MASSIVE Difference In Accuracy b/w wintrchess and chess.com

[IshanAsati]

(chess.com) (lichess.org) (wintrchess) i just want clarity on... whatever is happening that leads to this.

[milankubin]

1. Chess algorithm is not open source.
2. Chess.com uses 16.1 stockfish (click on settings icon above board) for game review with ~5 seconds run. You can change this if you are premium, but I haven't tried it out how much difference this makes.
   I do often see there's a lot of differences in analysis (stronger accuracy) than in game review, and this is known: Stockfish 17 evaluates stronger than 16: https://lichess.org/forum/game-analysis/stockfish-161-vs-stockfish-17-server-game-accuracy-analysis
3. It looks like wintrchess is using 17.1 for everything.

You can play around at Lichess with the various settings I think and see how it works, or fork this (or Lichess) and play around with engines and see how it turns out.
If you're not a developer I understand this is not an option, but you can pretty much verify on Lichess itself, that Accuracy is determined by which engine is used and how (depth of analysis, run time / number of line).
If you set it to depth 16 with Stockfish 17, running 5 seconds and 10 lines, im guessing it will likely not calculate all these lines before the run is over.
You can run it infinite (all day) and you will see dfferences.

I remember when Chess.com was still using Komodo as default engine (not sure why) but Komodo (I believe also older stockfish), would show a Greek Gift sacrifice as a mistake or something, even though its the best possible move and leads to certain victory. But turn the engine on bigger depth and it would calculate it correctly.

Basically it comes down to that.