Enchanter and Conjurer are way too luck dependent. And more.

[njvack]

IIRC, most of the clustering illusion questions come up with respect to item and monster generation. And since there's a new version of crawl literally every single day, questions like "did this change or is it just me" are pretty reasonable if you're not following the git logs. No one asks if something suddenly changed in Nethack's monster tables, because 3.4.3 has been the latest version since 2003.

I don't know why damage formulas in Crawl are the way they are, though.

[bel]

On damage numbers, going a bit offtopic: I played a bit of Sil recently. The transparency of the damage formulae was refreshing (it even gives combat rolls in a window!), but I did spend a lot of time in "analysis paralysis".

But I do that with Crawl as well, (though much less) so it might be a personal problem. I am a very slow and deliberative player in general (in chess I get into time trouble often).

I am not sure how a flurry of numbers would go for a newish player. But I'd guess that people who play crawl are nerds anyway, and don't mind some numbers thrown at them.

(btw: Sil's combat system is awesome, I would really like Crawl to incorporate some of it, but I am not sure if it really fits in Crawl as it exists now).

[mps]

for example the idea that estimation of probability distributions is something human beings are uniquely bad at

It is. And this isn't just a crawl thing. We're hyper-sensitive to patterns, and spot them when they aren't there. There have been studies of V2 bomb targets in Britain, gamblers, etc. It just happens. We're bad at estimating probability distributions.

If you want to talk about probabilities being wrong, there's a valid empirical way to do it. Record 100 consecutive results, and post THAT. That's actually data we can work with. 'I feel like my hexes fail way more than the number' isn't.

I would appreciate some indication of understanding the point made in the previous post before giving me a list of facts you know that you think other people don't.

Estimation of probability distributions is not an easy thing to do in general, but it's something that's been studied carefully for a long time. Some distributions can be usefully and accurately estimated by naive heuristics with relatively small numbers of samples. Others can't. This is not an observation in psychology. It's a fact of probability. Crawl combat uses the latter kind of distributions, whereas in systems used for the entertainment of human beings, the former is probably more appropriate.

As to probabilities like hex chances and spell fail rates given in game, I've looked at code for the latter and though I think they're probably right, I wouldn't bet much money on it. Hex chances are sometimes flat out wrong, apparently.

[njvack]

As I understand it, hex chances are wrong when a monster has extra MR but you don't know.

[ThreeInvisibleDucks]

The game also shows you have a chance of casting Ensorcelled Hibernation on a monster with a rC armour, even though Ash told you so. Lost a SpEn recently to it because I wasn't paying attention to the "unaffected" messages.

[archaeo]

Not all players know that they can look up this information in the source code. Not all players know how to do so if they want to. Not all players who know how to will bother. Those who do will not necessarily easily connect the properties of these complicated formulas with their experience playing because they're pretty damn complicated and relevant information that would help connect these two things is concealed in-game. All of these factors contribute to the difficulty in discussion of crawl combat damage in a way that favors keeping crawl combat damage the way it is.

That's not what I see happening, though. If you don't understand how some mechanics work, the easiest way to get help on Tavern is to say something wildly wrong and wait for duvessa to swoop in and tell you what a boner you are. Alternately, you can just ask. I bet if you politely asked, in ##crawl-dev, ##crawl, or on Tavern, where to find mechanical formulae and how to interpret them, you'd get help.

Feel free to correct me, but I think what you're criticizing is the fact that, because these mechanics are obscure, discourse re: changing them is stymied. Of course, that assumes that we need to know the precise mechanics to criticize how Crawl feels when you play it. Also, I assume we're just talking about design concerns, and not actual in-play concerns, unless you want Final Fantasy Tactics Crawl as well.

Whenever you say "Hey, you say something is hidden and that has negative effects, therefore you are saying there must be a nefarious agent responsible for the hiding who wills those negative effects, which sounds like a crazy conspiracy theory and can't possibly be true," you should slap yourself. People say this kind of thing a lot, so we'd live in a better world, but one where people slap themselves a lot more, if everyone followed my advice.

I miss the fun, irreverent mps instead of the current, mean mps.

When you say, and I'm quoting, "I'm pretty sure the reason numbers like this aren't made more available is that they're all over the place and would immediately draw criticism," that sounds conspiratorial to me. I was flippant about it, and it's possible you were being flippant too. I feel like we should be able to agree that the devs pretty obviously have used complicated mathy mechanics because they think it makes the game better, even if we disagree.

Actually, it would be more fair to say that many probability distributions that appear in crawl are hard to estimate based on samples as a matter of mathematics and their being far from normal -- which is arguably a questionable design choice. The clustering illusion trope has a spurious, built-in defense of that choice.

I still fail to see how it isn't operative, when we're discussing things that we have no reason to believe have meaningfully changed (spellpower, vault selection, etc.) that others are saying "oh, this is changed, I just had minmay_frog_pond a gajillion times." Sure, it gets overused, but it's probably fairly used as well.

This is inevitably where someone brings up the double damage bug, but I tend to see that as a success of the system instead of the failure. Lots of independent goodplayers all thought something was off, it was investigated, and it was fixed; the only reason it came off as a big deal was because it happened during the tournament.

Estimation of probability distributions is not an easy thing to do in general, but it's something that's been studied carefully for a long time. Some distributions can be usefully and accurately estimated by naive heuristics with relatively small numbers of samples. Others can't. This is not an observation in psychology. It's a fact of probability. Crawl combat uses the latter kind of distributions, whereas in systems used for the entertainment of human beings, the former is probably more appropriate.

This doesn't seem quite right. Does Crawl behave so counterintuitively? As someone who is outright ignorant of all things above Stats 101, I've been able to win, and win consistently, by predicting how Crawl will behave. Is my experience so out of wack?

[tasonir]

Imho a lot of the very complex formulas were probably inherited from ancient versions, and while I'm sure most devs would prefer simpler formulas for most things, it's a lot of work for marginal gain. And if you change them significantly then there's weeks of playtesting and balance concerns afterwards. It'll make some number crunchers happy, but players interested in statistics are the minority, even if it does include the most important player (that's me). No need to be hostile to any devs, current or otherwise. I have not tracked down the history of exactly how old the formulas are, feel free to let me know if you know.

[PowerOfKaishin]

You guys are nice to new players, but surprisingly hostile to one another. Not what I expected at all.

Is there a way to quickly kill ice trolls? Wandered into one in the ice cave without any rC (don't ask; I was doing just fine until that thing showed up thanks to confuse + dazzle).

And while I find it hard to believe fist only does one damage less, it doesn't really change the fact that a Tengu Conjurer can't one-on-one a Jackal.

Was it a troll simulacra? Those are slow, so you can walk away and hit them from a distance. Weak to fire and dispel undead.
I can't think of what else might be confused for an "Ice Troll," which I don't recall being an enemy in the game.

Sorry, it was an ice giant. My bad. Yeah, thing ripped me a new one, like that weapon from Ratchet and Clank.

As I understand it, hex chances are wrong when a monster has extra MR but you don't know.

Why not just display the correct percentage?

Imho a lot of the very complex formulas were probably inherited from ancient versions, and while I'm sure most devs would prefer simpler formulas for most things, it's a lot of work for marginal gain. And if you change them significantly then there's weeks of playtesting and balance concerns afterwards. It'll make some number crunchers happy, but players interested in statistics are the minority, even if it does include the most important player (that's me). No need to be hostile to any devs, current or otherwise. I have not tracked down the history of exactly how old the formulas are, feel free to let me know if you know.

Balance concerns... as long as it's not gamebreaking I don't really see what could be so imbalanced considering not all races were created equally and that was obviously intentional. I doubt a flat +2 bonus and a nerfed damage dice for magic dart is gonna snap the game in half. >_>

Though, if it completely overshadows the other level 1 spells, that could be a problem.

[jejorda2]

That Frost Giant is kind of the boss of the Ice Cave. Most characters are not ready for something that hard at that time, and should run.

Blowgun and Curare can take him out while you retreat. An identified wand of fire (or maybe fireball, but not flame) can be a big help, as could a lamp of fire. Brothers in Arms can be good, or Nemelex Summons.

Bolt of Cold can hit you through other creatures for 3d25, though, so it's an easy way to die.

[triorph]

Why not just display the correct percentage?

It could be nice to do this, but the problem is that it then telegraphs to the player that the monster is wearing something with MR+, when they had no way of knowing this prior. The current identification mini-game requires that these things be hidden, and by including them in to the hex calculation you would be subverting that. There might be an argument for removing that aspect of identification altogether (auto-id all ego types?)

[mps]

@archaeo: I think tasonir's basically correct about why there are weird, complicated formulas. People always say dcss was basically an autoexplore mod at its inception, but another major issue at the time was broken combat mechanics. There was apparently an overhaul that resulted in roughly what we have today. I wouldn't quite say it's broken, but it's definitely eccentric. Anything that's moderately complicated but kind of works for reasons that aren't easy to grasp will stay in software for a long time. It's not obviously true that weird formulas are in the game because any current dev thinks they're the best thing possible or even good, except to the extent that they kind of work. Basically, I think primordial crawl was full of weird, goofy formulas and ill-considered mechanics, along with some big ideas that made it kind of work anyway. I remember thinking this fifteen years ago when I was thinking about fiddling around with super-old crawl and making some kind of variant. A lot of the bad ideas and goofy formulas have been removed or smoothed out, but the legend continues (see combat system, food, curses, identification, etc.).

As for my contention that the numbers are hidden because they're obviously ridiculous if you have them right in front of you, I don't think that's a conspiracy, it's the obvious thing to do if you don't want to overhaul combat again. I realize there's an official explanation, but it pretty obviously doesn't wash. I mean, I can believe that certain people really think there's an aesthetic or usability argument for making combat damage less transparent than it is in dragon warrior and essentially every other commercial roleplaying title, but it's a really silly idea that I can't believe is the first mover. I would be willing to grant that there's a plausible alternative, though: Originally, not showing damage numbers was just what the other roguelikes were doing at the time. I think the connection between not showing damage numbers and the persistence of a system with wild damage distributions way outside the norms of the genre, whether you take that to be roguelikes or CRPGs more generally, is undeniable though.

I hope you're trolling me about double damage. I can't think of any other game where you could double player combat damage and it wouldn't be immediately, blindingly obvious to absolutely everyone, even without explicit damage numbers being available.

About clustering illusion, if your primary way to address discussion of probability is clustering illusion, that's a problem. I don't care how many times it's the right thing to say. A lot of times it isn't.

re: winning at crawl, you and I and probably everyone else who regularly wins and hasn't been playing since 2006 learned tactics from other people that allow one to deal with the crazy damage distributions of crawl. Part of it is that crawl just isn't that hard, in spite of the weird combat damage. I think complicated formulas and wild damage distributions are something you can learn to work around as a player, but the difficulty in analyzing them from a game balance/design perspective is unnecessary and limiting. The easier it is to wrap your brain around mechanics and get a feel for the probability distributions that come up, the easier it will be to think through new features or changes. The current combat system is not very user friendly, hard to think about, and is not better than combat systems in other games that don't have these problems.

[bel]

A bit of lateral perspective: combat in Sil can be even spikier than Crawl, since you don't gain any HP as you level, and some monsters (like Giants in Sil) are similar to Ettins in Crawl - huge damage, low accuracy. Except, if a giant hits you twice, it can mean you're dead. Not to mention critical hits. The other day I had the great experience of dying from half HP when two archers simultaneously scored a critical hit on me, when they had shot at me for 10 rounds before that for close to 0 damage. And Sil has completely transparent combat rolls.

I don't believe displaying numbers has anything to do with how wacky they can be. But to mix metaphors, this is a hobby horse horse of mps, and this horse has been beaten to death. I do not see any point in arguing about this in a DCA thread, since OP was talking about something else entirely (magic dart), which has many reasons to exist as it does.

[Sprucery]

I can't think of any other game where you could double player combat damage and it wouldn't be immediately, blindingly obvious to absolutely everyone, even without explicit damage numbers being available.

I can think of Nethack and ADoM.

[Sar]

I was talking to some person about Half-Life 2 another day and he was sure that it didn't have locational damage (basically, headshots) when it in fact does have locational damage (headshots) and they triple all the damage, too. He just never noticed it while playing it. I've also seen people being not aware that the first Half-Life has headshots (again, triple damage) after playing through the game. Like, maybe they played on Normal and there are a few weapons that one-shot certain enemy classes on Normal even without headshot, but those aren't always available. So, some people just don't pay attention, I guess.

[Hurkyl]

I hope you're trolling me about double damage. I can't think of any other game where you could double player combat damage and it wouldn't be immediately, blindingly obvious to absolutely everyone, even without explicit damage numbers being available.

It's the truth. People really overestimate the power of casual observation, especially about things they don't know to look for.

I'll admit I played quarter a game of SpEn at the time, and I did notice -- but only because I had numbers available. But then, I was still at the point where my best inference was that it was a combination of good look and that I was misremembering how much stabbing I needed to kill things, rather than inferring that I was actually doing more damage that game.

About clustering illusion, if your primary way to address discussion of probability is clustering illusion, that's a problem. I don't care how many times it's the right thing to say. A lot of times it isn't.

It's the right thing to say here. It might not be the right answer, but it is by far the best conclusion given the information we have, which is that one person claimed to be suspicious of the probability with no supporting numerical evidence.

[mps]

I was talking to some person about Half-Life 2 another day and he was sure that it didn't have locational damage (basically, headshots) when it in fact does have locational damage (headshots) and they triple all the damage, too. He just never noticed it while playing it. I've also seen people being not aware that the first Half-Life has headshots (again, triple damage) after playing through the game. Like, maybe they played on Normal and there are a few weapons that one-shot certain enemy classes on Normal even without headshot, but those aren't always available. So, some people just don't pay attention, I guess.

Before I tell you what I think about people who play Half-Life 2 and don't notice locational damage, I would first like to verify that the person you were talking to was not your grandmother.

[Sar]

I tried to make her play HL2 once but we didn't go far.

[archaeo]

I don't know which one of us is "right" mps, but I suspect we both would do well to stop coming on Tavern after we have too much coffee.

[Igxfl]

In the long run, you see luck dependence matter more with item drops than with the success rates of individual attacks. Conjurer and Enchanter can hold out longer with their starting equipment than most other backgrounds, even without god help. They're some of the least luck-dependent classes.

If, on the other hand, you care abiut early-game tactical luck, concerns like "You go down the stairs. Sigmund comes into view." matter a lot more than the precise success rate of EH.

Ultimately, Conjurer is going to get through most of the game on Mystic Blast, which is as reliable as attack spells get.

On Enchanter, it's Confuse and Tukima's you're really relying on. EH is for when you can get away with it.

[WingedEspeon]

\
Uh, except they're not actually hidden, since this is an open source game? It's not like dpeg guards the source code at his German castle or anything.

The devs aren't keeping anything a secret here. From what I understand of the reasoning, there's just some point in between "give players all information" and "say nothing about the game's mechanics" where they think you can find a happy medium where players have enough information to make decisions but not so much that crawl becomes less about playing the game and more about exhaustive minmaxing.

except it isn't a happy medium, It's just saying "Ha Ha, now you have to spend hours scouce diving to do your min maxing that you are going to do regardless of weather I show you the numbers." Hell, which are a lot harder to understand than how much damage something deals.

[savageorange]

\
Uh, except they're not actually hidden, since this is an open source game? It's not like dpeg guards the source code at his German castle or anything.

The devs aren't keeping anything a secret here. From what I understand of the reasoning, there's just some point in between "give players all information" and "say nothing about the game's mechanics" where they think you can find a happy medium where players have enough information to make decisions but not so much that crawl becomes less about playing the game and more about exhaustive minmaxing.

except it isn't a happy medium, It's just saying "Ha Ha, now you have to spend hours scouce diving to do your min maxing that you are going to do regardless of weather I show you the numbers." Hell, which are a lot harder to understand than how much damage something deals.

I'm pretty sure the correct response is, to reference Nethack: 'The punishment for pudding farming is pudding farming'.
The only reason that wouldn't hold true was if that information was something you actually needed, rather than something you thought you needed.

(yes, there is information that crawl reasonably should provide and doesn't. But exact damage numbers .. nope, not really on that list.)

[Hurkyl]

I'm pretty sure the correct response is, to reference Nethack: 'The punishment for pudding farming is pudding farming'.

Crawl has a different philosophy than nethack, and has eliminated a lot of instances where doing a horrible thing was better than not doing it. (e.g. once upon a time, Repel Missiles expired over time, and best play was almost surely to recast it every time it started expiring)

[byrel]

I think you misunderstand the reference. Pudding farming is a highly tedious theoretically endless method of farming a splitting monster (a pudding) which can, if done properly, result in having as many of any item you want, since they drop a random item 1/3 deaths. Pudding farming is clearly optimal, but insanely tedious. Some people claim it's cheating, but the general consensus is that anyone doing that is obviously already being punished quite enough by doing it.

Optimal but tedious exploits are their own punishment. Like old kill-outside-LOS summoners.

[Hurkyl]

I think you misunderstand the reference. Pudding farming is a highly tedious theoretically endless method of farming a splitting monster (a pudding) which can, if done properly, result in having as many of any item you want, since they drop a random item 1/3 deaths. Pudding farming is clearly optimal, but insanely tedious. Some people claim it's cheating, but the general consensus is that anyone doing that is obviously already being punished quite enough by doing it.

Optimal but tedious exploits are their own punishment. Like old kill-outside-LOS summoners.

A missing word probably made my post unclear -- my point is that the crawl devs have eliminated many optimal but tedious things. If crawl had pudding farming, the crawl devs would seek to eliminate it.

The thing the pudding farming philosophy fails to grasp is that some people (such as myself) find it quite unfun to deliberately sabotage their chances of winning for the sake of avoiding tedium.

[byrel]

OK, agreed. Which is why I play crawl instead of nethack.