Evaluation and computer lines

Sort:
Avatar of valexandrov

Hi.

I've always wondered the evaluations in chess.com.

In the analysis tab (computer lines) and the bar in the review, have so different numbers from time to time. Las game for example I was getting up to 0.8 difference.

Another one: Sometimes best move in review doesn't agree with computer best line.

Anyone has a clue? Thanks.

Avatar of KeSetoKaiba
valexandrov wrote:

Hi.

I've always wondered the evaluations in chess.com.

In the analysis tab (computer lines) and the bar in the review, have so different numbers from time to time. Las game for example I was getting up to 0.8 difference.

Another one: Sometimes best move in review doesn't agree with computer best line.

Anyone has a clue? Thanks.

I've discussed this a few times in past videos on my YouTube channel as well as in past chess.com forum threads.

The reason for this is because the lines are from the chess.com Analysis (Stockfish), whereas the "evaluation" and "best moves" are from chess.com's Game Report. They both usually align, but sometimes they don't. This is because Stockfish runs at a deeper depth than the Game Report Quick Analysis does. If you were to run a deeper analysis (chess.com Diamond Membership Max depth analysis is good, but a super computer like the non-desktop Stockfish or AlphaZero etc. would be even stronger), then eventually the fluctuations will dissipate and both the Analysis and Game Report will align.

A related problem is that the game report will sometimes "change" what it considers the "best move" and this is because it will pick the move it believes is "best" and then when you play a different move, you are basically telling the computer to analyze that option. If that option becomes a better evaluation than the "best move", then the game report will change and call your move the "best move" now. It isn't that the literal best move changed (it is the same position after all). It is just that the computer "changed its mind" once it analyzed deeper and recognized that its original assessment wasn't actually "best."

Avatar of valexandrov

Thanks for the answer @KeSetoKaiba.

Doesn't sound that complicated to set those lines to the same depth, and get consistency. It is very confusing to see all those different evaluations in the different modules, and computer changing his mind.

I hope Chess.com comes up with a solution some day.

Avatar of KeSetoKaiba

I know what you mean, but I don't think this is really a chess.com problem with game report; maybe chess.com could just spread a little more awareness about how the game report works and what the statistics actually mean.

Basically, the game report you run at the beginning and doesn't change (usually Quick Analysis) and the computer lines are more "up to date" because they update in real time as the computer keeps thinking "deeper" on the position.