what is the best engine for big material handicap?

[Ovyron]

Here's an example (3 0):

https://lichess.org/tournament/Zleaebtx

Notice how high rated players win most of their games, and upsets are rare. If this translated to the real world, I'd expect the strongest crazyhouse human player at 3 0 to beat everyone else in the world in most of their games. This is what I mean by the skill gap being much larger in crazyhouse than in normal chess.

Contrast that to this (4 2):

https://lichess.org/tournament/bEUNZinE

This one looks normal, what you'd expect to see in a chess tournament. It seems what I'm talking about doesn't apply if there's increment happening, so I hold increment is a must for crazyhouse for it to have any professional future. At least, nowadays the no increment tourney seems to have as much interest as the inc tourney, so over the years people have gotten used to it. This would break my claims so they'd only apply to players that do 3 0 or faster exclusively.

(to recapitulate: my claim is that in the 3 0 tourney, that 2300 player that won by a landslide would have won against the same people at 5 0 or 10 0, so if the winner is going to be the same it doesn't make sense to use longer time controls, and that's why they're so rare)

Do any other chess websites have specific Crazyhouse ratings?

The ICC, FICS and chess.com all have specific Crazyhouse ratings. There's also other places where you can play them at correspondence time controls (say, 2 days per move) and they have specific ratings, such as in brainking.com (which calls it "Loop Chess" because promoted pawns keep their rank when captured), Scheming Mind, or chessvariants.com (which also has "Chessgi" in where you can drop pawns on the first rank.)

[lkaufman]

So do any of the sites that have correspondence time controls for Crazyhouse (or one of these slight variants of it) have stats for White wins, draws, and losses? That might actually tell us something about the game.

[Ovyron]

Scheming Mind is a paid site so I can't check. Chessvariant has had only 10 games played since its inception. Brainking doesn't have statistics, but presumably games could be gathered from here to get them.

Please note that these correspondence games aren't of high quality, because it's not like someone has to move and then spends hours analyzing the positions to arrive at the best move and makes it, they just check the games where they have to move in and play moves in them (the point is then to be able to play people without having to be with them real time, not to spend time analyzing them.) Sorry if all this is useless

The most relevant things happening seem to be the Crazyhouse World Championship of lichess, and the $2,000 Crazyhouse Championship of chess.com, both played with 3 0 time control which seems to be the standard.

Something I just noticed is that I haven't found any evidence that the game has ever been played in actual chess boards, in person, in real life (or that equipment exists, this guy had to make his own board.) It seems Crazyhouse is an online thing.

High quality human games have yet to happen.

[lkaufman]

Scheming Mind is a paid site so I can't check. Chessvariant has had only 10 games played since its inception. Brainking doesn't have statistics, but presumably games could be gathered from here to get them.

Please note that these correspondence games aren't of high quality, because it's not like someone has to move and then spends hours analyzing the positions to arrive at the best move and makes it, they just check the games where they have to move in and play moves in them (the point is then to be able to play people without having to be with them real time, not to spend time analyzing them.) Sorry if all this is useless

The most relevant things happening seem to be the Crazyhouse World Championship of lichess, and the $2,000 Crazyhouse Championship of chess.com, both played with 3 0 time control which seems to be the standard.

Something I just noticed is that I haven't found any evidence that the game has ever been played in actual chess boards, in person, in real life (or that equipment exists, this guy had to make his own board.) It seems Crazyhouse is an online thing.

High quality human games have yet to happen.

I guess that the stats from the last stage or stages of the Crazyhouse World Championship, if available, would be the most useful human stats then. How about engine stats, limited to the top engines? Maybe engine play at blitz would give some idea as to how human GMs (if they existed in this game) would play with 2 hours per side. What is the strongest crazyhouse engine? Can the best human player ever win a game from it?

[Ovyron]

I guess that the stats from the last stage or stages of the Crazyhouse World Championship, if available, would be the most useful human stats then.

Here's the games between of the champion and second place of the world championship (my bad, it turns out they do use increment and games are 3 +2), so it's possible those are the strongest human zh players in the world (second place was also chess.com's world champion.) I don't know how to download them to get statistics.

If you filter lichess's database to top players you get this:

How about engine stats, limited to the top engines? Maybe engine play at blitz would give some idea as to how human GMs (if they existed in this game) would play with 2 hours per side. What is the strongest crazyhouse engine?

Crazyhouse hasn't been taken seriously by programmers. There were efforts made, and then one day Daniel Dugovic had the idea of adapting Stockfish to play 13 new variants, including crazyhouse, and his "MultiVariant" Stockfish was able to beat the second best engine 100% of the time (so imagine that in chess we only had Stockfish, the second best would be defeated 100% of the time by Stockfish, and the rest would be worse than this.)

Can the best human player ever win a game from it?

Doubtful. I don't think the best human player would be able to play stronger than Imortal v4.0, and Imortal already loses all games against the best engine, but nothing like this has been tried.

It'd be interesting to organize a time handicap match between MultiVariant Stockfish and world champion opperwezen (say, M Stockfish gets 1 minute and opperwezen gets 2 hours.)

Note that if M Stockfish is the strongest crazyhouse entity in the world, it doesn't have any problem winning from the black side.

[Guenther]

Crazyhouse hasn't been taken seriously by programmers. There were efforts made, and then one day Daniel Dugovic had the idea of adapting Stockfish to play 13 new variants, including crazyhouse, and his "MultiVariant" Stockfish was able to beat the second best engine 100% of the time (so imagine that in chess we only had Stockfish, the second best would be defeated 100% of the time by Stockfish, and the rest would be worse than this.)

This is quite outdated info. Since a year there is Crazy Ara (MCTS + NN) meanwhile.
I don't know exactly how strong it is compared to the current SF ZH version,
but of course it will not lose by such a margin. Probably they are even.

https://github.com/QueensGambit/CrazyAra
https://github.com/QueensGambit/CrazyAra/releases

Note that the XB/UCI chronology even has a section for Crazyhouse aka ZH since the end of 2016 IIRC.

https://docs.google.com/spreadsheets/d/ ... =338512849

[lkaufman]

I guess that the stats from the last stage or stages of the Crazyhouse World Championship, if available, would be the most useful human stats then.

Here's the games between of the champion and second place of the world championship (my bad, it turns out they do use increment and games are 3 +2), so it's possible those are the strongest human zh players in the world (second place was also chess.com's world champion.) I don't know how to download them to get statistics.

If you filter lichess's database to top players you get this:

How about engine stats, limited to the top engines? Maybe engine play at blitz would give some idea as to how human GMs (if they existed in this game) would play with 2 hours per side. What is the strongest crazyhouse engine?

Crazyhouse hasn't been taken seriously by programmers. There were efforts made, and then one day Daniel Dugovic had the idea of adapting Stockfish to play 13 new variants, including crazyhouse, and his "MultiVariant" Stockfish was able to beat the second best engine 100% of the time (so imagine that in chess we only had Stockfish, the second best would be defeated 100% of the time by Stockfish, and the rest would be worse than this.)

Can the best human player ever win a game from it?

Doubtful. I don't think the best human player would be able to play stronger than Imortal v4.0, and Imortal already loses all games against the best engine, but nothing like this has been tried.

It'd be interesting to organize a time handicap match between MultiVariant Stockfish and world champion opperwezen (say, M Stockfish gets 1 minute and opperwezen gets 2 hours.)

Note that if M Stockfish is the strongest crazyhouse entity in the world, it doesn't have any problem winning from the black side.

So based on those stats, Crazyhouse would be perfectly fair if you prohibit 1.e4 and 1.d4. Are similar stats available anywhere for engine vs engine games? As it is, I suspect that playing White in Crazyhouse is something like getting f2 pawn handicap or 3 moves in normal chess, although it's hard to compare since chess has a huge draw problem while Crazyhouse apparently does not. It is no surprise that M Stockfish can give other engines this handicap if it is something like a thousand elo stronger than they are! Stockfish and Komodo can give f2 pawn handicap or 3 moves in chess to 3000+ rated engines, no problem.

[EroSennin]

Here's an example (3 0):

https://lichess.org/tournament/Zleaebtx

Notice how high rated players win most of their games, and upsets are rare. If this translated to the real world, I'd expect the strongest crazyhouse human player at 3 0 to beat everyone else in the world in most of their games. This is what I mean by the skill gap being much larger in crazyhouse than in normal chess.

Contrast that to this (4 2):

https://lichess.org/tournament/bEUNZinE

This one looks normal, what you'd expect to see in a chess tournament. It seems what I'm talking about doesn't apply if there's increment happening, so I hold increment is a must for crazyhouse for it to have any professional future. At least, nowadays the no increment tourney seems to have as much interest as the inc tourney, so over the years people have gotten used to it. This would break my claims so they'd only apply to players that do 3 0 or faster exclusively.

(to recapitulate: my claim is that in the 3 0 tourney, that 2300 player that won by a landslide would have won against the same people at 5 0 or 10 0, so if the winner is going to be the same it doesn't make sense to use longer time controls, and that's why they're so rare)

Random assumptions. I don't see anything weird in the 3 0 tournament. After a few games some people lost and some won.

[Ovyron]

Are similar stats available anywhere for engine vs engine games?

Probably Ferdinand Mosca has them or can produce them on request.

it's hard to compare since chess has a huge draw problem while Crazyhouse apparently does not.

Yeah! That's why I propose switching games instead of trying to fix chess with armageddon or whatever. I still don't know if ZH is the right choice, if, say, one of the Billiards Chess variants didn't have the draw problem I'd support them over crazyhouse; but it's the most successful chess variant (other than Chess960), it just needs sponsorship.

It is no surprise that M Stockfish can give other engines this handicap if it is something like a thousand elo stronger than they are!

What is surprising is that the best programmers of the era that tried to make a crazyhouse engine couldn't produce anything stronger than the level of... Yace 0.99.87 of chess (imagine that before Stockfish appeared the best engine around was Yace.) Before M Stockfish appeared I assumed the state of the art of zh engines was much higher.

Random assumptions. I don't see anything weird in the 3 0 tournament.

It's just games being decided in favor of the losing player because the one that was winning didn't have enough time to make the moves, because the low time on the clock didn't let them convert the win so they blundered a won position and lost, or random upsets because games are decided at random on clock struggles (not unlike bullet chess.) It seems the long skill gap is widened because being good at crazyhouse isn't enough, you also have to have outstanding clock management to compete at the top level (...huh, maybe normal chess is like this as well, so 3 0 time control is the problem. if this is the case I could see why you don't see anything weird, as you wouldn't see anything "weird" in a bullet tournament either).

Even lichess saw the problem and added the 2 second increment on the world championship.

[todd]

Something I just noticed is that I haven't found any evidence that the game has ever been played in actual chess boards, in person, in real life (or that equipment exists, this guy had to make his own board.) It seems Crazyhouse is an online thing.

Crazyhouse is definitely mostly an online thing.

It has been played in person, though. I've played quite a few games myself.

On November 30, 2012, a Chicago-area chess club (South Suburban Chess Club in Oak Lawn) held an over the board crazyhouse tournament. Around a dozen people (including me) participated.