How reliably can an experienced player win?

[axiom]

What is an experienced/skilled player's success rate with say MiBe?

What about with a more average species/class like Human or something?

DCSS definitely has plenty of skill and luck, but how deep does it go?

[VeryAngryFelid]

As long as we have banishment/paralysis, luck is always required. Lack of bad lack more precisely.

[papilio]

D:1~5 is the hardest hurdle in this game.

Ijyb with wand of acid, D:1 adder, D:1 gnoll with +2 halberd of venom would be source of unavoidable deaths.

Beyond D:10, most death comes from mistakes and overconfidence.

[Siegurt]

D:1~5 is the hardest hurdle in this game.

Ijyb with wand of acid, D:1 adder, D:1 gnoll with +2 halberd of venom would be source of unavoidable deaths.

Beyond D:10, most death comes from mistakes and overconfidence.

Adders and Gnolls with weapons can't generate on D:1 in the first 700 turns (By which point an experienced player should be at D:2 or later 100% of the time)

Ijyb also can't generate on D:1, I don't know if him with a wand of acid is *unavoidable* death (Certainly if you're level 3 or 4 by the time you encounter him, you'll often have enough hps that you can take some positive action), but it's fairly likely that you won't have the consumables to properly deal with such a situation, so the only way to avoid a death from that is to explore *very very* carefully, and tediously, in a way that never leaves you vulnerable and without retreat options, and so that you will encounter anything nasty at maximum range, with maximum possible cover.

That being said, the first few levels of the game *are* where you have the least options, and the least control over how you deal with dangerous situations, so it's where an experience player will feel the most threatened.

A very experienced player will not reliably win most of the time, because an experienced player would get very bored with playing an easy combo in the least risky way over and over, how reliably they *could* win is therefore pretty hard to estimate.

[CanOfWorms]

gnolls with weapons can absolutely generate on D:1 immediately (are you maybe thinking of branded weapons?)

[Siegurt]

gnolls with weapons can absolutely generate on D:1 immediately (are you maybe thinking of branded weapons?)

Sorry, yes, i did mean branded weapons, although it is plausible that i am mistaken about that as well. I remember something gnoll related, but i could easily be misremembering what it was.

[Implojin]

Adders and Gnolls with weapons can't generate on D:1 in the first 700 turns (By which point an experienced player should be at D:2 or later 100% of the time)

https://crawl.jorgrun.rocks/morgue/hyperimplojin/morgue-hyperimplojin-20161119-073703.txt


edit:
In answer to the OP's question, DCSS winrate for the top veteran players is something very close to 100% when choosing their combos to play to racial strengths and playing to win. All you need do is look at any of the huge streaks to verify this.

Looking at top veteran hyperaccount winrates (for randomly chosen combos), expected winrate seems to be just shy of 70%. Those accounts are mostly older though, with no guarantee that the players piloting them were playing with winning as their primary goal -- and on top of that, the community standard of play has improved in the years since people started playing hyper accounts. It's probable that expected hyper winrate for optimal play is higher than 70%.

Needless to say, almost nobody makes a habit of trying to play optimally. Abusing all of the engine quirks to do so is unbelievably tedious.

[tasonir]

Pretty damned near 100%, as is demonstrated by players who have streaks over 20 games long. Myself I tend to lose focus and play fast and lazy, so my longest streak is 4, and will probably not ever get up to 10+ games long. But if you're both very experienced and very slow and methodical with a strong resistance to tedium, a win rate of 90% or higher is doable. I'll never be there, though I hold down tab way too much.

[Sprucery]

in the first 700 turns (By which point an experienced player should be at D:2 or later 100% of the time)

Is this really true? I am an experienced player. I know I am a slow player too, but why would I be much slower on D:1 than others? I use autoexplore and always clear D:1 fully. I checked the notes of my last 20 winners and 16 of them seemed to have been on D:1 more than 700 turns. The number is the biggest turn count found from D:1 before any notes on D:2:

  Code:

SpAE (722)
HaCK (1528)
HuSu (1455)
FeNe (1362)
DgEn (914)
FoGl (1159)
DrEE (1283)
MiSk (1131)
MfBe (1307)
CeWn (1031)
OgGl (1061)
SpWz (1376)
HaWr (978)
VpMo (1098)
GhGl (867)
MuVM (1635)

The 4 characters that definitely went to D:2 before turn 700 were DDAr, OpGl, GrFi, and TrMo. I use explore_wall_bias=20 but I'd think that it doesn't make autoexplore considerably slower in early D where there are usually lots of corridors and not much big open spaces. Maybe someone could find out from sequell how many turns 'experienced' players usually spend on D:1?

[stickyfingers]

Just yesterday I had a game with an adder being the first enemy I encountered (next to the entry room even), and a gnoll was second. Also my D:1 takes more than 1000 turns usually, due to no upstairs level to rest on.

[sanka]

I think that when experienced players want to streak they do not use autoexplore on D1.

[Sprucery]

I think that when experienced players want to streak they do not use autoexplore on D1.

Right, I guess it must be autoexplore that explains it. Maybe I should try once again doing D:1 without it :) But it would be so boring, I don't know if I could do it...

[sanka]

Well, I usually use autoexplore, but I rarely play even remotely well.

[papilio]

I always wondered why the wrong theory of "there are no D:1 adders til 700 turns" became widespread.

Image link: http://i.imgur.com/RCTifwo.png

Perhaps, there were some unrecognized patches (or unintended bugs) and outdated chaosforge wiki has not been fixed..

[tabstorm]

high 90%s if you play reasonable combos, as in, don't play things like DEFi, TeEE, MuAM, FoAE and so on.

[VeryAngryFelid]

By ignoring dodging. If you don't trust me, check the longest streak of the last tournament

[axiom]

By ignoring dodging. If you don't trust me, check the longest streak of the last tournament

Where do I find all this tournament data?

[VeryAngryFelid]

http://dobrazupa.org/tournament/0.19/overview.html

Edit. Actually I was joking, Dodging is very useful. But it is still amazing to see a player did a 15-win streak while ignoring Dodging in almost all of those games.

[njvack]

Wow. Aside from Fighting, this HuCj basically ignored defenses altogether.

Sometimes I think I don't know how Crawl works at all.

[axiom]

What is that software that everyone uses to run simulations to test out numbers/stats?

I'm starting to think that any sort of conventional knowledge or rules of thumb that people go by are worthless (or just more inaccurate than people realize). These top players clearly have a deeper understanding of the mechanics. I'm going to have to just rely on my own calculations from now on, which actually seems pretty cool because RPG type games are a lot more skillful when there aren't guides to tell you what is optimal.

[Tedronai]

At the 700 turn mark you better be prepared for water moccasins and orc priests on D:1. Puny adders should be the least of your concerns

[Lasty]

Almost all of the rules of thumb and detailed questions about how to min/max are for players whose grasp of the fundamentals (threat evaluation, positioning, noise, consumables use, exploration patterns) are weak or are intentionally playing sloppy. Good fundamentals are something like 90% of skill. Almost nothing else is a significant contributor.

"That player won an unlikely combo by doing something most players consider bad" isn't particularly surprising, considering the above.

[VeryAngryFelid]

That player won an unlikely combo by doing something most players consider bad" isn't particularly surprising, considering the above.

I think it has more to do with character power. I am sure many experienced players can win many powerful combos with self-imposed conducts, not training Dodging is one of them, it's just increasing difficulty level after decreasing it via picking an easy combo/god. Similarly players get 3 hell runes first without entering Lair/Orc/Vaults, nobody is doing it with FeCK, all pick OP combos like MiBe. That player did train Dodging for characters who were weak otherwise, I hardly can imagine winning FeBe with no Dodging, for example.

[axiom]

That player did train Dodging for characters who were weak otherwise, I hardly can imagine winning FeBe with no Dodging, for example.

Well of course, but my point is that these "unusual" skill training choices are indicative of a deeper understanding of the mechanics that goes beyond any sort of simplistic heuristics like "dodging is good", "dodging is bad", "AC is better than EV" etc.

[Sprucery]

I hardly can imagine winning FeBe with no Dodging, for example.

I'm sure that you could pull that off, it can't be worse than streaking mummies of Chei :)

[stickyfingers]

At the 700 turn mark you better be prepared for water moccasins and orc priests on D:1. Puny adders should be the least of your concerns

Now this is just ridiculous.

[dowan]

What is that software that everyone uses to run simulations to test out numbers/stats?

I'm starting to think that any sort of conventional knowledge or rules of thumb that people go by are worthless (or just more inaccurate than people realize). These top players clearly have a deeper understanding of the mechanics. I'm going to have to just rely on my own calculations from now on, which actually seems pretty cool because RPG type games are a lot more skillful when there aren't guides to tell you what is optimal.

Once in a while these top players will come here and talk about their strategies (Bart comes to mind). Then they get told that although they are obviously good at winning, they're still wrong because they go against that conventional knowledge and those rules of thumb. Then they just stop coming back. (I am not one of those top players, by the way)

So, if you have no idea what you're doing, come here, learn a little, and you'll probably do better. But like you said, you have to rely on your own experience and what makes sense to you, because that is what those really good players are doing. A part of that process is finding out where those rules of thumb are wrong, or not applicable.

An interesting note: There is a large segment of the playerbase from Korea. They have their own ideas about what is good and bad. In their community, Chei is considered a top god, overpowered even. Here, chei is considered worse than no god at all. Both are right, from their own point of view. Just realize that all those rules of thumb are also good from a certain point of view, and you generally have to figure that out for yourself.

By the way, the tool you're asking about is called FSIM, you can use it in wizard mode in the offline version. There's even a handy little text file telling you how to use it in the download package.

[axiom]

What is that software that everyone uses to run simulations to test out numbers/stats?

I'm starting to think that any sort of conventional knowledge or rules of thumb that people go by are worthless (or just more inaccurate than people realize). These top players clearly have a deeper understanding of the mechanics. I'm going to have to just rely on my own calculations from now on, which actually seems pretty cool because RPG type games are a lot more skillful when there aren't guides to tell you what is optimal.

Once in a while these top players will come here and talk about their strategies (Bart comes to mind). Then they get told that although they are obviously good at winning, they're still wrong because they go against that conventional knowledge and those rules of thumb. Then they just stop coming back. (I am not one of those top players, by the way)

So, if you have no idea what you're doing, come here, learn a little, and you'll probably do better. But like you said, you have to rely on your own experience and what makes sense to you, because that is what those really good players are doing. A part of that process is finding out where those rules of thumb are wrong, or not applicable.

An interesting note: There is a large segment of the playerbase from Korea. They have their own ideas about what is good and bad. In their community, Chei is considered a top god, overpowered even. Here, chei is considered worse than no god at all. Both are right, from their own point of view. Just realize that all those rules of thumb are also good from a certain point of view, and you generally have to figure that out for yourself.

By the way, the tool you're asking about is called FSIM, you can use it in wizard mode in the offline version. There's even a handy little text file telling you how to use it in the download package.

Thanks for this man. Super helpful.

Once in a while these top players will come here and talk about their strategies (Bart comes to mind). Then they get told that although they are obviously good at winning, they're still wrong because they go against that conventional knowledge and those rules of thumb. Then they just stop coming back.

LOL

Well it's cool to hear that there is a lot left to discover. That definitely makes me excited to start trying stuff out in FSIM. Of course I still need to learn a lot of fundamentals first.

[Gerad]

That player won an unlikely combo by doing something most players consider bad" isn't particularly surprising, considering the above.

I think it has more to do with character power. I am sure many experienced players can win many powerful combos with self-imposed conducts, not training Dodging is one of them, it's just increasing difficulty level after decreasing it via picking an easy combo/god. Similarly players get 3 hell runes first without entering Lair/Orc/Vaults, nobody is doing it with FeCK, all pick OP combos like MiBe. That player did train Dodging for characters who were weak otherwise, I hardly can imagine winning FeBe with no Dodging, for example.

Does GhFi count as overpowered?

[papilio]

Does GhFi count as overpowered?

cosmonaut got his avarice 3 + lord of darkness 3 with GhWr, too! Ghouls are bit underestimated in tavern, as they are greatly buffed in 0.19 with chunk healing.

[duvessa]

Once in a while these top players will come here and talk about their strategies (Bart comes to mind). Then they get told that although they are obviously good at winning, they're still wrong because they go against that conventional knowledge and those rules of thumb. Then they just stop coming back. (I am not one of those top players, by the way)
[...] In their community, Chei is considered a top god, overpowered even. Here, chei is considered worse than no god at all.

I'm pretty sure none of this is true.

[papilio]

Rating Chei depends on what kind of standards do you think.

If you think "I want to have longest winning streaks, and I want to max out my winning probs of any combos I pick" then Chei is terrible.
If you think "I want character who become eventually OP enough to steamroll everything" or "I want turncount speedrun" then Chei is OP.

At least, tavern's atmosphere on Chei that "Chei is f-in useless and worshipping no god is better than Chei" is completely far from the truth.

http://dobrazupa.org/tournament/0.19/species-backgrounds.html
http://dobrazupa.org/tournament/0.18/species-backgrounds.html
http://dobrazupa.org/tournament/0.17/species-backgrounds.html
http://dobrazupa.org/tournament/0.16/species-backgrounds.html

[VeryAngryFelid]

Well of course, but my point is that these "unusual" skill training choices are indicative of a deeper understanding of the mechanics that goes beyond any sort of simplistic heuristics like "dodging is good", "dodging is bad", "AC is better than EV" etc.

Not really. It is kind of protest of the player. Similarly you can rarely see me using sack of spiders but it does not mean that I have deep understanding of uselessness of the item

[VeryAngryFelid]

Does GhFi count as overpowered?

I am not sure if you intended it as joke but when I wanted a 15-rune win during my first tournament, guess which species I picked?
I even selected the hardest non-CK background for more points because Gh has very easy early game (not Tr of course, but much better than Gr for example).

Spoiler: show

GhAE http://dobrazupa.org/tournament/0.16/pl ... man25.html

Edit. Or maybe it was for RMsl, I don't remember now. One thing I am sure is that I wasn't going to stay on conjurer path to take advantage of that extra Int

[WalkerBoh]

In answer to the OP's question, DCSS winrate for the top veteran players is something very close to 100% when choosing their combos to play to racial strengths and playing to win. All you need do is look at any of the huge streaks to verify this.

Looking at top veteran hyperaccount winrates (for randomly chosen combos), expected winrate seems to be just shy of 70%. Those accounts are mostly older though, with no guarantee that the players piloting them were playing with winning as their primary goal -- and on top of that, the community standard of play has improved in the years since people started playing hyper accounts. It's probable that expected hyper winrate for optimal play is higher than 70%.
Needless to say, almost nobody makes a habit of trying to play optimally. Abusing all of the engine quirks to do so is unbelievably tedious.

This is the best answer in this thread. I consider myself an experienced player, and I managed a 67% win rate (12 wins in 18 games) in this last tournament with my hyper account (I also more or less did a 'take first god seen' conduct, which made it slightly harder than otherwise). Unless you have a really crap char or get very unlucky on D:1, the biggest test for most characters is if they can beat the first adder they see on D:2, and to a lesser extent the first orc wizard/priest they see on D:3. After that, if you don't make big mistakes and you play carefully, you should be able to win 90%+ of your games.

For strong characters that can deal with early game threats reliably, I would expect 95+% win-rate for top players - it takes some significantly dangerous or unusual encounter or particularly bad luck for you to die. For average characters, it's a lot more likely that you don't find !curing in time, or you have some bad rolls against an adder or so, so I expect a win rate closer to 70%. For bad characters, it will probably drop below 50%, and down from there for the worst characters. I think Implojin's estimate of 70% is probably about right. This is just based on my own personal experience.

[TonberryJam]

I autoexplore b/c i live dangerously.

[lethediver]

High winrate is more indicative of personality/priorities than experience imo.

Winning dcss reliably is boring to a large percentage of the gaming population, because the safest strategies tend to be the most tedious and the game is already too long as is. Overcoming this requires either extreme dedication or the rare ability to enjoy constant, methodical, painstaking luring/stairdancing/manual pathing/fight resetting for 50+ floors.

Experience is only the first hurdle.

[Magipi]

The original question was about the winrate of "experienced players" in general, but the responses were about the winrate of a tiny group of top players. There is a world of difference between those.

The correct answer to the original question is: "10 percent or less".

[Sprucery]

Yeah, the title in the OP is "How reliably can an experienced player win?" but then in the message part "success rate" is mentioned. These are two different things. For example, my overall winrate in Crawl is about 7%. As of DCSS 0.10, it is 17%. If I hadn't played any Octopodes or Felids it would be clearly more than 20%. I'm pretty sure that if I would only play strong combos and would be willing to play as tediously optimally as I could, even I could win more than every second game.

There is a big difference between "How reliably can an experienced player win" and "How reliably do experienced players generally win".