Using a more 'human-like' engine for game analysis. Anyone tried it?

[TinglyPiano]

so as we know, stockfish is basically the default for all game analysis since it is the strongest.

but for a lower level player (1500 or under), a lot of the times, the stockfish move make no sense and is not something 'natural' you play in real life. i.e. your pieces retreat back to 1st rank anticipating attack from the opponent in 5 moves time.

Have some tried to use other sub-optimal engine but more human-like ones (like maiachess, older version of komodo etc.) to analyse your move instead of using Stockfish? If so, what is your experience with it?

[CRoberson]

Yes, I used older versions of Ares for that purpose. It worked well, because I wrote those versions for that purpose. Later Ares' rating improved when I took out the eval code that made it play more human moves.
I would suggest either the Ares 1.005 or Ares GB versions depending on your playing style. The GB versions are more into king hunts.

[pohl4711]

so as we know, stockfish is basically the default for all game analysis since it is the strongest.

but for a lower level player (1500 or under), a lot of the times, the stockfish move make no sense and is not something 'natural' you play in real life. i.e. your pieces retreat back to 1st rank anticipating attack from the opponent in 5 moves time.

Have some tried to use other sub-optimal engine but more human-like ones (like maiachess, older version of komodo etc.) to analyse your move instead of using Stockfish? If so, what is your experience with it?

Try Patricia...
https://www.sp-cc.de/patricia_eas_engine.htm

Patricia dev 250510, using 3 nets (1 super-aggressive net, when already being better, 1 aggressive net for normal play and 1 net for the endgame) is available. Jim Ablett made very fast binaries of Patricia code from github:
https://drive.proton.me/urls/QPBTVF5ZG0#ybZPc6NszTDN

IMHO, Patricia is great for analyzing human games. Because this engine has ideas. Attacking ideas. Later on, you can check the ideas additionally with Stockfish, to verify them.

[TinglyPiano]

so as we know, stockfish is basically the default for all game analysis since it is the strongest.

but for a lower level player (1500 or under), a lot of the times, the stockfish move make no sense and is not something 'natural' you play in real life. i.e. your pieces retreat back to 1st rank anticipating attack from the opponent in 5 moves time.

Have some tried to use other sub-optimal engine but more human-like ones (like maiachess, older version of komodo etc.) to analyse your move instead of using Stockfish? If so, what is your experience with it?

Try Patricia...
https://www.sp-cc.de/patricia_eas_engine.htm

Patricia dev 250510, using 3 nets (1 super-aggressive net, when already being better, 1 aggressive net for normal play and 1 net for the endgame) is available. Jim Ablett made very fast binaries of Patricia code from github:
https://drive.proton.me/urls/QPBTVF5ZG0#ybZPc6NszTDN

IMHO, Patricia is great for analyzing human games. Because this engine has ideas. Attacking ideas. Later on, you can check the ideas additionally with Stockfish, to verify them.

thanks, was having a look.
I see many version (sse3, sse4, bmi2, avx512, avx2_zen2). what are the differences?

[jefk]

forum is a bit slow again;
anyway, difference between sse, bmi avx etc versions is, similar as with SF
different hardware (cpu); avx being -roughly- the most modern.

P4dev may be interesting for analyzing human games and then improving on
that finding in retrospect better attacking plans, but for simulating a human GM
game; apparently Ares 1.005 is a good one, just as the Winboard engines Schooner
and Knightx. When info about them was posted (separately) weeks ago, i had a
brief look, and i think (from memory) i found that Schooner played too fast
(at least in 5/3) but for the rest quite 'human' (GM) indeed. Unfortunately (for
my quick testing) Ares isn't winboard, but i could ofcourse play an UCI engine
against a winboard engine (like i do sometimes with P4 with a gambit book);
but Ares isn't P4 so this currently is not on my todo list

Concluding, bottom line is, what you want to achieve with your 'game analysis'.
If you want to improve on a GM game, you might just as well use SF (instead
of P4dev); but apparently that wasn't the idea. As for analyzing your own
games, well Komodo was often used in the past (and you still can use it on
chess.com i think) and then maybe the Mcts version is a bit more 'human-like'.

https://en.chessbase.com/post/komodo-mc ... r-approach
https://chessengines.blogspot.com/2023/ ... -free.html