 0 and watk < 10 and watk > BAIL_THRESHOLD:

Like ATTEMPTS, BAIL_THRESHOLD is gonna be a constant parameter that we can tune to our liking. The choice of bail threshold is somewhat arbitrary, but we’ll get back to it later.

Now comes the part where we actually, you know, use the CS. Just like in the “But what is a “Chaos Scroll 60%”, really…?” section above, the first thing that happens is that we lose a slot:

for _ in range(ATTEMPTS):
    watk, slots = 8, 4
    while slots > 0 and watk < 10 and watk > BAIL_THRESHOLD:
        slots -= 1

Okay, easy enough. Now we need to see whether the CS succeeds. Python’s random module of the standard library provides functions for several commonly-used probability distributions like e.g. the beta distribution, the normal distribution, the Pareto distribution, &c.. But our good friend the Bernoulli distribution appears to be missing! What gives?

Well, we can actually use the inconspicuously-named random.random function to easily obtain a Bernoulli distribution. Calling random.random (which takes no arguments) gets us a floating-point number that is uniformly distributed^[4] over the interval $\left[0,1\right)$^[5]. By checking whether the result is strictly less than the $𝑝$ of our Bernoulli distribution (in our case, $𝑝=\mathrm{60\%}$), we get what we wanted:

for _ in range(ATTEMPTS):
    watk, slots = 8, 4
    while slots > 0 and watk < 10 and watk > BAIL_THRESHOLD:
        slots -= 1
        if random.random() < 0.6:

Naturally, if the CS passes, then we adjust the WAtk value of the cape. In this case, Python’s random module has our back, with a function called random.randint that does exactly what we want:

for _ in range(ATTEMPTS):
    watk, slots = 8, 4
    while slots > 0 and watk < 10 and watk > BAIL_THRESHOLD:
        slots -= 1
        if random.random() < 0.6:
            watk += random.randint(-5, 5)

We don’t have to worry about WAtk values below 1 being an issue (at least, not yet), as we definitely bail if we get that low. But we do need to keep track of how many successful capes we make, & we also need to keep track of how many CSes we use:

successes, cs_used = 0, 0
for _ in range(ATTEMPTS):
    watk, slots = 8, 4
    while slots > 0 and watk < 10 and watk > BAIL_THRESHOLD:
        slots -= 1
        cs_used += 1
        if random.random() < 0.6:
            watk += random.randint(-5, 5)

    if watk >= 10:
        successes += 1

After we do all that calculation, we can output the raw data with some simple print statements:

print(f"Successes: {successes}")
print(f"CSes used: {cs_used}")

This is already useful, but we’re gonna need to do some nasty arithmetic to get to the pCoin values that we’re really after. So, let’s make our shitty Python script do the dirty work there, as well. Consider what proportion of our capes are successful:

success_rate = successes / ATTEMPTS

Also consider the average number of CSes that we use per attempt, regardless of whether or not the attempt is successful:

cs_per_attempt = cs_used / ATTEMPTS

If we wanna get the average number of CSes that we use per successful attempt, then we have to divide this out by the success rate. If the success rate is quite low, then the average number of CSes per success is gonna be quite high, as we’re wasting a ton of CSes on failures before we ever get a success:

cs_per_success = cs_per_attempt / success_rate

But we want the average number of pCoins per success, not the average number of CSes. We already assumed that we’re buying CS svc at 93⁢ pCoins a pop, so that’s easy:

pc_per_success = cs_per_success * 93

And finally, we can print that out as well, as it’s the number that we care about the most:

print(f"Expected cost per success: {pc_per_success} pCoins")

We’ve still two constants that we’ve not explicitly given values to: ATTEMPTS & BAIL_THRESHOLD. The value of ATTEMPTS is just some big integer like a million or whatever, as I mentioned before. Making ATTEMPTS smaller makes the program run faster (fewer samples), but larger values make the law of large numbers kick in more, making the results generally more accurate.

BAIL_THRESHOLD is a little trickier. The general vibe that we get is that larger bail thresholds will cause our ultimate result — measured in pCoins — to be smaller, & vice versā. The reason for this is that starting over with a new cape costs nothing in our model, so if we bail out as soon as things start going downhill, then we tend to use fewer CSes. Of course, we can’t set the bail threshold to be greater than or equal to our starting WAtk, as we would bail out before even trying! Furthermore, a bail threshold of less than 1 is totally useless. We’ll consider two possible bail thresholds, each with its own justification:

• We can set the bail threshold to one less than our starting WAtk (BAIL_THRESHOLD = 7), so that we bail any time that we’re actually losing WAtk relative to our starting point.
• We can set the bail threshold to the WAtk of a perfect clean PGC (BAIL_THRESHOLD = 4), so that we bail any time that the cape gets to be strictly worse than a perfect clean PGC, and thus we’d rather just buy one of those & start over.

Either way, we’re actually still missing something slightly more subtle here, as far as bailing goes. If we have a cape that has enough WAtk to still be above the bail threshold, but only very few (e.g. 1) slots left, then it might not even be possible to reach the goal, which is a very good reason to bail indeed. In this case, it’s not actually possible for this to happen, so we’re not really going to worry about it. The fact that a single CS can possibly increase the WAtk by a whopping 5(!!) makes this situation unlikely or impossible, unless your goal is much higher than your starting point (or your bail threshold is unreasonably low)… Which it isn’t, in our case.

Nonetheless, this reveals a flaw in exactly how we choose to keep going or to bail out: to “rationally” know whether or not we “should” bail out, we actually need to know the exact probability of the current cape succeeding, as well as the average number of CSes that we use for such a success, as well as the “CS per success” that we’d get if we started over with a fresh 8⧸4. Or something like that. The point is, we need to assess not just the WAtk, but also the slots, & the exact value of starting over, to know how dire our situation is, & whether “CS per success” is better minimised by bailing, or by continuing with the current cape. But that sounds complicated… Wah. 😢 Plus, this model is already so bad to begin with, that adding that level of sophistication to this one specific aspect of it is truly silly. Let’s just keep it simple, & hope that the result is still helpful…

Let’s give it a go with BAIL_THRESHOLD = 7:

Successes: 421793
CSes used: 1899694
Expected cost per success: 418.8584021071948 pCoins

Whew, alright. That’s 418 or 419⁢ pCoins. That’s… a lot of pCoins. Don’t worry, though — it gets worse! Let’s try with BAIL_THRESHOLD = 4:

Successes: 485535
CSes used: 2429241
Expected cost per success: 465.2999536593655 pCoins

As expected, the cost is even greater: 465⁢ pCoins. That’s, uh, relatively not terrible though, as it’s only about ≈1.1 times as many pCoins as with BAIL_THRESHOLD = 7.

Note that both of the above are just particular runs of the program; with the same starting values, you will still get something slightly different every time. That’s the power of Monte Carlo!

Still, this is potentially useful stuff. It’s unclear whether my model is “a bit shit”, “shit”, or “unsalvageable nonsense”, but so long as it’s not the last one of these things, the results may have some use after all. That being said, the output of the program is a bit opaque at the moment. If we keep track of more data, then we can get an idea of what the capes are looking like when we stop scrolling them — either because the cape succeeded, or because it “failed” & we bailed.

To keep track of the resulting WAtk values of the capes, we can store an array of natural numbers: the 0^th number in the array counts how many capes ended with 0⁢ WAtk, the 1^st number in the array counts how many capes ended with 1⁢ WAtk, & so on. And we can do the same with slots, as well:

successes, cs_used = 0, 0
watk_results = [0] * (SUCCESS_THRESHOLD + 5)
slots_results = [0] * STARTING_SLOTS
for _ in range(ATTEMPTS):
    […]

Instead of writing 10 everywhere, we can use a constant called SUCCESS_THRESHOLD, & likewise for the number of slots that we start with (STARTING_SLOTS = 4). The + 5 here is because 5 is the most that the WAtk can increase as the result of a single CS. Note that in Python, multiplying (*) an array by an integer concatenates the array with itself that many times, e.g. [1, 2] * 3 == [1, 2, 1, 2, 1, 2].

Now, negative WAtk values are actually a problem, so we clamp those to 0⁢ WAtk, and then we update our watk_results & slots_results arrays:

for _ in range(ATTEMPTS):
    […]

    watk = max(watk, 0)
    if watk >= SUCCESS_THRESHOLD:
        successes += 1

    watk_results[watk] += 1
    slots_results[slots] += 1

Let’s try just looking at averages:

avg_watk = sum(watk * n for watk, n in enumerate(watk_results)) / ATTEMPTS
avg_slots = sum(slots * n for slots, n in enumerate(slots_results)) / ATTEMPTS

print(f"Successes: {successes}")
print(f"CSes used: {cs_used}")
print(f"Avg resulting WAtk: {avg_watk}")
print(f"Avg resulting slots: {avg_slots}")

And we give it another go with BAIL_THRESHOLD = 7:

Successes: 422184
CSes used: 1901113
Avg resulting WAtk: 7.99923
Avg resulting slots: 2.098887
Expected cost per success: 418.7830637826161 pCoins

Perhaps unsurprisingly, the average WAtk that we end with is the same as what we started with. This is a reflexion of the fact that that’s exactly what Chaos Scrolls do, on average: nothing^[6]. Well, except that they lower the number of slots by one… And indeed, on average, we’re losing about ≈1.9⁢ slots before bailing out or succeeding.

Let’s try BAIL_THRESHOLD = 4 again?:

Successes: 485721
CSes used: 2430081
Avg resulting WAtk: 7.999964
Avg resulting slots: 1.569919
Expected cost per success: 465.2826066816135 pCoins

Again, the average WAtk that we end with is the same as what we started with. This time, though, we’re losing considerably more slots: about ≈2.437⁢ slots lost, on average. That’s a little more than half a slot on top of what we were losing with BAIL_THRESHOLD = 7.

I’m not satisfied with seeing that the WAtk doesn’t change on average, though. I wanna see something more than just an average; so let’s plot it, instead (BAIL_THRESHOLD = 7):

Although this is obtained via a Monte Carlo method, you should read this in basically the same way that you’d read a PMF.

Let’s again try BAIL_THRESHOLD = 4, as well:

Check out cape_scroller.py for the full code listing.

I’m not convinced!

I’m not convinced, & you can too shouldn’t be either! In my defence, I think I wrote the first version of this Python script in like 15 minutes after that person DM’d me on Discord™^®.

Nevertheless, it’s a bit of fun, & maybe we learned something. Like… that the Monte Carlo Casino isn’t just for rich people, if you’re willing to write a little garbage code here or there. Or… maybe that I should be offering like 400〜500⁢ pCoins to top up for 10〜11⁢ WAtk capes…? 300〜600……? Eh… I guess the real lesson is that WAtk capes are like the Monte Carlo of MapleStory: gambling is the name of the game, & the only people playing the game are rich!! 🤑

Footnotes for “Une randonnée à Monaco”

1. [↑] In this context, $𝑤$⧸$𝑠$ means “$𝑤$⁢ WAtk, $𝑠$⁢ slot [cape]”.

2. [↑] The words “some” and “sometimes” are doing a lot of heavy lifting here…

3. [↑] No longer observed by MapleLegends…

4. [↑] Here, I’m linking to the Wikipedia article on the continuous uniform distribution. Of course, random.random cannot possibly actually sample from such a distribution, as it would need infinitely many bits to represent each real-valued sample. Thus, one might be tempted to instead link to the Wikipedia article on the discrete uniform distribution, but that would actually be even more wrong (in a sense). This is because floating-point numbers are inherently non-uniform in their distribution, by design. The further away that you get from zero, the more sparse they get — that is, the larger the increase in numerical value when you go from one floating-point number to the next-larger floating-point number.

   It’s totally possible — indeed, it’s not even particularly difficult — to write a function that returns a floating-point number selected with a discrete uniform distribution over a given interval. That is, it selects each floating-point number that lies within that interval with equal probability. But this is almost never what you actually want. Instead, functions like random.random (which are included in many standard libraries of many programming languages, albeit under various names) have a different logic to them: instead of selecting any floating-point number within that interval with equal probability, they strategically eliminate certain floating-point numbers in that interval from consideration, & then act as something like a discrete uniform distribution.

   By strategically eliminating certain numbers, the result is uniform, effectively flattening the exponential curve that’s inherent to the floating-point representation. The end goal is essentially to approximate a continuous uniform distribution, which is why it’s perhaps better to ✨⁠spiritually⁠✨ think of it as sampling from a continuous uniform distribution, rather than from a discrete one — because it’s not a discrete uniform distribution from the perspective of the floating-point representation! It’s only a discrete uniform distribution if we get to pick & choose which numbers are included…

   See also (archived).

   A footnote within a footnote

   Does it make sense to go at this from the other direction? Functions like random.random, <rand::distributions::Standard as Distribution<f64>>::sample, Math.random, &c. seem to do basically what I described above; but it seems to me that it’s possible (in principle, at least) to eliminate bias in the numerical magnitude of the result — i.e. maintaining a good approximation of the spiritually-underlying continuous uniform distribution — whilst retaining a nonzero probability of selecting any given floating-point number within the desired interval. Again, ✨⁠spiritually⁠✨, we’d think of it like so: we sample from the desired continuous uniform distribution, & then return the floating-point number that is closest to the real number that we sampled (tie-breaking is irrelevant because we almost never have to do it). Obviously, doing it in exactly this way is impossible, but it intuitively seems to me that an implementation that does effectively the same thing — but in a practical manner — should be possible.

   For example, if the desired interval is $\left[0,1\right)$, then an individual floating-point number that’s closer to 0 is less likely to be returned than an individual number that’s closer to 1, because the numbers closer to 0 are denser as a result of the floating-point representation. If we plotted all floating-point numbers that could possibly be returned from such a function on a bar chart, spacing them out equally on the x-axis (abscissa), & giving each one a bar whose height corresponds to that number’s probability of being returned from the function, then we’d get something that basically looks (if you squint just a little) like an exponential curve.

   Does that make sense? It would be uniform in numeric value, but you’d have kind of a weird effect where the precision of the result depends a lot on how far the result just-so-happened to be away from zero — a weird effect that the usual Math.random way doesn’t produce. But maybe there are places where it makes more sense — after all, it’s actually using the full dynamic range of the floating-point representation, which is what floating-point numbers are for! If anyone knows of a function out there that does this, or something like this, please do let me know!

   N.B.: If we literally do things the “floating-point number that is closest to the real number that we sampled” way, then there’s no difference between closed & open intervals; e.g. $\left[0,1\right]$, $\left(0,1\right]$, $\left[0,1\right)$, & $\left(0,1\right)$ all give the same results.

5. [↑] See also, although “unit interval” more usually refers specifically to $\left[0,1\right]$.

6. [↑] This is not strictly true in all cases, because a CS cannot lower a stat to a negative value — negative stat bonuses from equipment items do not exist in MapleStory in general. However, in this case, we’re starting with WAtk values that are so far away from zero that it turns out to not matter.

7. [↑] Using $𝒰\left\{𝑛,𝑚\right\}$ to denote the discrete uniform distribution over the integers $𝑘$ such that $𝑛\le 𝑘\le 𝑚$.

No physician can cure this ailment

Hi, hello, welcome. Get comfy, because this section is going to be about a less-than-comfy topic…!

Footnotes for “No physician can cure this ailment”

1. [↑] This fact is perhaps most clearly evinced by iron(wo)man styles of play, wherein a PC swears to never trade nor party with any other PCs. The reality of iron(wo)man — and, more generally, “metallic” styles of gameplay — is somewhat more complicated, but the point is that iron is very much intended to close off all meaningful game-mechanical interactions with other PCs. It’s also worth noting, however, that truly single-player MapleStory also very much exists (and has existed); typically, all it takes is to run the game server & the game client on the same machine, for example. Obviously, the “on the same machine” part is not a hard requirement by any means, but is one possible manifestation of this phenomenon.
2. [↑] I’m tempted to say “assumed to be heterosexual”, but my memory is too fuzzy. It’s not unlikely that something that David said in the initial smega strongly implied that he was hetero.
3. [↑] Please, help me out here.
4. [↑] Much worse, unfortunately. 🫤

Wiktowia Aiwand…

Wewwcowm bauwuck two Wiktowia Aiwand…

Ahem. Sorry. Welcome back to Victoria Island, everyone’s favourite island. Probably. Maple Island is a close runner-up, but it doesn’t have sadistic jump quests nor backstabbing traitors, so it can’t quite snag that top spot.

On my vicloc dagger spearwoman d34r, I teamed up with fellow viclockers bak2monke the assassin and xXCrookXx the bandit, to go hunting for some Jrogs…

Luckily for us, they were actually there! All nice & spawned, in The Cursed Sanctuary! You’ll notice the unexpected presence of a Taurospear in the image above; MapleLegends somewhat recently changed this map (and The Cave of Pianus, in a similar way & for similar reasons — but with Pinbooms) so that it has two Taurospear spawns, mostly to allow PCs to claim mapownership^[1] even when Jrog has yet to spawn. For our purposes, though, they were actually kinda annoying for our ranged friend bak2monke, who was often stuck attacking Jrog’s Taurospear meatshields instead of Jrog itself.

My presence was particularly important here, as Jrog is not fucking around with the magic attacks — with 605 MATK, it’s capable of hitting up to 2.9k damage!! Probably more like 2.8k after MDEF, but still. HB was very necessary. And even with HB, it’s not easy staying alive when you can easily get murderificated^[2] in just two attacks…:

R.I.P. bak2monke… 🪦

It’s okay though, bak2monke got back on his feet, and we got ’em:

Yes, all of ’em. All eight. Do not fuck with Victoria; we will steal your area bosses take back the area bosses that are rightfully ours…!

Also on d34r, I had an epic battle with our good frenemy Grog, alongside osslock ranger Mekhane (Sunken), & outland marauder SignMyPirate!:

And finally, after some more Physical Fitness Challenges courtesy of the MapleLegends GMs, it was time to do another Scroll of Secrets (SoS) run, on my vicloc clericlet d33r:

Alright well, the top WDEF 60% goes right on the ground, where it belongs. But the helm INT 30%… very nice!! I would use it, but I don’t have any hats worth booming, except for the one that I’m wearing… and I ain’t booming that shit. Yeah, I wear a Rudolph’s Toy Hat (6⧸6⧸8⧸6 stats!) of slightly questionable legality, but I can’t even wear normal mage hats, goddammit! I’m too stupid!! Gimme a break!!!

Footnotes for “Wiktowia Aiwand…”

1. [↑] “Mapowner” is a technical term used by MapleLegends in the context of their built-in map ownership system. If a PC is sufficiently high level, killing a monster grants them ownership of the map that they’re in (i.e. grants them mapowner), for five minutes. The only exception is if another PC already has mapowner, in which case mapowner is “sticky” and is retained by that other PC. Killing a monster when one already has mapowner simply refreshes the five-minute duration, so it’s possible to hold ownership indefinitely if one continues to kill stuff. The basic idea is to give some kind of system for negotiating the use of maps between potentially competing players, with the owner of the map wielding the right to exclude other players from using the map — should they choose to exercise said right. Mapowner is often impossible to enforce in the context of relatively weak area bosses; I’ve certainly had issues with people swooping in and stealing kills on maps that I own, and leaving before I can so much as copy their IGN. But, you know. In principle, you’re not supposed to do that.
2. [↑] A very real word. Trust me, I am can English goodly.

Seduce me if you dare

When it comes to bosses in MapleLegends, the ones that I haven’t really been able to experience — yet! — happen to be those that I would classify as what I like to call “ultra-difficult” or “ultra-hard” content: Pink Bean and Aufheben. I’ve had a little trial run of Aufheben, but that’s it. And certainly no Pink Beaning whatsoever. That being said, excluding these two bosses leaves… well, most of the bosses in the game. And for each of those bosses, I generally have a pretty crisp idea of how much I enjoy fighting it. Actually, let’s go ahead and make one of them fancy “tier lists” that are hip with the kids these days. And heck, maybe I can even update it in the future, as my opinions change…:

1. S tier: Horntail, Zakum, Jiaoceng^[1], Core Blaze.
2. A tier: Bergamot, Ergoth.
3. B tier: Capt. Latanica, Nameless Magic Monster, Nibelung, Scarga^[2].
4. C tier: Dunas 1, Headless Horseman, Bigfoot, Pianus^[3], Black Crow, Kacchuu Musha^[4], Ravana, Papulatus, Scarlion, Targa, CWKPQ^[5], Leviathan, Manon, Griffey.
5. F tier: Dunas 2, The Boss^[6], Female Boss^[7].

Why is ⟨S⟩ at the top? In what alphabet is the letter ⟨S⟩ the first letter? I don’t know, that’s just how tier lists work. I don’t make the rules. I’m only including level ≥100 bosses. I omitted Castellan (Toad) because although I have done an entire Toad run before, it was only one time, and I was suffering too much from slideshow-level framerates — amongst other things — at the time to give it a fair shake. I also omitted Spirit of Rock (SoR) because it never spawns. I also omitted all ToT bosses, because I don’t really clearly remember fighting them (Lyka does have seduce, which is fun), and I’m never going out of my way to fight them — they’re basically quest-only, as far as I care. Ergoth is in the A tier mostly because I’m not just including the “proper” boss fights here, but rather, whole-ass runs (boss fights included, of course).

Anyways, some of these bosses are really different from each other, so it’s very situational as to which one I might “enjoy more”, or “would rather fight”, or what have you. With that being said, maybe there’s a good reason for randomly using a letter that isn’t ⟨A⟩ to represent the highest tier: the bosses in the highest tier really are blessed in a way that the other bosses are not. I am always tryna do S-tier bosses, if I’m tryna do any bosses at all. Likewise, the F tier is there — skipping ⟨D⟩ and ⟨E⟩ — to contain the truly cursed bosses; the ⟨F⟩ stands for “fuck these bosses”…

In any case, the point of talking about boss tiers is to lead to a particular S-tier boss that is the subject of this section: Horntail! Everybody’s favourite three-headed lizard!! Perhaps familiar to the Ukrainian reader as Змій Горинич ⟨Zmij Horynyč⟩ /zʲmij ɦɔrɪnɪt͡ʃ/…

I did some HT on my darksterity knight rusa as an attacker, including one with a level 188 corsair by the name of Danger:

Yeah! So, it’s somewhat unusual to find PCs in MapleLegends that are such an extremely high level, but nevertheless with quite low MAXHP that can be achieved with little or no HP washing. That is, of course, because of the prevalence of HP washing in MapleLegends, especially for the purpose of fighting bosses that cannot be survived (or cannot practically be survived) without the extra MAXHP available through washing. And so my utmost respect goes to Danger (a fitting name, come to think of it…) for being sufficiently determined, brave, and certainly skilled, to play the game this way! Plus, you know, it means a lot to me^[8] to feel like my HB is truly useful, and to feel like I really am protecting my party when I HB, IW, & Power Crash.

And… we made it!!

First clear! :O

On another day, I was unexpectedly recruited at the last minute into a Funk guild HT run that was intended to be something like a 20-Mapler run. Standardly (at least in MapleLegends), HT runs come in basically three flavours: 6-Mapler, 10-Mapler, and 12-Mapler. Having 12 runners is kinda the default, because it’s exactly enough to fill two parties, which is generally enough to clear the run successfully — and if you’re particularly adept, quickly as well. Having 10 runners is basically similar, but with five runners per party instead of two full parties; it may be slower and/or require more firepower, but in exchange, the loot is split amongst fewer people, so each individual benefits more. Plus, it’s easier to recruit & organise 10 people than it is 12. The logical conclusion of reducing the size of your team, then, is simply to reduce it to one party! Thus, the 6-Mapler run is a natural choice for particularly strong parties who are willing to put in the extra time, and perhaps the extra Onyx Apples as well. By comparison, then, a 20-Mapler HT run is positively gargantuan: all Funkers are invited! Well, I’m not in Funk, but I can certainly pretend.

We found out very early on in the first run that F/P archmage Baboom (Babeem) was running the game on a vaguely recent MacBook™ that was… struggling. Because the MacBook would naturally be running macOS™ as its operating system, running a Windows™ program (and quite an old one, at that) meant running it in Wine, or something like that.^[9] And uhh… something wasn’t translating well. As soon as we summoned the first prehead and started attacking (all 18 of us, or whatever…), Baboom was frozen solid. At first, we assumed that she had completely disconnected from the game due to a client crash, because she was totally frozen in place and not taking damage from anything — not even from the falling rocks that she was standing under. It was confirmed in voice chat that her screen was totally frozen on her end. The whole “not taking damage” thing is a common artefact of client crashes, along with the similar artefact of never leaving “alert” stance^[10]; because damage is calculated by the client and sent to the server, a PC associated with a crashed client naturally cannot take damage — the client can’t calculate damage and send packets, because it is deceased!

Nevertheless, we tried reviving her client by stopping all attacks — you know, just in case she hasn’t actually crashed yet, and is just dropping an unlimitedly large number of frames:

Bizarrely enough, it actually… kinda worked? She definitely had not crashed, and was able to briefly recover for a second or two — at least long enough to correctly take damage from the falling rocks, and get pushed to the right a bit. But also, it kinda didn’t work. She froze again, and we gave up trying to revive her and focused once again on actually killing the damn prehead. Then, in less than a fraction of a second, this happened:

It’s a little hard to see in the image, but you can see her Funk guild tag and her pet Jam?’s nametag at the bottom-right of the map. What you can’t see is that, well, she’s dead. She kinda just… teleported from the top-left of the map to the bottom-right, and a gravestone just fell on her from the sky. What a way to die.

But, again, she had not crashed! We were able to revive her once the first prehead had died, and she had no problem getting through the portal into the second prehead map. Bizarrely enough, she was briefly able to attack the second prehead normally before, as we naturally suspected would happen, she froze again. And died:

That’s really the fun of a 20-Mapler run: there’s so many damn people, that not only can you not see yourself, but there’s virtually always at least one person doing something goofy. Here, we can see bishop Pengebingen (Palladino) repeatedly firing Angel Ray into… well, the wall. The opposite way of where Horntail is…:

Speaking of goofy, we may or may not have triggered mass seduce early:

Oh, my. Both the mid and right heads are still up……

In the end, though, we made it. We might have seen some wack-ass shit, died a lot of times, failed to actually shoot in Horntail’s direction, and griefed our own run by continually cleaving wings + left arm even when heads weren’t weapon-cancelled despite not even being sed target^[11], but that lizard is dead. A good day for Funk’s guild coffers.

But that’s not all — there’s more Horntailing where that came from. And we’re making it spicy: I’m a sed target now! I’ve talked a lot about running HT as sed target on rusa in previous diary entries, so this is nothing particularly new. Nevertheless, things are a little different this time around, as we’ll later see. To start things off, I ran as sed target again for Diggy’s runs:

I’ve grown accustomed to wearing my Valentine’s day outfit when running as sed target, as the giant goofy bright red heart-shaped hat makes me easy to spot, both for my own sake and for the sake of my bishop.

I’ve mentioned before that the right side of the map can be dangerous for me. In particular, I’ve previously died as a result of not being able to Hero’s Will before a 1⧸1 whilst seduced on the top-right rope, then getting hit by that 1⧸1, and then quickly being slain by a wing slap. Getting 1⧸1’d frequently does not kill me, as a result of damage mitigation from skills like Achilles and Elemental Resistance; however, taking touch damage from almost any of HT’s body parts is basically a no-go in this situation, as I’m working with precious little HP, and Elemental Resistance will not help me. So, a wing slap is particularly brutal, as it causes me to take touch damage despite not actually bumping into HT myself — I am, after all, perfectly still, hanging motionless & seduced on my rope. What I’ve found, then, is that although roping is very instinctively appealing when you’re about to get seduced, and there is a rope right there immediately next to where I’m attacking from, I never actually use that rope when seduced (except possibly when arm(s) is/are the only body part(s) alive, in which case there’s no actual danger anyways).

So, if I’m attacking from the right side, and I’m about to get seduced, what do I do, then? Well, it often looks something like this:

I can use the bottom rope, instead. It seems simple, but after a lot of practice and some thought, it’s more difficult to consistently execute well than you expect. The basic idea is simple: if I do this, then I am virtually guaranteed not to take any physical damage from HT whilst seduced — the wings definitely can’t hit me from here, and both legs & tail are dead. And certainly, this strategy is clearly superior to grabbing the top-right rope, as I can say from experience, and as you might be able to figure out analytically if you think about it really hard. However, the phrase “if you think about it really hard” in the previous sentence is doing a lot of heavy lifting here; frequently, the harder that you think about it, the less obvious it seems to identify “the best strategy”.

Being in the position that you see above has clear advantages. I already mentioned that it avoids the issue of taking physical damage after getting 1⧸1’d, but of course, what I’m most worried about is getting 1⧸1’d in the first place. If I’m grabbing the top of this rope, I’m generally high enough above the floor of the map to not be in striking distance of the 1⧸1s on the floor. 1⧸1s also don’t target the platform that the rope is connected to, so I’m all good on that front. But again, “I’m generally high enough above the floor of the map” has its caveats. Consider the lightning-bolt-adorned 1⧸1 markers at the bottom of the above image; that’s what it looks like when the right head uses its 1⧸1. Also notice the position that the right arm is in: it is outstretched, angled ever-so-slightly downwards, has a clenched fist, and has an ornate red-tinted orb thingy around it. This is its mana drain animation, and you can see the red stabby thing above my head as it hits me (there are actually three stabby things, but I took this screenshot too late to clearly see them). Also, it does the exactly 1 damage that you see above. A mana drain lowers your MP by 3k, so it’s guaranteed to lower my MP to exactly zero (0), due to my MAXMP being below 3k. Having zero MP means zero ability to use Hero’s Will, so I can no longer Will in anticipation of being 1⧸1’d. Although I’m obviously fine in the above image, as none of the 1⧸1 markers are close enough to hit me, the point is that the lightning 1⧸1s have unlimited vertical range; they stretch from the very bottom to the very top of the map. This makes the right head particularly terrifying, as it’s the most likely to actually threaten me when it decides to cast its 1⧸1. Furthermore, the other 1⧸1s can have deceptive vertical ranges. The left head casts ice 1⧸1s, the mid head casts fire 1⧸1s, and neither have the unlimited vertical range of the right head’s lightning 1⧸1s. Nevertheless, I’ve recently found that fire 1⧸1s reach significantly higher than I previously thought, and considerably higher than the actual animation visually suggests.

On the other hand, trying to position myself high up so as to be out of range of non-lightning 1⧸1s can be somewhat problematic. In particular, in the hopefully rare case that I actually have to eat 1⧸1, my bishop is going to find it more difficult to assist me (read: Heal my sorry ass) as my elevation increases. Bishops are almost always on the floor, and climbing upwards is neither trivial nor particularly instantaneous. Nevertheless, it’s worth noting that a single Heal is virtually always more than enough to fix up a 1⧸1 that I just tanked, so hitting me with a jump-Heal is perfectly acceptable, if perhaps not absolutely & totally ideal. Furthermore, it can be surprisingly difficult to position at the very top of the rope without missing it entirely, especially with the precious little time that I have between the seduce animation starting and the seduce itself starting. This is worsened by the fact that my JUMP and SPEED values can be a bit all over the place for various reasons (did the thief in my party miss me with Haste? Are they consistently not hasting me, except for this one time?? Did my Victoria’s Amorian Basket run out??? Did someone’s Haste override the SPEED portion of my basket and then expire before the basket did???? Did I use a Speed Pill?????). And judging whether or not a 1⧸1 is close enough to hit you is harder than you might think, in marginal cases…

You know, come to think of it, “marginal cases” isn’t really doing it justice. Even beyond the difficulty of judging whether or not a 1⧸1 can hit me when I’m roped, there are plenty of cases where I’m seduced on the floor, and if that seduce is a left seduce or a right seduce, it can be difficult to tell if a 1⧸1 can hit me. If I’m running into the wall (e.g. I’m near the right wall and get right sed), the edges of the map are invisible, thanks to the cave walls obscuring all vision. I’m usually fine, due to the falling rocks giving me iframes, but I think(??) that it’s still technically possible to get 1⧸1’d there — which is problematic, as there’s no way to see the marker. In the case where I get seduced to the opposite side of the map, I am just cruising along… walkin’ it up. The 1⧸1 animations are a little obscure when it comes to exactly which frame of the animation the 1⧸1 effect occurs on, and if I’m walking past it, it can be difficult to time where exactly I’m going to be when the 1⧸1 actually occurs. Adding iframes into the mix makes things more difficult, as I sometimes think I’m about to be 1⧸1’d but I get iframes in time for it to whiff, or I think my iframes are going to last long enough to avoid the 1⧸1… and then they don’t. The good news with the particular case of being seduced to the opposite side of the map is that it’s quite visible; it’s very likely that my bishop sees me zip past them and unwillingly cross to the other side. Nevertheless, I am always ready to Will if I think that I’m about to tank a 1⧸1.

Consider three possible manoeuvres that I might attempt when attacking from the right side and seeing that I’m about to get seduced:

• Attempt to hold the top of the bottom-right rope.

  • 👍 If my Will isn’t on cooldown, being roped immobilises me and makes the damage (particularly 1⧸1s) that I might take more predictable, enhancing my ability to choose between Willing to save myself, and conserving my Will.

  • 👍 Assuming that I don’t get knocked off the rope (likely, given my avoidability & Power Stance), the only sources of damage are magic attacks from the heads — or, in rarer cases, attacks (touch or otherwise) from Wyverns. These attacks are relatively sparse, and if they aren’t 1⧸1s, they don’t do much to me. In particular, I’m avoiding taking touch damage from HT’s body parts (ouch!), and avoiding taking damage from falling rocks (not very painful individually, but very consistent).

  • 💩 If push comes to shove and I tank a 1⧸1 (my Will is on cooldown, I got mana-drained, whatever…), my bishop is less likely to both (a.) correctly identify that I might need help, and (b.) actually land a Heal on me, as a result. I say “less likely” because this is (usually) still better than being stuck in an even more elevated position (jump sed on the platform, roped on the top-right rope, etc.). Because I’m roped, I’m not very visible (read: not moving), and I just look like I’m absolutely chillaxing and having a great time enjoying a faceful of rope. When, in reality, I might actually be enjoying a faceful of 1⧸1. Plus, even if my bishop does try to Heal me, they have to make sure that they jump-Heal.

  • 💩 This is probably the most difficult manoeuvre to pull off, requiring a lot of precision to not (a.) miss the rope, nor (b.) go too far down the rope. Not to mention how fast you have to react to the seduce animation…

    Failing in the former way “(a.)” can be problematic particularly if it’s jump sed; usually, this means that I’m jump seduced on top of the platform that the bottom-right rope is fastened to. This is probably the worst case (other than grabbing top-right rope or being way up top) for my bishop if they want to Heal me, as a jump-Heal is unlikely to do the trick. And although I can’t get 1⧸1’d by the other heads, I can — of course — still get 1⧸1’d by the right head. That being said, it’s still not the end of the world, as I should have little issue knowing whether or not I should Will, and the wings cannot slap me from here.

    Failing in the latter way “(b.)” can be problematic particularly if the mid head’s 1⧸1 hits me when it otherwise wouldn’t have. This 1⧸1 has limited vertical range, of course, but the vertical range is still considerable.

• Run to the right, dropping down onto the platform to which the bottom-right rope is fastened.
  • 👍 There’s virtually no way for me to take touch damage from HT.
  • 👍 Because I’m not roped, my movement makes me more visible.
  • 👍 Unless it’s jump sed, my bishop should see me fall to the floor and have an easy time Healing me if necessary.
  • 👍 This manoeuvre is relatively foolproof; that is, easy to pull off.
  • 🤷🏽‍♀️ If I perform this manoeuvre, it may or may not be on purpose — it’s possible that I tried the manoeuvre above and simply failed to grab the rope in time.
  • 💩 If it’s jump sed, I run into the problem mentioned above, where I’m jumping from too high of an elevation to reliably get help from my bishop.
  • 💩 Unless it’s jump sed, it’s unlikely that I will evade a hypothetical mid head 1⧸1 by dint of being far above the floor, because I probably am on the floor — unless the 1⧸1 happened quite early on.
  • 💩 Unless it’s jump sed, there’s a distinct chance that I will be on the floor (walking across the map, or tryna hug the right wall) and wondering if that 1⧸1 is actually gonna make contact with me or not.
• Immediately downjump to the smol platform below, and then evacuate said platform in an appropriate way (downjumping again, walking left, or walking right).^[12]
  • 👍 There’s virtually no way for me to take touch damage from HT, provided that I pull the manoeuvre off successfully.
  • 👍 Because I’m not roped, my movement makes me more visible.
  • 👍 My bishop should see me fall to the floor and have an easy time Healing me if necessary.
  • 👍 Provided that I pull the manoeuvre off as intended, it’s very unlikely that a jump sed is any more dangerous than a left sed.
  • 🤷🏽‍♀️ This manoeuvre is only moderately difficult to pull off; the only real risk is that I get seduced just after downjumping the first time, meaning that I failed to evacuate the bottom platform in time.
  • 💩 If it’s jump sed, then depending on how things work out and on how I decide to fall to the floor, there’s a possibility of ending up jump sedded underneath the right arm. Depending on my JUMP value, this can be surprisingly perilous if I jump high enough to take touch damage from the arm. Note that it’s possible to mitigate this in some cases by strategically right-clicking (cancelling) Power Stance, but it’s not easy to pull off.
  • 💩 Although it’s unlikely, failure to evacuate the platform — in combination with jump sed — can leave me jumping on the platform. This puts me at mild risk of getting wing-slapped (although if I get wing-slapped without first being 1⧸1’d, it’s unlikely to be particularly dangerous), and mild risk of my bishop trying but failing to Heal me.
  • 💩 Unless it’s jump sed, it’s unlikely that I will evade a hypothetical mid head 1⧸1 by dint of being far above the floor, because I probably am on the floor — unless the 1⧸1 happened quite early on.
  • 💩 Unless it’s jump sed, there’s a distinct chance that I will be on the floor (walking across the map, or tryna hug the right wall) and wondering if that 1⧸1 is actually gonna make contact with me or not.

Now, all this is uhm… quite a lot to unpack. Don’t worry, though; it gets worse! The three manoeuvres listed above are obviously not exhaustive. Furthermore, consider situational factors… Which sed (left, right, or jump) is this? (You may have noticed the significance of jump sed!) Do I already see 1⧸1 markers as I’m getting seduced? Even furthermore, all of this is assuming that we’re starting from the typical attacking position that lets me cleave wings, right arm, & right head at the same time. Oh, my lord.

Am I thinking about it too hard? Maybe. But also, maybe not. I don’t really die, so that must mean that when I do die, it’s probably a particularly nasty storm of multiple conspiring, confounding, & coincidental factors — it can get pretty hairy out there, particularly because HT has so many body parts that operate independently of one other. Ideally, I have a technique and instinct that is refined enough to keep me alive & doing my job, even through a tumultuous & eventful hour or so of continuous fighting! Without reducing my brain to a fine mush!!

Speaking of conspiring, confounding, & coincidental, have you ever seen HT perfect-frame your bishops? I have!:

I was paying good attention, so it felt like I was watching it in slow motion; I directly witnessed the exact point at which both of our bishops were alive and going about business as usual, as it transitioned over the course of one (1) frame (just shy of 17 milliseconds) to those same two bishops being deader than a pair of doornails. If you’re not familiar with the term “[to] perfect-frame”, the basic idea is simple: because the dispelling ability that is cast by the mid head (and by the left arm, but the left arm is typically not the culprit) is not an attack, it does not give afflicted PCs iframes. Thus, it’s technically possible to get dispelled by the mid head (which is a map-wide dispel that has a ≈90% chance of hitting each PC in the map) and take damage within the same frame — or perhaps separated by just one frame. This is problematic particularly for mage types like bishops, because they may have insufficient MAXHP to survive some attacks, unless buffed by Magic Guard (MG). Thus, if the dispel and the powerful attack hit at the same time, the bishop cannot possibly react in time to re-cast their dispelled MG before getting hit by the attack. The resulting effect is frightening: one frame, your bishop is chilling at 100% HP and operating as usual, and the very next frame, they are dead!!

You’ll notice that I carefully stated that mages “may have insufficient MAXHP to survive some attacks, unless buffed by [MG]”. For one, not all attacks are equally damaging, so even a mage with modest MAXHP will be able to survive some attacks sans MG, assuming that they are sufficiently close to 100% HP before being struck. Furthermore, HP washing (and, to a lesser extent, equipment that grants MAXHP^[13]) enables mages to artificially bolster their MAXHP so that they may survive a wide range of attacks sans MG — again, assuming that their HP is sufficiently close to 100% upon being struck.

There’s a significance to the fact that we were on the right side when this happened. You’ll notice that all three heads are still alive; that is, we started on the right side immediately after dispatching the legs & tail. My understanding is that the choice of which side to start on (left vs. right) is essentially left up to the bishop(s); yes, the host of the run is ultimately the shot-caller, but the host’s understanding of which side is best to start on is based on the bishops — particularly, the bishops’ MAXHP values. I’ve only ever played rusa in HT, and I’ve never hosted HT before, so I don’t know what the significance is of starting on the left vs. starting on the right. But the basic vibe that I got in this case is that starting on the right was at least slightly preferable to the left, and it was just a matter of asking the bishops if they’re okay with the increased probability of getting perfect-framed. I don’t know what it is about the right side that makes perfect-framing more likely (or possible in the first place?), but perhaps it has something to do with one of the right head’s ranged attacks being elementally untyped (no particular elemental affinity, even though you would expect the right head’s attacks to be lightning), or something like that. Of course, ultimately you have to go to the right to kill the right head anyways, but as the fight goes on, the bishops might have more opportunity to anticipate dispels by purposely running/Teleporting/jumping into Wyverns, or whatever. I don’t really know.

So, considering that both bishops died simultaneously to a perfect-framing, one wonders why we started on the right side. Then again, perfect frames are rare, so it’s hard to really anticipate one — I think that this was the first time that I ever witnessed one with my own two eyes. Anywho, I guess I just have to go ⅔ of the main body fight without any bishops… I mean, it sucks for everyone (not my HS!!), but it sucks a little extra for me, considering that I’m sed target…

Right, so, we have no bishops, which means no one Genesising/Big Banging the Wyverns to death. Because I’m a melee job whose attacks naturally cleave, I’m okay with having to deal with Wyverns flying willy-nilly all over the place. For our ranged expedition members, on the other hand, it’s a bit unfortunate. But they’re all the way up there hitting heads, anyways. Me? I’m down here getting my entire body continuously and lightly pummelled by these god damn dragonbirds.

Unfortunately for all of us, however, things were about to get worse. Our bishops were not the only ones sent to an early grave; our bowmaster bit the dust:

So did the other DK:

Having started with 11 members, we were now down to a comical 7 — a massive reduction of our forces, but nevertheless just shy of allowing us to merge into a single party. In the end, though, we — eventually!! — killed the lizard:

Gee golly gosh, Horntail! Thanks for the 🔥🗑️🔥hot fucking garbage🔥🗑️🔥!!! AAAAAAAAAAAAAAAAAAA

Oh, well. At least we got a pouch. Oh, and I actually survived??

The Wyverns were certainly annoying, particularly the Dark Wyverns. Because they can dispel, I frequently found myself scrambling to recast at least some my buffs so that I would be able to kill the birds, and be more likely to survive in the event of a seduce. Unfortunately, this saps the time that I actually have to attack them, so some of my party members had to help me two or three times… In any case, annoying as they may be, I partially attribute my survival to the damned things. If I’m constantly taking touch damage from Wyverns, HT’s got nothin’ on me!! This run was a real test of my abilities as sed target, so I’m proud to say that I survived! [':

On another run as sed target, I hit level 179 (…and corsair Grrrk (Gurk, Gork, Gyrk, Girk) hit 162)!!:

Just one more level for the big 180! 😌

As a result of a slight quirk in recruitment, I ended up doing one run with not one, but two sed bishops: tomtomgun & AINTNOWAY!

In the above image, you can see just the tops of the heads of our three bishops. From left to right: AINTNOWAY (of whom you can mostly just see his HP bar), tomtomgun, & Wook123 (our regular bishop). You can also see a demonstration of what happens if I get jump-seduced on top of the right side’s large bottom platform without roping.

This was also a neat run because we had a HTP buyer: F/P archmage pyrOs, perhaps better known as MitoAdmita!

On another run, nightlord AYEMWHY accidentally entered the second prehead map before I did, so she inadvertently got a little taste of being sed target…:

As sed target, some things are much scarier than others. For example, being a warrior sed target, the tail — and the legs — are particularly scary because I don’t have Dark Sight (nor Oak Barrel, for that matter). Thus, if I get right sed and I’m not securely roped, I risk running my entire-ass body into the legs & tail! Ow!! This kills the rusa!!! For reference, the tail touches an average character for probably around ≈25k damage when unempowered, and considerably more when empowered. With the legs, it’s probably more like ≈11k unempowered. Now, I have 30k MAXHP with HB, and I have Achilles on my side to reduce all sources of damage, but I’m pretty sure that still likely kills me; chances are, I’m going to be taking more than just one touch attack before my seduce’s duration runs out. I’ve never actually had this issue, but you know, I prefer to not learn about this particular scenario from experience…

Furthermore, the legs make things considerably scarier than the tail already is. Not only is it another source of touch attacks should I get right sed, but it has pretty nasty ranged attacks that can hit a typical character for around ≈9.5k a pop, and it can stun. The damage from its ranged attacks (i.e. its stomps) is mostly an issue if I’m left sed and am not securely roped; you have to be standing on a foothold for it to hit you, so being roped or jump-sed is frequently (but not necessarily) enough to avoid it. And the stun… well… I think I’ve mentioned this before, but getting stunned immediately before a seduce animation is particularly bad, as it implies that I cannot move onto the rope in anticipation of the seduce itself. Nevertheless, if my Will is not on cooldown, I like to at least attack the wings whilst legs and/or tail are still alive:

You can see that I’m on the far right edge of the larger platform, rather than standing on the smoller platform that is a bit lower and closer to the centre of the map. Because my Will isn’t on cooldown, I can use it to save myself if I get in trouble (e.g. I don’t manage to rope in time for a sed). Nevertheless, I prefer not to make this excessively risky and greatly increase the chance that I have to pop my Will (which has a six-minute cooldown, mind you), so this particular position serves as a kind of middle ground: I’m able to hit wings 100% of the time that I’m attacking, I hit left head any time that it cranes down to attack, and I hit left arm any time that it stretches far enough to the left and/or I am Shielded and can attack “normally” from the smol platform — and furthermore, there is now almost no chance (other than a particularly unfortunately-timed stun, assuming legs are still alive) that I don’t manage to rope before getting seduced. Once legs & tail are both destroyed, I can start attacking normally.

Speaking of attacking normally, there’s naturally a problem of what body part(s) I’m attacking. I’m often cleaving wings & arm; this might sound like an invitation for early mass seduce, but it generally works out quite well. I keep an eye on how much damage these body parts are taking, and I know how to identify when an arm is crossing an HP boundary (e.g. seeing when left arm crosses into dispel range). Combined with information that I can get at a glance — particularly, which body parts are alive and how much total HP Horntail has remaining — I know when to take it easy and focus heads if possible.

Speaking of focusing heads, attacking heads is naturally difficult for me. Being so far away from my bishop is unappealing, and furthermore, I have to be careful not to get seduced into a head and take touch damage that way (remember — I don’t have Dark Sight!). That being said, if my Will is not on cooldown and I’m willing to be extra careful about watching the arms really intensely in case of a seduce animation, I do somewhat frequently attack heads (especially later into the run, for obvious reasons):

On the left side, it’s really not so bad. Getting into position to cleave left & mid heads simultaneously is relatively easy, and I probably don’t have to worry about right head 1⧸1s. On the right side, if I’m honest, I usually avoid it. Not even because of right head 1⧸1s, but because getting into position to cleave right & mid heads simultaneously is so painfully difficult and takes so long that it feels like a stressful waste of time. The positioning is pixel-perfect, somewhat dangerous to manoeuvre into, the whole time that I’m trying to get into position is time that I can’t really watch for animations very well (as I’m scrutinising my own position at a subpixel level…), and by the time that I actually do get into position, some random-ass off-cooldown seduce gets cast by right arm, or someone crosses and triggers left arm sed… Yucky.

I’ve grown accustomed to using Victoria’s Amorian Baskets when running as sed target during the main body fight. I get these damn things from APQing anyways, so I may as well eat ’em, right? Once HT’s body parts start to become sparse (e.g. right head is dead), I generally just fall back to Archer Elixirs, and maybe a sneaky Speed Pill here or there. The great thing about these baskets is of course the +40 AVOID, but I will say that the +40 SPEED really doesn’t hurt either. If you’ve run HT before, you may or may not have been given advice like “right-click [i.e. cancel] your Haste” in cases where one or both arms are casting mass seduce. The idea here is to lower your own SPEED value, causing you to take longer to arrive at the far left (in the case of mass left sed) or the far right (in the case of mass right sed), either of which threaten you with falling rocks. The falling rocks don’t do all that much damage, but it’s certainly more damage than the arms themselves can deal (barring touch attacks). For me, though, I really just want my SPEED to be maxed at all times; this enhances my ability to quickly & easily manoeuvre myself as necessary, particularly in anticipation of seduction. It’s convenient that the basket gives loads of SPEED but no JUMP at all, as I also generally want my JUMP to be on the lower end of things. This is because, as mentioned previously, jump sed is of special significance in being somewhat extra tricky & dangerous, e.g. the unfortunate position of being jump-sed directly beneath an arm. A low enough JUMP value prevents actually taking touch damage in such a case.

Overall, I think that one of the greatest strengths of playing rusa in the sed target role is simple: the greatest threat to any non-shadower sed target — viz. 1⧸1s — is relatively(!) easy to survive, for me, assuming that my technique is decently refined. Even if I have to accept a 1⧸1 connecting with me whilst I’m seduced (which is just inevitable sometimes; the cooldown of Hero’s Will is just too lengthy), I’m probably fine. For starters, my avoidability is good enough that I have a decent chance of not being 1⧸1’d at all; it may just “MISS” me. Even if it doesn’t miss, Achilles mitigates the impact, leaving me with a small but usable amount of HP. Being a DK, Elemental Resistance takes the pain out of almost all magic attacks, especially in combination with Achilles. Furthermore, my avoidability is still in effect, so any attack (other than the falling rock map hazards on the far left & far right) has a solid chance of missing me altogether. Also, being a DK, I have small but significant self-healing that heals me even whilst I’m seduced.

This brings me to the topic of sed bishops. The screenshots above document just a few moments of a few of the Horntail runs that I’ve done recently as sed target. In fact, I’ve run so many such runs now, that I have experience with a wide variety of sed bishops, and I can safely say that this wide variety comes with widely varying experiences. In MapleLegends at least, two-party HT runs (e.g. 12-Mapler or 10-Mapler runs) conventionally refer to the role filled by the bishop who is in the same party as the sed target as “sed bishop”; conversely, the role filled by the bishop in the other party (which does not have the sed target) is referred to as “regular” (or “reg”, for short) bishop. This nomenclature reflects not only the additional responsibilities that are borne by the sed bishop in comparison to their reg counterpart (e.g. using HSh much more strategically), but also the special relationship that exists between the sed bishop and the sed target. When I say “my bishop”, I use the ktetic determiner “my” in the same way that one might say “my advisor”; obviously, the expression is not one of literal ownership/possession, but rather, is indicative of a joint bidirectional relationship. I babysit the group by absorbing all of the seduces, and in turn, my sed bishop is there to babysit me by Shielding me and occasionally Healing me as appropriate. It’s babysitting all the way down.^[14]

Some sed bishops are more consistent with knowing when to Shield (and doing it quickly enough) than others. Some fear for my life even more than I do (an impressive feat), and Heal me to no end. Others have a habit of Healing me two or so times immediately after I get seduced, and then going back to business as usual. The funny part about this is that I almost never need Healing towards the beginning of a seduce; I have multiple HP pots bound to my keyboard and nervously watch for seduce animations, careful to heal myself to full HP (I’m zerking, so this is nontrivial) before the seduction. Then, on the occasions where I do have to tank a 1⧸1 during my state of seducèdness, they are ironically nowhere to be found! Good thing I can survive on my own!! Some sed bishops are more prompt to respond to my cries of “hlep” than others, although I usually try to cry out early if possible, knowing in advance that a 1⧸1 is going to connect with me. It’s not terribly uncommon that I cry “hrelp” and then “jk it missed me” in quick succession. I admit to sometimes not saying anything when I tank a 1⧸1 because I’m fine… I DON’t NEED HELP, OKAY?? I could maybe use a smol Heal, though… If you’re not busy…

As you can see, the sed target–sed bishop relationship is somewhat complicated by the sed bishop also having other responsibilities unrelated to seduction. The more that the sed bishop has to worry about me and do things like Healing, the less time that they have to do other stuff like attacking or buffing. Furthermore, I am absolutely required to have some level of self-sufficiency, even if my bishop is with me the whole time, because chances are that they simply will not always be able to save me. If they die, then they will be inactive for the duration that they are dead (which, in the worst case, is until the end of the run, as we have seen…); if they are in the middle of their Genesis animation, then they are effectively stunned for some time; if they are very far away from me on the map, then they cannot immediately reach me with their spells; etc.

Speaking of Healing me, I once heard someone talking to my sed bishop in voice chat, telling them that they could definitely Heal me any time that I’m not attacking. Naturally, it takes some getting used to a DK who sometimes acts like a normal DK and is allergic to Heal, and other times enjoys being Healed during sed — a lot of bishops probably have an instinct to not Heal at all in the presence of DKs. This is basically correct, so I didn’t say anything… at least, not until someone followed up by suggesting that I’m “probably not zerking” because I’m sed target. Not true!! I most definitely am zerking, virtually the same amount of zerking that I do when running as a normal (non-sed-target) attacker — except, of course, when I’m actually seduced or about to be seduced. Perhaps the simplest way that I can put it is thus:

• I’m attacking: Don’t Heal me.
• I’m not attacking and my HP is ≥50%: You can Heal me, at your discretion.
• I’m not attacking and my HP is <50%, and/or I say “help”: Toss like two Heals my way, please.

As explained above, Horntail is an “S-tier” boss for me, so I always enjoy running it. Furthermore, running as a sed target enhances the experience in many ways: it is really very exciting, and I also get to feel more useful — both to my party, and even to myself, in realising the full potential of the job that I’ve chosen for my character (darksterity knight, in this case). With that in mind, I’m not gonna lie; I did quite a few of these runs, and at some point, perhaps too many. Strange as it may sound, there was a brief moment one night, as I made efforts to put myself to sleep, where I realised that the stress was actually getting to me and affecting my mood & thought patterns. So, you know, I took a bit of a break LMAO. That’s not to detract from how fun it is, but rather, to make clear that it really can be stressful, especially if you are trying absolutely way too hard, like I am… 🙄

The fact that I can even be “trying too hard” in the first place is enabled (but naturally, not strictly caused/required…) by how chaotic & random the game mechanics at work are, and by how thin my attention/concentration can sometimes be spread. In the end, I’m really never quite sure how “good” or “not so good” I am at it, but I certainly do try, and I think that I have a pretty good track record at this point. So, you know. Seduce me, if you dare…

Footnotes for “Seduce me if you dare”

1. [↑] You may know this boss as Wǔlín Yāosēng (武林妖僧; Standard Beijing Mandarin /u˨˩˦lin˧˥ jau̯˥səŋ˥/; “martial arts demon-monk”).

2. [↑] Scarga is, strictly speaking, not a boss in & of itself. It’s, you know, two bosses that have been summoned simultaneously. Nevertheless, because I’m considering entire runs rather than just boss monsters per se, it deserves to be listed independently from its constituents.

3. [↑] Unlike Scarlion/Targa, the Pianuses ((R) & (L)) are area bosses that spawn in the same map, so I have coalesced them into exactly one entry.

4. [↑] You may know this boss (甲(かっ)冑(ちゅう)武(む)者(しゃ); ⟨katchū musha⟩; /katːɕɯː mɯɕa/; “helmet-and-armour warrior”) as simply Samurai, Samu, or Sam. See pts. lxxvii & lxxxi of this diary.

5. [↑] CWKPQ’s boss stage summons five bosses at once: Red Nirg, Margana, Rellik, Hsalf, & Heron (perhaps better known as “Nagrom”). I personally only have experience fighting the “cleave” bosses on the bottom floor of the map, which are Red Nirg, Rellik, & Hsalf.

6. [↑] I prefer to refer to The Boss as “TB”, but it’s more commonly referred to as “BGA” because Bodyguard A is its first “body”, followed by Bodyguard B (BGB), and finally The Boss.

7. [↑] Perhaps better known as Anego (姉(あね)御(ご); /anego/; “female yakuza boss”, literally “older sister”).

8. [↑] And I really mean it, too. More on this subject in the next diary entry…

9. [↑] See “Running MapleStory requires highly advanced technology” from pt. xciii of this diary. Also see MapleStory in the year 2999.

10. [↑] You know, when you attack and/or get hit by something, you go into alert stance — also known as “catching one’s breath”. In alert stance, you can’t drop items, can’t change equipment, can’t change channels, etc.

11. [↑] Yeah, for the record, not me. I’m not going to name any names, but if you’re reading this, you know who you are…

12. [↑] You might worry that walking off of the left side of the platform (manually, or even due to a left seduce) risks touching wing. However, as far as I can tell, this is not actually possible, so the manoeuvre itself is perfectly safe.

13. [↑] The usual suspects are equipment items such as:

    • Monster Book Ring,
    • pet equips that have been scrolled for MAXHP,
    • Speargrave (Wand)/Flairgrave (Wand),
    • eye accessories (e.g. Raccoon Mask) scrolled for MAXHP,
    • face accessories (e.g. Smiling Mask) scrolled for MAXHP,
    • ⋮

14. [↑] See the footnotes for “Wooza” in pt. xciv of this diary.

täˈrän.drʊs

Somewhat regrettably, this is not exactly an episode of Questing With tara™. But this section is starring ya girl tarandus the shield pugilist. As discussed previously, I am on that grind now. Not a lot of quests left to be done…

Whilst grinding Himes with dagger DK daggerknight (Gumby, j0hnny, Jonathan, Kimberly, Mommy), we ran into a dastardly Black Crow trolling the Encounter with the Buddha, searching for potential victims… But with some help from shadower Sanakoh (Yuriel), we were able to slay the beast! And I levelled up!!:

I also headed to CDs with daggerknight to grind some more EXP, where I hit level 112~!:

I previously talked about the holy trinity of pugilist equips for tara, and I’ve already obtained the MoN and the Scar Hat… which just leaves the RBD (not to be confused with RBD). Now, an average clean RBD has 7 STR, 3 DEX, and 2 AVOID, which means that a perfect clean one has 8 STR, 4 DEX, and 3 AVOID. If we ignore the AVOID for now (I do care about it, but just to briefly simplify things), then we have 12 total stats (that is, STR + DEX) perfect clean. I thought that gaining something like ≈20 total stats from scrolling was a good threshold to aim for, as it’s quite good without being extremely difficult (luck-wise) to obtain; for example, passing two 30% scrolls and five 60% (and/or 70%) scrolls — thus leaving three failed slots — is enough to do it. Heck, in a previous diary entry, I made a 26 DEX bathrobe! So even if I do way worse than that, I can still make a 12 + 20 = 32 total stat RBD!

Naturally, I consulted my program scroll_strategist (mirror) to get an idea of how likely I was to succeed on any given attempt. Note that scroll_strategist only cares about getting to ≥32 stats, and will do anything — no matter how costly — to bring the single equipment item that it is given to this goal. We can represent the problem for scroll_strategist, assuming that we start with 8 STR and 4 DEX clean, like so:

{
  "stats": [12],
  "slots": 10,
  "scrolls": [
    { "percent": 10, "dark": false, "cost":  800000, "stats": [5] },
    { "percent": 30, "dark": true,  "cost": 7000000, "stats": [5] },
    { "percent": 60, "dark": false, "cost":  600000, "stats": [2] },
    { "percent": 70, "dark": true,  "cost":  500000, "stats": [2] }
  ],
  "goal": [32]
}

The cost values here are assuming that I’m scrolling for STR, and are based on rough estimates of market prices, in mesos, on MapleLegends, at about the time of this writing. If we give this JSON to scroll_strategist_cli (mirror), the basic output looks like so:

Probability of success: 12.232%
Expected cost (scrolls only): 17511978.9
Next scroll to use: 60%

Okay, so, if we only assume that I follow scroll_strategist blindly and that I start with 12 stats clean, I have a 12.232% chance of success. That’s about a one in ≈8.2 chance. Not bad, right? I can tolerate trying that many times, for sure.

Of course, not every RBD is 12 stats clean. Maybe I want to consider starting with RBDs that are 11 stats clean — either 7 STR & 4 DEX, or 8 STR & 3 DEX. All that we have to do is change the stats from [12] to [11], and scroll_strategist_cli gives us this:

Probability of success: 9.508%
Expected cost (scrolls only): 17458427.1
Next scroll to use: 30%

Okay, so, as expected, we now have a lower probability of success: 9.508%. That’s about a one in ≈10.5 chance. Still totally tolerable!

When I first started scrolling, I started trying to scroll only for STR. In the end, I would want both a STR and a DEX one; DEX for when I need the WACC, and otherwise STR for the superior damage. Since I’m usually fighting things that I can hit 100% of the time without any assistance from my overall, STR seemed like the obvious first one to go for, as I would use it more frequently. Unfortunately for me, STR scrolls are like two or three times as expensive on the market as DEX scrolls are. But again, I do really want both, so I’ll have to take the hit at some point.

By default, scroll_strategist_cli is interactive, so when it suggests a “Next scroll to use”, you can tell it whether the scroll passed, missed (i.e. failed but did not boom), or boomed. It then gives an updated probability of success, expected cost, and next scroll to use. At first, I pretty much followed exactly what scroll_strategist_cli said to do, knowing that for any given overall, it was mathematically optimal. At some point, it would become too costly, because my only choice would be to use a 30% scroll to desperately try to rescue an already mostly-failed overall. This is kind of a waste of a 30% scroll if there’s a very small probability of success, so I’d usually rather NPC the RBD instead of taking another hit to my meso purse. Luckily for me, scroll_strategist_cli can very well quantify how failed my overall is by giving me a probability of success under the assumption of optimal scrolling (read: doing everything that scroll_strategist says to do). A good heuristic is, if scroll_strategist says that the probability of success is less than the probability of success for a fresh new 11-stat RBD (≈9.5% or so), then I should probably give up on that overall and just start a new one.

I later realised that, if I want to bail out early sometimes to save my mesos from utter destruction, then I’m better off 70%ing the first slot when scroll_strategist says to 60% the first slot. This is not because scroll_strategist is wrong, but because scroll_strategist doesn’t care about my meso purse; it is given exactly one (1) RBD, and intends to absolutely maximise the probability that that specific RBD becomes a ≥32-stat RBD. In my case, clean RBDs on the market are generally not expensive enough (although this is starting to no longer be the case…) to justify continuing after failing the first slot. Thus, a 70% tends to be better than a 60%, just because it succeeds more often; sure, it can boom, but I was going to NPC it anyways unless it succeeded.

Fast-forward through a lot of pain and destruction of my mesos, and I’m starting to think that I can’t actually make an RBD. At this point, I’ve spent so many mesos on overall STR scrolls that I’ve single-handedly inflated their prices. The ultimate irony of using mathematical methods to optimise your gambling (let’s be real — scrolling items is just pixelated gambling) is that it’s still gambling! Nothing can stop you from just having shite luck!! HAHAHAH!!!!!!!

Ahem. Okay, so, maybe all of my scrolls fail, and so now I’ve made many more attempts than I statistically need to, on average, to get ≥1 successes, but I can’t let that defeat me, right? Maybe I’ll try scrolling for DEX, since I’ll be doing that anyways. Less meso devastation is always nice…

Right, of course. Of course I get it on the first or second try. Well, sweet. DEX RBD done & done.

$ ssc -j rbd.json
=== Results ===
Probability of success: 9.508%
Expected cost (scrolls only): 19810425.7
Next scroll to use: 30%
Did the scroll pass? [y/n/b]: y

Probability of success: 23.735%
Expected cost (scrolls only): 15977384.6
Next scroll to use: 60%
Did the scroll pass? [y/n/b]: y

Probability of success: 28.213%
Expected cost (scrolls only): 14078721.1
Next scroll to use: 60%
Did the scroll pass? [y/n/b]: y

Probability of success: 33.704%
Expected cost (scrolls only): 11768071.2
Next scroll to use: 60%
Did the scroll pass? [y/n/b]: y

Probability of success: 40.666%
Expected cost (scrolls only): 9274520.0
Next scroll to use: 60%
Did the scroll pass? [y/n/b]: y

Probability of success: 49.536%
Expected cost (scrolls only): 6525120.0
Next scroll to use: 60%
Did the scroll pass? [y/n/b]: n

Probability of success: 32.310%
Expected cost (scrolls only): 8521800.0
Next scroll to use: 60%
Did the scroll pass? [y/n/b]: y

Probability of success: 40.950%
Expected cost (scrolls only): 4599000.0
Next scroll to use: 70%
Did the scroll pass? [y/n/b]: y

Probability of success: 54.000%
Expected cost (scrolls only): 4220000.0
Next scroll to use: 60%
Did the scroll pass? [y/n/b]: y

Probability of success: 70.000%
Expected cost (scrolls only): 700000.0
Next scroll to use: 70%
Did the scroll pass? [y/n/b]: y

Final stats: 32

You’ll notice that scroll_strategist decided to switch from using 60%s to using a 70% on the 8^th slot (3 slots remaining), then went for a 60% when the 70% passed, and then 70%ed the final slot. This might seem weird, but a bit of consideration reveals the logic here. With 3 slots remaining and a total of 26 stats, there’s a possibility of simply passing some combination of 60%s & 70%s on all 3 remaining slots, thus reaching the goal without using any 30%s. Because the odds are actually not so bad anymore — it’s “only” 3 slots — scroll_strategist determines that it’s now worth risking the 15% boom chance of using a 70%. More importantly, if it were to use a 60% and fail, it would have no choice but to pass a 30% on both remaining slots to reach the goal, which is really bad, further justifying the 15% boom chance of the 70%. Once that 70% passes, we are only 4 stats away from the goal, which could be satisfied by passing some combination of 60%s & 70%s on both remaining slots, or could be satisfied by just passing a single 30%, regardless of what happens to the other slot — unless it booms on that other slot. Because of this flexibility (using 60%s/70%s or passing just one 30%), and because booming before getting to use that flexibility is the main mode of failure at this point, scroll_strategist backs off a bit and just uses a 60%, as the 15% boom chance of a 70% is no longer worth it. Once that passes, we are 2 stats away from the goal, so 70% is the only option — a 60% would work, but if the 60% fails, then the whole item is a complete failure, so it may as well boom.

Unfortunately for me, even a half-finished STR RBD gives me better damage range. So I really want that STR one now…

Aquatic questation

Well, let’s take a break from the abject misery of scrolling, and do something actually fun: questing! I still haven’t wrapped up Carta the Sea Witch. It’s only a level 90 quest, but I needed to kill Pianus to get Pianus’s Scream, so I had to put it off for some levels. With the Pianus’s Scream that I got in the previous diary entry, all that I need now is to kill some Bone Fish (or, as Mekhane likes to call them, “Bone Fucks”, thanks to their incredibly annoying weapon-cancels) for their Cold Tears:

Oh, yes, and I also need to break open some of the chests in Deep Aqua to get Sea Dust:

The Bone Fucks were being a bit fucky, but eventually I was able to extract a sufficient quantity of Cold Tears:

And with that… Red Musketeer Cape~!:

Very stylish.

A return to misery

Okay, back to more scrolling misery — on a quest to obtain a ≥32-stat STR RBD! Wait a sec… where are all the overall STR 30% scrolls…? Le sigh. It looks like between me buying them, and someone else manipulating the market, overall STR 30% are over. Done. Well, you know what that means: it’s time to buckle up and farm some myself, with my own bare hands. As it turns out, only three monsters drop overall STR 30%s:

• Jiaoceng,
• Horntail,
• & Red Dragon Turtles.

As I elaborated in the “Seduce me if you dare” section above, both Jiaoceng and HT are what I consider “S-tier” bosses according to my taste. That’s great, but neither of them are exactly solo content. Horntail drops get split across the runners as mesos, and I never really get to do Jiaoceng these days. So I guess it’s time to kill some RDTs…?

Regrettably, there are no good maps for RDTs. Go figure! The only two options are Dragon Forest I and The Hidden Dragon Tomb I, which have RDT populations of 52% and 47%, respectively. Both maps have pretty bad layouts; The Hidden Dragon Tomb I looks like it has an okay layout, until you get to the bottom… 😬

Dragon Forest I is the more diverse of the two (with four species represented; twice as many as are represented in The Hidden Dragon Tomb I), and it has a larger population fraction of RDTs, so I went with that one for a while:

Don’t get me wrong — I also gave The Hidden Dragon Tomb I a fair shake, grinding there for a solid chunk of time on multiple occasions.

Early on in my time at RDTs, a wild Joyce appeared!!!:

Long-time readers of this diary may recognise Joyce — IGN ducklings — as the STRginner with whom I MPQed on my darksterity knight rusa back when she was still a baby DEXgon knight. It was back then that I also met xBowtjuhNL (PriestjuhNL), Harlez (VigiI, Murhata), Gruzz (Furca), Eoka, RosePink, and all the rest. So seeing Joyce log on again was a really cool surprise!!

Yep. Only 6〜8 weeks, and rusa will surely be level 200…! Never mind the fact that level 187 is the halfway point between level 1 and level 200…… ;~;

In any case, Joyce wanted to come help me farm RDTs at Dragon Forest I, so that’s exactly what we did:

As you can see, she does 1 053 times more damage than I do, so it really sped up the farming. Well… there was one caveat:

Ouf. R.I.P.… 🪦

I was actually farming at RDTs as “extra” grinding on top of the daily challenge that I set out for myself: 60 minutes of EXP grind (usually solo Himes, but any EXP-heavy map like CDs/Gallos/Petris/etc. solo or in a party also works), plus two runs of Capt. Latanica, per day. Again with the solo Himes, I hit level 113~!:

And, after many Capt. Latanica solos, ’Tanica finally treated me with a precious MCP1!:

Which, naturally — because I opened it — had exactly one pCoin in it 🙄:

I love MapleStory!

Each ’Tanica solo takes me five minutes, which sounds pretty good, until you factor in the time to do the repeatable for a White Essence, plus travel time… I mean, I’m used to it. It’s a bit of fun, and a good break from proper grinding.

Again at solo Himes, I hit level 114~!:

Cancel Carp

Down under the Aqua Road, I did a fun Pianus (R) duo with DEX WK SwordFurbs (Yoshis, Furbs, Fabiennes, SwordFurb)!:

It was a tough fight, but we did it. Pianus being weak to fire certainly helped… And, I hit level 115 as a result!:

Now that I hit level 115, I had some SP left with nothing whatsoever to spend it on other than þͤ olde Shockwave — perhaps better known as “memewave”…

Don’t get me wrong; Shockwave is really cool. But it only works when Transformed, and my god does the animation take forever…

Over at Himes again, I got the chance to duo with marksman Level1Crook (xXCrookXx, Lvl1Crook), whom you can see using Piercing Arrow (PA) in the image below:

Most marks(wo)men don’t prioritise PA this early on, but fuck it, right? Way cooler than Snipe anyways…

I continued doing a lot of RDT grinding, and I was joined by shadower Harlez on multiple occasions. I was also joined by STRginner Taima (Yunchang, Boymoder, Tacgnol, Nyanners, Hanyou), as well:

And I… hit level 116…:

And 117………:

My god, I have killed a lot of fucking RDTs. They don’t even give that good of EXP!

Also with Harlez, and joined again by daggerknight, I did a trio Pianus (R) & Pianus (L) run:

After which, daggerknight & I learned that “sit death” is — believe it or not — a real thing in MapleStory. You see, when Energy Charge (EC) is all charged up, the bubble that surrounds your character also causes you to deal damage to any monster that touches you. This is vaguely similar to one aspect of the Power Guard (PG) skill; however, unlike PG, the damage that EC deals is not based on the damage dealt to the marauder/buccaneer at all. Furthermore, there is no particular limit on how much damage EC can deal, whereas PG’s per-hit damage is limited to ≤10% of the monster’s MAXHP. This means that, in the presence of particularly low-level monsters, I can actually straight-up one-shot kill them by simply touching them whilst my EC is charged.

Remember in the “Wiktowia Aiwand…” section above, when I mentioned that Taurospears now spawn in The Cursed Sanctuary? Well, The Cave of Pianus had a similar issue, so two Pinboom spawns were added to the bottom centre of that map. These Pinbooms have three qualities that are particularly relevant here:

• They are low-level monsters with barely any HP (viz. 400 MAXHP).
• They are immobile.
• They respawn quite quickly when slain.

So, the question is simple: if I charge up my EC, and then sit in a chair (e.g. the Sky-blue Wooden Chair) directly on top of one of the Pinboom spawns, can I kill a Pinboom whilst sitting down?:

As it turns out, the answer is actually yes! At first, I doubted that it would be possible, considering that EC fails to do anything if I am mounted (on my Hog). Then, upon actual testing, it seemed like it really wasn’t possible. Then, I realised that the damage that I take when the Pinboom initially spawns cannot damage the Pinboom, because it “isn’t finished spawning yet”, in some sense. So, in order to maintain my position in my chair, it has to initially “MISS” me. Then, the second touch attack against me causes the Pinboom to take damage from EC. So, there you have it. I still haven’t seen the “sit death” meme image, but I have seen sit death in MapleStory…

And, finally, we’re back to the bread & butter: solo grinding at Himes. Level 118~!:

Oh boy… You know what that means: just two more levels before a certain juicy job advancement…… Stay tuned!!

Oh, and in case you were wondering: no, I have not found even a single overall STR 30% scroll at RDTs. And yes, I have spent like two or three levels’ worth of EXP there. I love MapleStory!! I guess I’m just never making a fucking STR RBD!!!! 😤😤😤

drugs

drogas.

drugs.

alces my beloved

In previous diary entries, I’ve focused just a little bit on getting some cards for my daggerlord alces. She is, after all, dreadfully squishy, and that STR/DEX/LUK is lookin’ real joocy.

Actually, probably most of the card sets that I got this time were farmed on tara. I gave myself some Hime cards, and would you look at that! A lightbulb already!:

That’s tier three, bay bee 😎

In addition to picking up some of the cards that dropped at Dragon Forest I, I continued a bit of card-hunting on alces in Nihal Desert. Particularly, Magatia — like this Iron Mutæ set that I finished:

I also did the Reinforced Iron Mutæ set at Area A-3 of the Alcadno Research Institute Laboratory:

And finally, the Mithril Mutæ^[1] set, over at Area B-1:

That’s all for now. [': Hopefully I do some more card-hunting… in the future… rings plz… 🥹

Footnotes for “alces my beloved”

1. [↑] This is a bit of a weird one. As it turns out, the name of this monster species in the game data (using ␣ to represent a space) is not Mithril␣Mutae, but rather, Mithril␣␣Mutae. This is one of a surprisingly large number of whitespace mishaps that can be found in MapleStory game data. If you search the MapleLegends Library for Mithril␣Mutae, the monster species does not come up, but their card does, and so do the related species Reinforced Mithril Mutae and Reinforced Mithril Mutae. Bizarrely enough, the latter species looks like a Mithril Mutae, rather than looking like a standard Reinforced Mithril Mutae (the former species, which naturally spawns in Area A-1). I think it’s an MPQ monster, but I can’t be sure.

It is Valentine’s Day, and love is dead

On February 14^th, the day of Valentine, I attended an osslock wedding between priest j0hnny (daggerknight, Gumby, Jonathan, Kimberly) and thief twilight (chwis, chwistophwer, Christopher):

Ouch. So much for V-day… 💔

ボス

Alright, folx. Put on your bossy pants, ’cuz it’s bossin’ time.

First up, we have me playing my pugilist tarandus, attending a pair of osslocked Papulatus runs alongside outland STRginner Jestterz (Jestterzz, Jestterzzz, Jestterzzzz), osslock ranger Mekhane (Sunken, CaptainNemo), F/P perma-mage WarLikeFire, and osslock DK yeIlo (bak2monke, snakewave, VAPORWAVE):

Uh-oh… Spooky 2^nd body…

Don’t worry, we all made it out alive. Probably.

I also had the pleasure of doing a very fun pair of all-odd Rāvaṇa runs on my woodsmaster capreolina, with guildmates Yoshis (SwordFurbs, Furbs, Fabiennes, SwordFurb, PoultryWoman) the swashbuckler, and LawdHeComin (AppleBasket, Snel) the F/P archgish!!:

Naturally, none of us could pin, so we kinda just had to deal with Rav goofin’ off, waddling all over the map the whole time, and attacking us with a variety of garbage.

Here, you can see LawdHeComin using his Meteor Shower, and Yoshis using her Aerial Strike, to clear out some of the panty gobs that Rav likes to summon:

These runs were actually pretty decent! We cleared in a reasonable amount of time, and… not to brag, but… I did white both times… 😏

Speaking of Rav, I had the pleasure of doing my first Rav runs on my daggerlord alces! Poor alces is so squishy that she can’t actually survive Rav without a little of that HB juice, and somewhat unfortunately for me, my darksterity knight rusa is on the same account. That’s okay though, because that means it’s someone else’s turn to do the HBing! In this case, that “someone else” is dagger DK daggerknight (j0hnny, Gumby, Jonathan, Kimberly)! Here I am with daggerknight, and shadower Harlez (VigiI, Murhata), trio Ravvin’ it up!:

I even did some more trio Ravving, but this time with daggerknight and buccaneer Buccon (Buppoon, Buwwon, Bushroom)!:

Eventually, I remembered to actually do the goddamn Rav prequests on alces, so that I could at least get a Magical Stone Piece^[1]…:

And I switched over to capreolina to do a pair of duo Papu runs with Buccon, as well!:

Also on capreolina, I did some SURRA×Oddjobs crossover content with paladin Fuadili (who was streaming to Twitch.tv at the time^[2]), buccaneer Seiyaru, buccaneer dori of Favela, nightlord TerribleFool (iLoveShafts) of Chroma, and bishop misandrist (xRook) of Nice — Krexel!!:

Naptime tree~ 😊

it me wuza!

Also with misandrist and TerribleFool, I did a bit of Zakking on rusa. As discussed in previous diary entries, I fulfilled my role as trash woman^[3] in Zak by dutifully Monster Magneting the summons that Zak likes to spew regularly. My party members were impressed, and asked me about my job:

I should have said that my job is wastewoman… missed opportunity, I guess.

Speaking of Zakking on rusa, I did a duo (yes, you read that right: duo) Zak run with marksman xBowtjuhNL (PriestjuhNL), bringing along my STR bishop cervid for HS & HSh:

I felt pretty good about the arms stage, considering that I usually white one or two arms, even in a run with five other folx in my party. So, say goodbye to arm three^[4]!:

Indeed, we witnessed the power of the cleave just before the arms stage ended, as we had three arms alive (1, 5, & 7) and barely a sliver of HP between the three of them combined!:

Then it was time for the three bodies, where I got to test my multi-clienting skills by trying to HSh at appropriate times 🥲:

I think I was able to fulfil my roles reasonably well, even including some level of garbage collection during weapon cancels. In the end, the run took about 73 minutes — not bad for a combined total of one Energizer and one MMF between the two of us! And of course, we were rewarded for our efforts… with hot garbage:

Le sigh. At least the EXP was really good. 😅

In similar fashion, we also did a duo run of Nameless Magic Monster (NMM, Namie, Nameless)…

…With me on “pinning” duty:

At first, we were unsure whether or not I would be able to control & damage NMM enough to make the fight work out; I had run NMM as the only pinner multiple times before, but that was always with bowmasters (like Flurri and MitoAdmita) who could Hamstring the boss to make things easier. I can’t actually knockback NMM (that would require consistently dealing damage lines of 50k or more…), so I rely on strategic Rushing to constrain NMM’s movement. Being a marksman, xBowtjuhNL has Blind instead of Hamstring, so that wasn’t going to help manage NMM’s movement. Nevertheless, it still worked surprisingly well, with the occasional hiccup here or there, and xBowtjuhNL was kinda half-AFK most of the time.

✨Naturally✨, we were rewarded for our efforts with even MOAR hot garbage!!!:

Oh no, indeed.

Speaking of duo NT, I did the occasional duo ᚾᛁᚠᛚᚢᚾᚴᛅᚱ^[5] with Harlez:

We really need to get Siegfried up in here to take care of this futuristic spaceship once & for all…

Plus, we duoed a single Dunas II (better known as “2nas”). Unlike Dunas I, Dunas II comes in two stages. The first & preliminary stage would require slaying the Dunas Unit, which is the little floating spaceship thingy with the antenna:

Luckily for me, the ever-so-slightly improved vertical reach of Crusher would mean that actually hitting the Dunas Unit — as opposed to attacking just underneath it 🙄 — got a lot easier than it was previously, as illustrated by the above image. Unfortunately for both of us, however, Dunas Unit is not the only thing that we have to deal with. We also have to deal with Fake 2nas (Shielded 2nas), which is invincible and will start attacking & dispelling us willy-nilly if we manage to so much as aggro it once. So… we just didn’t aggro it:

There were a few close calls here & there, but we did make it through the entire Dunas Unit fight without aggroing Shielded Dunas at all.

Now that the Dunas Unit is out of commission, we get to fight the real-ass 2nas. Which means a lot of fun with jumping:

As you can see, we are both jumping for joy as a result of having the pleasure of fighting 2nas. Just kidding, we’re trying to dodge 2nas’s dispel, a fun little move that 2nas likes to use every four seconds or so. The problem with this technique is that networking latency exists. Timing your jump depends a lot on how long a round-trip takes between your client and the server, so if you guess wrong about how long that round-trip takes (an easy “mistake” to make, as any of a huge number of factors may influence the latency at any time), the server might have the wrong idea about your exact coordinates at the time that the dispel actually occurred. Fun!!

After a whole lot of Rushing to keep 2nas from wandering away, recasting countless buffs, nearly dying several times, and an altogether uncomfortably lengthy battle against this smug little twerp, we were rewarded with— Go on… guess……

🌡️🔥🌡️ HOT fUCKING GARBAGE!!! 🌡️🔥🌡️

Don’t worry, I’m fine. We’re fine. It’s… fine. As long as I never have to fight 2nas ever again so long as I live… 🫠

Footnotes for “ボス”

1. [↑] Come to think of it, I don’t think I’ve even once heard someone refer to this item as “Magical Stone Piece” or “MSP”. I’ve heard “Rav stone”, which I used for a bit myself, until I found people confusing that terminology with Sunburst. Sunburst is yet another ETC item associated with Rav, and if you squint a little bit and ignore its name, it could maybe be a really shiny polished spherical stone. Nowadays, I refer to the Magical Stone Piece as a Rav “cube” due to its cubic appearance, and because the Sunburst is clearly spherical, thus avoiding that kind of confusion. Because one requirement for entering the Rav map is that the party leader has a Magical Stone Piece in their inventory, we say that someone who has a Magical Stone Piece in their inventory is “cubed”. Note that, although you do have to drop a Sunburst in the Rav map to summon the boss itself, Sunbursts are in no way required to enter the map (contrast this with Capt. Latanica and its Engine Room). Furthermore, only the party leader needs to be cubed.
2. [↑] https://www.twitch.tv/fuadili1
3. [↑] 4.1×10¹¹ dead cops.
4. [↑] Counting arms from left-to-right, top-to-bottom.
5. [↑] This is Old Norse ⟨Niflungar⟩ */ni(ː)flu(ː)ngɑ(ː)r/ — the Old Norse equivalent of German (Middle High German) Nibelung — written in Younger Futhark. This corresponds with the use of the term in Atlakviða (ca. 900 AD).

(…cnvpstdf…)