winrates and "winscumming"

[Aule]

{MOD NOTE: Split from another topic.}

It seems pretty clear to me that the existence of long streaks and high winrates disprove your "helpless" claims.

Actually, it should be clear that the existence of streaks amongst such numbers of deaths more likely proves the influence of randomness. Randomness almost guarantees streaks will appear infrequently. Also, the high winrates you see are likely the product of winscumming. I'd be willing to wager there are very few with a high winrate who have used the same name for all their games. Switching to build a better winrate after learning the game better is winscumming, plain and simple. In addition, one can play offline, try things out in wizmode, play a few games to get the hang of a combo, and only then get online to get the "win." This is also winscumming, plain and simple.

I think you find far too much to admire in winrates and streaks. A genuine assessment of Crawl should never derive from such antics.

[Arrhythmia]

"winscumming"

[Bloax]

it's like thankscumming except even more sinister because now it actually lets you swing around your e-schlong for real

[kroki]

It seems pretty clear to me that the existence of long streaks and high winrates disprove your "helpless" claims.

Actually, it should be clear that the existence of streaks amongst such numbers of deaths more likely proves the influence of randomness. Randomness almost guarantees streaks will appear infrequently. Also, the high winrates you see are likely the product of winscumming. I'd be willing to wager there are very few with a high winrate who have used the same name for all their games. Switching to build a better winrate after learning the game better is winscumming, plain and simple. In addition, one can play offline, try things out in wizmode, play a few games to get the hang of a combo, and only then get online to get the "win." This is also winscumming, plain and simple.

I think you find far too much to admire in winrates and streaks. A genuine assessment of Crawl should never derive from such antics.

winscumming? what the fuck man. how is playing playing to your best ability without exploiting the game *scumming?

[Bloax]

It's a joke about how most high-winrate accounts aren't actually the person's original account they started out with that has all their early splats.

[Sar]

I don't see the point? Like yeah, maybe they didn't start being good at Crawl, but now they are... how is that scumming exactly?

The only kind of "winscumming" I can think of is starting games on multiple servers, getting to D:1 with the Orb and then winning them all at once rapidly to get your streak. That's easy to check by looking at game started - game finished dates though.

[Sandman25]

I don't see the point? Like yeah, maybe they didn't start being good at Crawl, but now they are... how is that scumming exactly?

Yesterday I got something like "Wow, 75% wins. Who are you?" in game chat. That was pleasant Yes, it's winscumming and does have a point.

[Aule]

winscumming? what the fuck man. how is playing playing to your best ability without exploiting the game *scumming?

I don't see the point? Like yeah, maybe they didn't start being good at Crawl, but now they are... how is that scumming exactly?

It's really simple, leading me to conclude that the gasps here probably reveal its practitioners. Using multiple accounts, wizmode and offline playing solely to bolster an online winrate makes that winrate an invalid metric, again, plain and simple. It's not real. It's dishonest. And it is exploitation that is not appreciably different from savescumming. If online winrates are to be used as a deciding metric for game design discussions, then the only way they can be valid is if they are immune to this kind of scummy behavior, which Crawl's design philosophy expressly forbids. Unfortunately, there is no easy way to prevent it, and it happens not in the game, but the IRC channels and the bots within them, so Crawl development can't change it. The best approach is to simply ignore winrate arguments altogether as they are meritless.

If you want winrates to have merit, you have to prevent offline play. My solution would be twofold: 1) close the open source, but still make it accessible to registered accounts, who would then be eliminated from any standings or bot-generated charts, and: 2) make trunk unavailable for download as binaries, and calculate winrates solely from trunk. Otherwise, its meaningfulness will remain little more than dick-diddling shennanigans.

But please do feel free to ignore me and the elephant in the room; I just have a thing against glaring double standards.

[Aule]

Yesterday I got something like "Wow, 75% wins. Who are you?" in game chat. That was pleasant Yes, it's winscumming and does have a point.

Oh, man, I so greatly admire and appreciate your honesty. Perhaps others will be as forthcoming and this charade will finally be revealed.

@bloax: sorry for the derail. I haven't tried your mods, but I know you must have put a lot into them, which gets even more admiration from me.

[Sandman25]

Using multiple accounts, wizmode and offline playing solely to bolster an online winrate makes that winrate an invalid metric, again, plain and simple. It's not real. It's dishonest.

I disagree. You mention very different things here.
1) Wizmode is a tool to make crawl more user-friendly. Examples: fsim. +8 spear of pain at min delay or +6 lajatang of speed. Actually +2 lajatang of speed, 4 scrolls of enchant weapons and 0 skill levels in Staves (but 8 from crosstraining). Before making an important decision I need all info I can get and unfortunately crawl provides it in wizard mode only. Another example is check of AC/EV if I want to buy some expensive body armour in a shop, is it better than my current armour? I can easily calculate it for AC but not for EV
2) Offline play. As others mentioned, players become better. If I have 75% win rate now, why should I have 1% win rate? Just because I started playing in version where monsters didn't have their spell listed, had their stats completely hidden and distortion could be present as every runed weapon? Erolcha, ok, who is that unique, what does she do? I don't know. What is her HP/AC/EV/MR? Let's retreat behind a corner and try to melee her. Ouch, she has a bolt of lightning, standing near the wall was a "mistake".
Next game. Ok, Erolcha again. Let's try to confuse her. How "easily resists"? Ouch, where am I? What is that? Executioner? Ouch. Next game etc.
3) Multiple accounts. My Sandman25 has 40 games, 17 of them were played during CSDC tournament, they had extra rules which made it easy to die. Another 15+ games my attempts to win combos which I lost in the tournament (DrCK, for example). It means there were only about 6-8 games where it was me who selected combo. Do you really think it correctly reflects my real current win rate?

[Sar]

It is literally impossible to use offline to boost your online winrate. It does not make sense.

By your "logic" an Olympic record runner is a fraud because he used to crawl when he was a child and he doesn't run to get his groceries at the top of his ability. Except that's stupid, isn't it? You should judge an account for what it is - an account. A player creates it and he plays it - he might have an account where he plays random combos, he might have an account where he will casually quit a character that is not fun to play, he might have an account where he will startscum to get a monstrous Demonspawn, and he might have an account where he plays to the best of his ability and streaks multiple games. Just like a runner doesn't put in his best effort when he's walking down the street to buy some veggies, but when he does it he's really good at that. That is impressive, and that should be judge regardless of his other accounts or offline games because that is the only way that makes any kind of sense.

Disclaimer: I am not trying to imply that playing Crawl is in the same ballpark as being an athlete. It isn't. Please don't try to ridicule my argument by attacking that part.

[Sandman25]

Yesterday I got something like "Wow, 75% wins. Who are you?" in game chat. That was pleasant Yes, it's winscumming and does have a point.

Oh, man, I so greatly admire and appreciate your honesty. Perhaps others will be as forthcoming and this charade will finally be revealed.

Of course, I am honest. 75% looks right for my win rate when I don't choose stupid combos like MiWz and try to play it as typical Wz.

[Aule]

Using multiple accounts, wizmode and offline playing solely to bolster an online winrate makes that winrate an invalid metric, again, plain and simple. It's not real. It's dishonest.

I disagree. You mention very different things here.
1) Wizmode is a tool to make crawl more user-friendly. Examples: fsim. +8 spear of pain at min delay or +6 lajatang of speed. Actually +2 lajatang of speed, 4 scrolls of enchant weapons and 0 skill levels in Staves (but 8 from crosstraining). Before making an important decision I need all info I can get and unfortunately crawl provides it in wizard mode only. Another example is check of AC/EV if I want to buy some expensive body armour in a shop, is it better than my current armour? I can easily calculate it for AC but not for EV

This is a fair use of wizmode, but wizmode is by no means limited to fair use.

2) Offline play. As others mentioned, players become better. If I have 75% win rate now, why should I have 1% win rate? Just because I started playing in version where monsters didn't have their spell listed, had their stats completely hidden and distortion could be present as every runed weapon? Erolcha, ok, who is that unique, what does she do? I don't know. What is her HP/AC/EV/MR? Let's retreat behind a corner and try to melee her. Ouch, she has a bolt of lightning, standing near the wall was a "mistake".
Next game. Ok, Erolcha again. Let's try to confuse her. How "easily resists"? Ouch, where am I? What is that? Executioner? Ouch. Next game etc.

Exactly. That would make "winrate" a valid metric, because that is how one must learn crawl in an "unspoilery" way. It is how I play, and why my metrics are accurate. I have not played a single offline game since going online. My winrate (deathrate, more accurately) reflects the real state of Crawl's difficulty, which is what elicited my original response.

3) Multiple accounts. My Sandman25 has 40 games, 17 of them were played during CSDC tournament, they had extra rules which made it easy to die. Another 15+ games my attempts to win combos which I lost in the tournament (DrCK, for example). It means there were only about 6-8 games where it was me who selected combo. Do you really think it correctly reflects my real current win rate?

The only valid reflection of your actual winrate would combine every game you played online, and this still would leave out offline games, which cannot be adjudicated at all. Thus, there is no valid measure given by online winrates that can be used in any guiding decisions about Crawl's difficulty, unless the design philosophy is amended to allow for scummy behavior in preparation for the game.



@mods: can this be split off? I feel bad now for derailing bloax's thread.

[Aule]

By your "logic" an Olympic record runner is a fraud because he used to crawl when he was a child and he doesn't run to get his groceries at the top of his ability.

Not even close, Sar. Shame on you. Crawl is not an evolution. Every game is counted. Artificially manipulating those counts is disingenuous where those results are used to "prove" something. There are no "free" games in crawl. No walks to the corner store. Every "step" in the player's life is a death-defying race. Your analogy would hold water only if every step the athlete took was in a competition, because every Crawl game is counted in the competition. Back to you, Jim.

[wheals]

By your "logic" an Olympic record runner is a fraud because he used to crawl when he was a child and he doesn't run to get his groceries at the top of his ability.

Not even close, Sar. Shame on you. Crawl is not an evolution. Every game is counted. Artificially manipulating those counts is disingenuous where those results are used to "prove" something. There are no "free" games in crawl. No walks to the corner store. Every "step" in the player's life is a death-defying race. Your analogy would hold water only if every step the athlete took was in a competition, because every Crawl game is counted in the competition. Back to you, Jim.

for a second there i really thought you were saying all that stuff above seriously. Poe's law'd again I guess.

[cerebovssquire]

leading me to conclude that the gasps here probably reveal its practitioners.

Not even close, Sar. Shame on you. Crawl is not an evolution. Every game is counted. Artificially manipulating those counts is disingenuous where those results are used to "prove" something. There are no "free" games in crawl. No walks to the corner store. Every "step" in the player's life is a death-defying race. Your analogy would hold water only if every step the athlete took was in a competition, because every Crawl game is counted in the competition. Back to you, Jim.

i kind of picture you with a flustered face, in a toga, and in the middle of a square in ancient rome

[Aule]

for a second there i really thought you were saying all that stuff above seriously. Poe's law'd again I guess.

I was, and I am. I looked up Poe's law, and no, I'm quite serious. It's really not all that difficult. If you have a specific question, I can answer it for you.

[DrKe]

tbh it's a valid point but i'm not really sure why one would care about the practice unless the "winscummer" is openly bragging about it

[Sandman25]

It is literally impossible to use offline to boost your online winrate. It does not make sense.

It is literally required to use offline with optimal play. Example from my last online game. OpTm got storm bow from Oka on D13. How much damage do UA/the bow do? Is it worthwhile to switch? fsim answered yes. Another question appeared a bit later - is it a good idea to go for Statue Form with the bow? How much Transmutations/Earth would I need for that (with and w/o wizardry)? How much AC would I get? What about Stoneskin, how much AC will it give me at those Transm./Earth levels?

[wheals]

@Aule

OK, well, then, I think you have the wrong idea. Sar made the point of not every game being a competition, but that's wrong; _no_ game is a competition unless you want it to be. If you want to afflict yourself by fsimming every weapon combination under the sun, it doesn't affect me. Similarly, if you want to open lots of new accounts I certainly don't mind -- personally, I'm still on my own account, and I've only played ~5 games offline (though I have used wizmode extensively in code testing). It makes a lot of people feel good to see a nice big number after their name, and not necessarily because they compare their number to other people's.

Crawl is a singleplayer game, and it always will be. If any number of people want to compare their scores, they can choose any arbitrary group of games to compare, and if any person is unhappy with the comparison, they have every right to be -- but it isn't right to barge in on other people's fun and tell them that they're doing everything wrong because you want to take a more holistic view. Everyone plays in their own way, and you can feel free to play however you want, but don't accuse other people of cheating (which *scumming gives at the very least a connotation of) because you prefer it differently.

It is literally impossible to use offline to boost your online winrate. It does not make sense.

It is literally required to use offline with optimal play. Example from my last online game. OpTm got storm bow from Oka in Lair. How much damage do UA/the bow do? Is it worthwhile to switch? fsim answered yes.

No, it answered with a flood of numbers. If, because of your skill (I don't deny that you're a better player than I) you interpreted those well, congratulations.

e: I don't deny that in your situation using wizmode is theoretically optimal, but in any case where it does mean the difference between life and death an experienced player will be able to estimate well enough. If your optimality only means you have an extra potion of hasting at the end of the game, it had no effect...

[Sivar]

you guys are really stretching the "or whatever" in the thread title about as far as it should go so if you're not done discussing competitive epistemology or whatever can you please take it to another thread, preferably by moving the posts already in this one

[Sandman25]

No, it answered with a flood of numbers. If, because of your skill (I don't deny that you're a better player than I) you interpreted those well, congratulations.

e: I don't deny that in your situation using wizmode is theoretically optimal, but in any case where it does mean the difference between life and death an experienced player will be able to estimate well enough. If your optimality only means you have an extra potion of hasting at the end of the game, it had no effect...

I keep hearing that argument about experienced players being able to estimate damage and I always wonder - how do they do it? In this particular example - do they shoot the bow at skill level 5 (thanks to heroism) and compare it with UC 13 and UC 18 (my current skill level in UC was 13). How does it help to know if bows 16 is better than UC 22?
Or do you mean the experienced players already had this situation in one of their previous games so they know it from experience? What if it is your first time when you get the storm bow and you have no idea how much damage it causes to rElec and non-rElec monsters at high skills? Is fsim still an unnecessary luxury with optimal approach in this case?

I agree with the rest of your post.

[DrKe]

No, it answered with a flood of numbers. If, because of your skill (I don't deny that you're a better player than I) you interpreted those well, congratulations.

e: I don't deny that in your situation using wizmode is theoretically optimal, but in any case where it does mean the difference between life and death an experienced player will be able to estimate well enough. If your optimality only means you have an extra potion of hasting at the end of the game, it had no effect...

I keep hearing that argument about experienced players being able to estimate damage and I always wonder - how do they do it? In this particular example - do they shoot the bow at skill level 5 (thanks to heroism) and compare it with UC 13 and UC 18 (my current skill level in UC was 13). How does it help to know if bows 16 is better than UC 22?
Or do you mean the experienced players already had this situation in one of their previous games so they know it from experience? What if it is your first time when you get the storm bow and you have no idea how much damage it causes to rElec and non-rElec monsters at high skills? Is fsim still an unnecessary luxury with optimal approach in this case?

I agree with the rest of your post.

having a ranged weapon adds a lot of utility, so even if i was actually wrong about whether or not i was too invested in UC already, it's not like that is a strictly bad decision to put some skill in it

in the case of UC vs a melee weapon i think you could make a bad decision but it would most likely not make you lose the game

[Sandman25]

having a ranged weapon adds a lot of utility, so even if i was actually wrong about whether or not i was too invested in UC already, it's not like that is a strictly bad decision to put some skill in it

in the case of UC vs a melee weapon i think you could make a bad decision but it would most likely not make you lose the game

It could cost me a game, Tm is XP-starved background, even for Op. Also we are talking about optimal play here, we all know that it's possible to win despite multiple skilling mistakes.

Edit. The ranged weapon forced me to amnesia Animate Skeleton so I am not even sure it would be good as utility (Tab does not trigger warning if my ally is behind target monster as I learned, got penance twice- first time I wasn't even sure what happened and suspected I pressed a wrong button as answer). After getting Titans and co. twice on Snake 1/2 I amnesiad skeletons.

[DrKe]

having a ranged weapon adds a lot of utility, so even if i was actually wrong about whether or not i was too invested in UC already, it's not like that is a strictly bad decision to put some skill in it

in the case of UC vs a melee weapon i think you could make a bad decision but it would most likely not make you lose the game

It could cost me a game, Tm is XP-starved background, even for Op. Also we are talking about optimal play here, we all know that it's possible to win despite multiple skilling mistakes.

Edit. The ranged weapon forced me to amnesia Animate Skeleton so I am not even sure it would be good as utility (Tab does not trigger warning if my ally is behind target monster as I learned, got penance twice- first time I wasn't even sure what happened and suspected I pressed a wrong button as answer). After getting Titans and co. twice on Snake 1/2 I amnesiad skeletons.

yeah i got oka wrath from shooting a skeleton with a blowgun needle, no warning

it just comes down to how interested you are in playing 100% optimally, i suppose

[Sandman25]

it just comes down to how interested you are in playing 100% optimally, i suppose

Right. Even if storm bow was slightly worse, I would switch anyway, I always wanted to try storm bow with Op slaying (+7 slaying so far). But only if "slightly"

[Aule]

@Aule

OK, well, then, I think you have the wrong idea. Sar made the point of not every game being a competition, but that's wrong; _no_ game is a competition unless you want it to be. If you want to afflict yourself by fsimming every weapon combination under the sun, it doesn't affect me. Similarly, if you want to open lots of new accounts I certainly don't mind -- personally, I'm still on my own account, and I've only played ~5 games offline (though I have used wizmode extensively in code testing). It makes a lot of people feel good to see a nice big number after their name, and not necessarily because they compare their number to other people's.

Crawl is a singleplayer game, and it always will be. If any number of people want to compare their scores, they can choose any arbitrary group of games to compare, and if any person is unhappy with the comparison, they have every right to be -- but it isn't right to barge in on other people's fun and tell them that they're doing everything wrong because you want to take a more holistic view. Everyone plays in their own way, and you can feel free to play however you want, but don't accuse other people of cheating (which *scumming gives at the very least a connotation of) because you prefer it differently.

I agree that it's none of my business how others get off. The issue isn't purity or holistics, though, but the fact that these winrate arguments are repeatedly raised in support of game-defining changes, when they should not be. The test of real difficulty could only be measured by such drastic purity, and that was my sole point.

Scumming never implied cheating, to me, because it's used in many situations that are not considered cheating. Zigscumming is only one example. To me, the "scummy" connotation is a pure pejorative loosely defined as manipulating tactics to achieve unnatural results. I'm certainly not trying to say everyone should play my way, but only it is folly to consider one's self-fondled penis length as some measure of superiority over another who just lets it hang naturally. Certainly, the game should not be defined based on such things.

I also support moving this discussion, as I've requested.

[and into]

True Crawl Confessions:

Before tournaments, I go up into the mountains to train DCSS for three weeks, then come back down right before it starts. I have even dabbled in injecting EPO and steroids, and most of my wins involved a lot of caffeine abuse.

My winrate is a sham. I hope the community I have lied to for so long can forgive me.

[Tedronai]

So we are doing crawl confessions now? Here is mine: The combined winrate of all my accounts is 64.89% (and I have a total of 2 offline wins from before I started playing online).

[Lasty]

I keep hearing that argument about experienced players being able to estimate damage and I always wonder - how do they do it? In this particular example - do they shoot the bow at skill level 5 (thanks to heroism) and compare it with UC 13 and UC 18 (my current skill level in UC was 13). How does it help to know if bows 16 is better than UC 22?
Or do you mean the experienced players already had this situation in one of their previous games so they know it from experience? What if it is your first time when you get the storm bow and you have no idea how much damage it causes to rElec and non-rElec monsters at high skills? Is fsim still an unnecessary luxury with optimal approach in this case?

The argument is not that experienced players know the exact damage storm bow and UC will inflict on a variety of monsters given a variety of skill levels and attributes. The argument is that given two alternatives, if one is not clearly better than the other based on observable phenomena and available data, it doesn't particularly matter which one you choose. Your choice may technically turn out to be non-optimal, but unless you let yourself get into situations where a tiny difference in damage/accuracy is the only thing separating life from death, it doesn't matter. You yourself have pointed out that an experienced player can win the game with a flaming spear, which is obviously a bad weapon; having any kind of reasonable weapon is good enough. There's no prize for having chosen the exact best weapon choice. Of course, that's regarding immediately comparable weapons, and your question relates to comparing weapons when you don't have skill in one of the weapon classes.

It's absolutely true that players won't always have enough information to make the best choices the first time they play the game. The game is intentionally built around the idea that you will learn by experience, and usually what that means is learning a few things, sticking with those as a foundation and trying to build off of it, and eventually getting to the point where you're confident about those things, and then branching out and trying something slightly different that you can contrast with that existing knowledge. You can choose to spend time code diving or running fsim or watching other players or whatever you like to accelerate that learning process without having to actually play the game, but it's not necessary, and it's entirely possible to learn entirely misleading things by doing so and understanding/applying the information wrong.

So an experienced player might already have a good sense of how weapons will scale with skill based on previous experience, and an inexperienced one may not. The inexperienced player is probably best served (on several levels) by sticking to one class of weapons per character until they get enough experience with how to compare weapons while keeping skill level as a constant, at which point they can have a reasonable expectation of how differing weapons scale between differing skill levels. All the classes of weapons work similarly enough that knowledge gained about one class can be applied fairly reliably to other classes as well.

[johlstei]

To the extent that you are using winrate as a proxy for how good/consistent a player is, it is inappropriate it include games from when they first started, or at least to weight them equally to more recent games. Recent games are going to capture a players ability to play at present much better than ancient games are. From this perspective, some sort of moving average of winrates that discounts older games is a much better factor in predicting the players likelihood to win a given game. I don't think the rate at which a player learned the game is a particularly interesting thing to measure or talk about in this context, and that's all you are really measuring if you talk about all of the games in which a player learned.

I would call "winscumming" playing on every server and then just never finishing the game if you think you're going to die. There is clearly only finite capacity for that though and thus it can't possibly have that large an effect.

[XuaXua]

I have 12 fingers.

This has not helped me win.

[Sar]

I actually wrote a pretty angry and toxic reply to Aule but the site died and now I don't even care anymore.

gg I guess, you trolled me hard

[DDDD]

gg I guess, you trolled me hard

If someone has better arguments that doesn't necessarily mean he's trolling you.

[Sar]

"if"

[johlstei]

I have a truly marvelous demonstration of your foolishness which this post is too short to contain.

[tasonir]

I'm at 47 wins with a glorious 1.8% winrate, I've only ever used one account online.

I've thought about making a second one for high win rate play, but I didn't mostly just because I'm too lazy. I don't think there's anything wrong with having a separate account for attempting streaks and such. About the only abuse here would be the already mentioned playing games on multiple servers at the same time so you can win them back to back. Hopefully that can be patched at some point when the account system gets figured out, but even if it doesn't, I don't think it happens very often, and players can check for it.

I've always thought mikee's 22 win streak was interesting, because it starts in .10, continues into .11, and then suddenly when he plays NaTm it goes back to .10 for constriction. And then back to .11 after that game There are plenty of things you can do if you're looking to be sneaky...

[Zooty]

So we are doing crawl confessions now? Here is mine: The combined winrate of all my accounts is 64.89% (and I have a total of 2 offline wins from before I started playing online).

Once I quit a character to keep my uniqueplayer status.

[Arrhythmia]

it's like thankscumming except even more sinister because now it actually lets you swing around your e-schlong for real

As a noted thank-scummer I'd appreciate it if you didn't equate my art with the non-idea of "winscumming".