What is the best engine to predict human moves

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

Moderator: Ras

Uri Blass
Posts: 10790
Joined: Thu Mar 09, 2006 12:37 am
Location: Tel-Aviv Israel

What is the best engine to predict human moves

Post by Uri Blass »

For example level 2000-2100 fide rating list at classical time control.
The target is to guess correctly the move of the human player based on a lot of games.

Is there a data about different chess engines to know what is the percentage of human moves they get correctly?
jefk
Posts: 941
Joined: Sun Jul 25, 2010 10:07 pm
Location: the Netherlands
Full name: Jef Kaan

Re: What is the best engine to predict human moves

Post by jefk »

it used to be Rodent, i think.

But nowadays probably Fritz20 (in combination with Chessbase18)

https://www.perplexity.ai/search/with-c ... pfaID71Q#0

although i haven't tried it myself (did try getting some style reports with CB18 for certain players and
that worked fine, sometimes a bit slow, depending on nr of games (then leave out blitz).
As for the infamous fritz, well i still 'only' have F19. Sure CB isn't that bad, they provide regular
updates, both for CB18 and still F19 as well; but i would expect with normal software procedures
that with a major update i then would get F20 (as major update). Not so, apparently there's a new
Fritzie every three months and then they ask me to pay 60 bucks again every time? urgh.
:evil:
Lunar
Posts: 12
Joined: Wed May 21, 2025 12:32 pm
Full name: Patrick Hilhorst

Re: What is the best engine to predict human moves

Post by Lunar »

I think Maia is an interesting development: https://www.maiachess.com/

Explicitly trained on multiple Elo ranges of players
Uri Blass
Posts: 10790
Joined: Thu Mar 09, 2006 12:37 am
Location: Tel-Aviv Israel

Re: What is the best engine to predict human moves

Post by Uri Blass »

Lunar wrote: Mon Jun 16, 2025 6:46 pm I think Maia is an interesting development: https://www.maiachess.com/

Explicitly trained on multiple Elo ranges of players
Unfortunately it seems Maia is limited to 1900 and does not try to predict mistakes of stronger human players.

I am not sure but I guess it is also based on a lot of blitz games.
carldaman
Posts: 2284
Joined: Sat Jun 02, 2012 2:13 am

Re: What is the best engine to predict human moves

Post by carldaman »

jefk wrote: Mon Jun 16, 2025 12:30 pm it used to be Rodent, i think.

But nowadays probably Fritz20 (in combination with Chessbase18)

https://www.perplexity.ai/search/with-c ... pfaID71Q#0

although i haven't tried it myself (did try getting some style reports with CB18 for certain players and
that worked fine, sometimes a bit slow, depending on nr of games (then leave out blitz).
As for the infamous fritz, well i still 'only' have F19. Sure CB isn't that bad, they provide regular
updates, both for CB18 and still F19 as well; but i would expect with normal software procedures
that with a major update i then would get F20 (as major update). Not so, apparently there's a new
Fritzie every three months and then they ask me to pay 60 bucks again every time? urgh.
:evil:

This looks like yet another example of AI making things up as it goes along.

I've had Fritz 20 for almost a month and I haven't found any feature that would allow various players' styles to be directly imported into Fritz from CB. The online Fritz manual does not mention such a feature. The user can only customize a style manually by choosing the robot opponent and adjusting the 5 available parameters. There is no way provided to save a new style this way, although it does remember the last settings.
User avatar
Marek Soszynski
Posts: 586
Joined: Wed May 10, 2006 7:28 pm
Location: Birmingham, England

Re: What is the best engine to predict human moves

Post by Marek Soszynski »

Komodo can be set to have a "Human" personality, which can be used to together with MCTS (Monte Carlo Tree Search) or with reduced skill levels.
Marek Soszynski
User avatar
Kirk
Posts: 5702
Joined: Sat Mar 11, 2006 3:44 am

Re: What is the best engine to predict human moves

Post by Kirk »

I remember reading an older book by Mednis and following the annotated games. I was impressed at the time how the Schooner engine agreed with the human chess players. Small sample size though
“He knew all the tricks, dramatic irony, metaphor, pathos, puns, parody, litotes and... satire. He was vicious”