what is the best engine for big material handicap?

Discussion of anything and everything relating to chess playing software and machines.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
User avatar
Nordlandia
Posts: 2406
Joined: Fri Sep 25, 2015 7:38 pm
Location: Sortland, Norway

Re: what is the best engine for big material handicap?

Post by Nordlandia » Wed Oct 09, 2019 2:32 am

Crazyhouse is the most played variant on lichess. About 1% of all Crazyhouse games compared to classic. In other words CH take up 1 percent of all classical games. Despite this, the lichess staff don't prioritize adding more variants, saying people don't play variants in large scales. This is very true. Many want more variants available at popular chess websites like S-Chess (Sharper Chess) among others.

PK
Posts: 820
Joined: Mon Jan 15, 2007 10:23 am
Location: Warsza
Contact:

Re: what is the best engine for big material handicap?

Post by PK » Wed Oct 09, 2019 6:29 am

Handicap go is still go. Handicap shogi is still shogi. Hearts of Iron on "civilian" settings with several buffs is still Hearts of Iron. I don't buy the idea that handicap chess is not chess. Game mechanics and rules remain the same, so it is the same game.

Otherwise you would have to argue that chess puzzles are not chess puzzles, since they change starting position. What's more, tactics books are not about chess. Endgame books are not about chess etc.

User avatar
Ovyron
Posts: 2495
Joined: Tue Jul 03, 2007 2:30 am

Re: what is the best engine for big material handicap?

Post by Ovyron » Wed Oct 09, 2019 6:39 am

One thing about Crazyhouse that I've noticed, is that, if you'e going to play a blunder, you're going to play it, and more time doesn't help.

The variant lives to its name, if there's some crazy variation where your opponent sacrifices their queen then Drops a pawn in a weird place and somehow no matter what you do from there, they got some insane forced mate, if you can't see this on 3 0 time control, you wouldn't see it on 5 0, or on 10 0.

I think this is the main reason people don't bother with slower time controls, not even engines can see the intricacies of some positions, and will insist one side is completely winning when it's the other one.

Hmmm, isn't this like Go? A position is very hard to evaluate, so whatever evaluation is at the leaf node of your tree, the position from where your score comes from, could be completely wrong. There's no such thing as a quiet position, because what looks quiet could be a mate in something (not surprising if white has a forced mate from the opening position!). It seems this is a perfect game for a NN to destroy.

Personally, I quit Crazyhouse because I don't feel satisfaction about winning on time, and I don't like the stress of having to move very fast. I'd want 10 0 time control, at least, but then most of the time on the clock is wasted because games are decided very soon because a single pattern someone has that the other doesn't is enough to decide a game.

Looks like low time + some increment (say, 1 minute +7 seconds increment) would be ideal, but the people that tend to enjoy Crazyhouse also tend to enjoy the adrenaline rush of having to move fast at the end of the game (the time struggle is a core part of close games to them) so tournaments with increment fail to get any traction.

Chess960 has been getting a lot of attention lately, perhaps one day Crazyhouse will have its moment.
PK wrote:
Wed Oct 09, 2019 6:29 am
Handicap go is still go. Handicap shogi is still shogi. Hearts of Iron on "civilian" settings with several buffs is still Hearts of Iron. I don't buy the idea that handicap chess is not chess. Game mechanics and rules remain the same, so it is the same game.

Otherwise you would have to argue that chess puzzles are not chess puzzles, since they change starting position. What's more, tactics books are not about chess. Endgame books are not about chess etc.
So, if I invite you to a chess tournament and you register, and then you shake the hands of your opponent and the game starts, wouldn't you be shocked if the TD came to your board and set up some random chess puzzle from where you have to play, and are informed that the next 19 rounds would also start from a random chess puzzle position?

Or they take a rook away from you and say all the rest of the tournament you'll have to play with that rook missing from the start?

Or your opponent drops a knight all of nowhere to checkmate you?

You have to draw the line somewhere, at the point you would complain to me that I should have informed you about the abnormal conditions of the tournament.

Then, wherever you draw the line, when you think you should be informed about the abnormality before registering, that's not chess anymore.
Great spirits have always encountered violent opposition from mediocre minds.

PK
Posts: 820
Joined: Mon Jan 15, 2007 10:23 am
Location: Warsza
Contact:

Re: what is the best engine for big material handicap?

Post by PK » Wed Oct 09, 2019 6:46 am

So, if I invite you to a chess tournament and you register, and then you shake the hands of your opponent and the game starts, wouldn't you be shocked if the TD came to your board and set up some random chess puzzle from where you have to play, and are informed that the next 19 rounds would also start from a random position?
If it is specified in the tournament rules, then why not. I have played in many go tournaments, and before entering I always had information like "no handicap, komi 6,5" or "no handicap, komi 7,5" or "normal hadicap" or "hadicap -2" or even "board 9x9". And whereas I know many people who dislike handicap games, and I am not particularily fond of 9x9 go, I would not say it is not the same game.

User avatar
Ovyron
Posts: 2495
Joined: Tue Jul 03, 2007 2:30 am

Re: what is the best engine for big material handicap?

Post by Ovyron » Wed Oct 09, 2019 7:10 am

PK wrote:
Wed Oct 09, 2019 6:46 am
before entering I always had information like "no handicap, komi 6,5" or "no handicap, komi 7,5" or "normal hadicap" or "hadicap -2" or even "board 9x9".
Note even they agree with me with "this is a handicap and this is not", or they wouldn't use the word "normal". There's a definition of a "normal handicap", you know a priori what they're referring to, any other kind of handicap is not a "normal handicap" and the difference needs to be specified.

All the times I've been talking about chess on this thread, I've been talking about "normal chess."

Where we disagree is that I claim if it's not "normal chess", then it's not chess, while you say it's still chess. But you can pretend in all my arguments I added the word "normal" before chess, and they'd make sense.

Now, can we agree a tournament where players have to start from random chess puzzle positions isn't "normal chess" in the same way a handicap of -2 isn't a "normal handicap"?
Great spirits have always encountered violent opposition from mediocre minds.

PK
Posts: 820
Joined: Mon Jan 15, 2007 10:23 am
Location: Warsza
Contact:

Re: what is the best engine for big material handicap?

Post by PK » Wed Oct 09, 2019 7:52 am

OK, this makes sense. For me chess per se ends where different movement rules are introduced, and Your "normal chess"
may well end sooner.

lkaufman
Posts: 3676
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Re: what is the best engine for big material handicap?

Post by lkaufman » Thu Oct 10, 2019 4:21 am

Nordlandia wrote:
Wed Oct 09, 2019 2:32 am
Crazyhouse is the most played variant on lichess. About 1% of all Crazyhouse games compared to classic. In other words CH take up 1 percent of all classical games. Despite this, the lichess staff don't prioritize adding more variants, saying people don't play variants in large scales. This is very true. Many want more variants available at popular chess websites like S-Chess (Sharper Chess) among others.
What is the draw percentage? I'm guessing it's like in shogi, between 1 and 2%. But maybe it's higher in engine games? Any data on this?

I'm certainly no expert on Crazyhouse, but if it is anything like shogi, the notion that you can't play better with more time is absurd. I would imagine that the quality of games between strong players at say one hour per side would be many hundreds of elo (maybe even thousands?) higher than at say 3' per side. If someone starting offering $100,000 prizes for over-the-board Crazyhouse tournaments at long time limits, there would soon be grandmasters of that game, massive theory would develop, and probably the White win percentage between these GMs would reach the point where special rules would need to be introduced, like prohibiting 1.e4, or giving Black double the time, or ... who knows? It would probably be like GO, where the second player gets 6.5 or 7.5 points added at the end to make the game fair. Or maybe a better analogy is Renju (go-moku, 5 in a row) where they do prohibit the best opening moves to make for an even game. Note that shogi doesn't have this problem, because there are no queens, nothing stronger than the rook in the start position, so despite the drops there are few short games and the first player advantage isn't so overwhelming. None of this is a criticism of Crazyhouse, it's just a forecast of what may happen if it catches on big time.
Komodo rules!

User avatar
Nordlandia
Posts: 2406
Joined: Fri Sep 25, 2015 7:38 pm
Location: Sortland, Norway

Re: what is the best engine for big material handicap?

Post by Nordlandia » Thu Oct 10, 2019 5:57 am

According lichess database, i can't see the actual draw percentage for crazyhouse so it has to be 1% +/-

It it very pity that money has to be involved to force GMs to play and hopefully popularize trendy variants. Carlsen and Nakamura seems to have some magical spells over them, if they play something new, many of the fans will soon follow and maybe it may catch on. But again it seems like they avoid doing so on live streams because of this.

Capablanca tried it in the 1920s but failed big time. Yasser Seirawan made Sharper-Chess that limit the game to standard 8x8 board but again, no special attention is given to that unless it's implmented into a fairly populated chess website. There is few website available at this point that offers many variants but not many play there.

For example pychess (it supports shogi) https://pychess-variants.herokuapp.com/ and green chess - https://greenchess.net/index.php

User avatar
Ovyron
Posts: 2495
Joined: Tue Jul 03, 2007 2:30 am

Re: what is the best engine for big material handicap?

Post by Ovyron » Thu Oct 10, 2019 2:52 pm

lkaufman wrote:
Thu Oct 10, 2019 4:21 am
I'm certainly no expert on Crazyhouse, but if it is anything like shogi, the notion that you can't play better with more time is absurd.
That's not what I mean. What I have observed is that in Crazyhouse the difference in skill is enormous. Like, in chess, you can pair a 1500 player with a 1530 player, and you can expect the 1500 one to win, because they're not too far apart, basically anything can happen. In Crazyhouse, a 1530 player is going to beat the 1500 most of the time, and draws will be rare. There's something that makes it higher rated, and the 1530 guy can exploit that almost every game.

The 1530 rated player's rating does not explode because if it gets paired against someone with 1560 rating, she will give them a beating and keep their rating in check. If someone is unlucky or lucky and gets paired against much weaker and stronger players, their rating would get distorted, so someone can have 1530 rating and actually be 1470 (they were lucky to get lots of weaker players) and then the 1500 guy is going to beat them badly, but these observations would hold if the ratings were produced against about the same amount of weaker/stronger opponents.

If my observations about this are correct, then what I'd be saying is that if the 1530 rated player is going to beat the 1500 rated one, it's going to beat them almost as badly at 3 0, 5 0 or 10 0 time controls, so my theory is that the reason Crazyhouse never picked up with slower time controls is that the strongest player is most likely to win regardless of time control and then you can play 3 times the games so why bother with 10 0?

What has happened in lichess is that when a player is paired against a weaker player, they know they're going to win, so they "Berzerk" and cut their time in half, because if they win with half the time they'll get bonus points and are more likely to win the tournament. Since the weaker player knows they're going to lose, there's nothing to lose, so they might cut their time in half themselves in case they manage to win, so 5 0 crazyhouse becomes effectively 2.5 0 crazyhouse.

10 0 is the slowest time control I've seen for Crazyhouse, it doesn't seem like the people currently interested in crazyhouse are interested about sitting for hours and think about the positions, and I suspect it's because the decided games at 2 hours time controls could have happened at 3 0 (they key position would have been resolved the same way.) But probably a new generation of crazyhouse players could be educated about thinking more about the positions they play, it seems current games are played on intuition (they don't see if an attack would work till the end, they just see the beginning of an attack and play it, if their opponent is weaker they probably won't see the defense either.)

I wonder if the difference in skill could be reflected on the ratings, maybe multiplying them by 6 or something (so a 1500 v 1530 match looks more like a 9000 v 9180 rated one, then the expected 73.8% performance by the stronger player would be closer to the truth.)
Great spirits have always encountered violent opposition from mediocre minds.

lkaufman
Posts: 3676
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Re: what is the best engine for big material handicap?

Post by lkaufman » Thu Oct 10, 2019 7:02 pm

Ovyron wrote:
Thu Oct 10, 2019 2:52 pm
lkaufman wrote:
Thu Oct 10, 2019 4:21 am
I'm certainly no expert on Crazyhouse, but if it is anything like shogi, the notion that you can't play better with more time is absurd.
That's not what I mean. What I have observed is that in Crazyhouse the difference in skill is enormous. Like, in chess, you can pair a 1500 player with a 1530 player, and you can expect the 1500 one to win, because they're not too far apart, basically anything can happen. In Crazyhouse, a 1530 player is going to beat the 1500 most of the time, and draws will be rare. There's something that makes it higher rated, and the 1530 guy can exploit that almost every game.

The 1530 rated player's rating does not explode because if it gets paired against someone with 1560 rating, she will give them a beating and keep their rating in check. If someone is unlucky or lucky and gets paired against much weaker and stronger players, their rating would get distorted, so someone can have 1530 rating and actually be 1470 (they were lucky to get lots of weaker players) and then the 1500 guy is going to beat them badly, but these observations would hold if the ratings were produced against about the same amount of weaker/stronger opponents.

If my observations about this are correct, then what I'd be saying is that if the 1530 rated player is going to beat the 1500 rated one, it's going to beat them almost as badly at 3 0, 5 0 or 10 0 time controls, so my theory is that the reason Crazyhouse never picked up with slower time controls is that the strongest player is most likely to win regardless of time control and then you can play 3 times the games so why bother with 10 0?

What has happened in lichess is that when a player is paired against a weaker player, they know they're going to win, so they "Berzerk" and cut their time in half, because if they win with half the time they'll get bonus points and are more likely to win the tournament. Since the weaker player knows they're going to lose, there's nothing to lose, so they might cut their time in half themselves in case they manage to win, so 5 0 crazyhouse becomes effectively 2.5 0 crazyhouse.

10 0 is the slowest time control I've seen for Crazyhouse, it doesn't seem like the people currently interested in crazyhouse are interested about sitting for hours and think about the positions, and I suspect it's because the decided games at 2 hours time controls could have happened at 3 0 (they key position would have been resolved the same way.) But probably a new generation of crazyhouse players could be educated about thinking more about the positions they play, it seems current games are played on intuition (they don't see if an attack would work till the end, they just see the beginning of an attack and play it, if their opponent is weaker they probably won't see the defense either.)

I wonder if the difference in skill could be reflected on the ratings, maybe multiplying them by 6 or something (so a 1500 v 1530 match looks more like a 9000 v 9180 rated one, then the expected 73.8% performance by the stronger player would be closer to the truth.)
It sounds from your comments like lichess doesn't handle ratings of newcomers properly. The only way you would get a situation where established players would be just thirty elo above other established players despite a 70% + win percentage is if you have a huge influx of new players all the time starting at 1500 whose strength may be anything, and the games are rated as if the 1500 rating was already real. If they are not doing this, then I can't see how what you describe could happen. I was chairman of USCF ratings committe for many years, so I know something about ratings. If ratings are done properly, win percentages should lead to proper rating differences once players have played enough games. It certainly works that way for normal chess on the servers.
Do any other chess websites have specific Crazyhouse ratings?
Komodo rules!

Post Reply