What does chess.com use for its analysis and game review features?
Sort:
Leto
3 days ago
0
#21
Computer competition.
Leto wrote:
somebody plays ICCF: person might use “motorbike”: from 0% to 100%. And if person won the game - well, maybe he used available supercomputer 24 hours against laptop which calculates moves for 2 minutes only. And there is NO space for human at all.
If you want to be a correspondence world champion, or even GM, you should use a supercomputer, not a laptop. And you should use the engine with human assistance, as cleverly as you can, otherwise you'll get slaughtered.
3 days ago
0
#23
playchessordie19 wrote:
@magipi agreed. Side question: why do I get different evaluations of my games when I'm using stockfish through chessis than the game review features? Don't they use stockfish for their review also?
Game Review uses Stockfish and the evaluations will depend on the strength of the review in your settings. That says, a local analysis can run longer that the strongest level, so even that can have differences.
Local analysis on the site uses the engine you have configured. That can be Stockfish, Torch, or Komodo. The evaluations on that will depend on your client; it runs in the browser band can be influenced by loads
3 days ago
0
#24
playchessordie19 wrote:
@magipi agreed. Side question: why do I get different evaluations of my games when I'm using stockfish through chessis than the game review features? Don't they use stockfish for their review also?
Game Review uses Stockfish and the evaluations will depend on the strength of the review in your settings. That said, a local analysis can run longer that the strongest level, so even that can have differences.
Local analysis on the site uses the engine you have configured. That can be Stockfish, Torch, or Komodo. The evaluations on that will depend on your client; it runs in the browser band can be influenced by client load
3 days ago
0
#25
Thanks Martin. I'm going back through many of my games now and using a buddy of mine to tweak the process a bit and try to understand why either the moves I made did not work in the position or if they did, what the overall theme and strategy is for why they worked. We're still looking at a training plan to try to correct some of the mistakes but already in a month of doing this, both my daily chess and OTB games have shown improvement.
2 days ago
0
#26
playchessordie19 wrote:
@magipi agreed. Side question: why do I get different evaluations of my games when I'm using stockfish through chessis than the game review features? Don't they use stockfish for their review also?
For their game review chess.com uses stockfish 16. They give you 4 options.
Fast analyzes at depth 18
Normal analyzes at depth 22
Deep analyzes at depth 24
Maximum analyzes at depth 26
So yeah, there are multiple factors why the evaluation is different, like you using stockfish 17 on chessis instead of stockfish 16, game review being coded differently on chesiss, and you using a higher depth on chessis.
1 day ago
0
#28
Omed wrote:
playchessordie19 wrote:
@magipi agreed. Side question: why do I get different evaluations of my games when I'm using stockfish through chessis than the game review features? Don't they use stockfish for their review also?
For their game review chess.com uses stockfish 16. They give you 4 options.
Fast analyzes at depth 18
Normal analyzes at depth 22
Deep analyzes at depth 24
Maximum analyzes at depth 26
So yeah, there are multiple factors why the evaluation is different, like you using stockfish 17 on chessis instead of stockfish 16, game review being coded differently on chesiss, and you using a higher depth on chessis.
Depth hasn't been used for a while now, so it may get less deep in some positions and deeper in others.
1 day ago
0
#29
Magipi:
Let’s skip GMs and World champions in correspondence games. If you or me would do it it shows that it’s even more ridiculous.
Let’s take more common situation. We play several games against each other with such clown rules - we may use computer. Somebody won the match. What does it show? It just shows that somebody used better computer. Or even better: 1500+stockfish plays against CM/FM who does not use anything. Result is predicted, 1500 is “champion” in this match.
Let’s skip GMs and World champions in correspondence games. If you or me would do it it shows that it’s even more ridiculous.
Let’s take more common situation. We play several games against each other with such clown rules - we may use computer. Somebody won the match. What does it show? It just shows that somebody used better computer. Or even better: 1500+stockfish plays against CM/FM who does not use anything. Result is predicted, 1500 is “champion” in this match.
Leto wrote:
Magipi:
Let’s skip GMs and World champions in correspondence games. If you or me would do it it shows that it’s even more ridiculous.
Let’s take more common situation. We play several games against each other with such clown rules - we may use computer. Somebody won the match. What does it show? It just shows that somebody used better computer. Or even better: 1500+stockfish plays against CM/FM who does not use anything. Result is predicted, 1500 is “champion” in this match.
Let’s skip GMs and World champions in correspondence games. If you or me would do it it shows that it’s even more ridiculous.
Let’s take more common situation. We play several games against each other with such clown rules - we may use computer. Somebody won the match. What does it show? It just shows that somebody used better computer. Or even better: 1500+stockfish plays against CM/FM who does not use anything. Result is predicted, 1500 is “champion” in this match.
You and I don't have the skills to add useful input to engine analysis. But the best correspondence players do have that.
If you play against a correspondence GM and you have a slightly better computer, you'll still lose every game.
1 day ago
0
#31
I understand this. Of course, strong player with good skills of computer usage will show good game.
But majority of players play at their “ordinary” level. From 200 to 2300, let’s say. And I am sure that at these levels, computer will give 90% of success, so rating different will not be important.
But majority of players play at their “ordinary” level. From 200 to 2300, let’s say. And I am sure that at these levels, computer will give 90% of success, so rating different will not be important.
Leto wrote:
I understand this. Of course, strong player with good skills of computer usage will show good game.
But majority of players play at their “ordinary” level. From 200 to 2300, let’s say. And I am sure that at these levels, computer will give 90% of success, so rating different will not be important.
But majority of players play at their “ordinary” level. From 200 to 2300, let’s say. And I am sure that at these levels, computer will give 90% of success, so rating different will not be important.
Chess.com also understands this, that's why engines are banned in daily games. I would guess that other chess sites do the same for the same reason.