lichess.org
Donate

Stockfish 17 cant beat Stockfish 15.1 on my laptop

I played at least 10 games - no wins for SF 17, and at least 2 wins for Sf 15.1.
I even included opening book in SF 17 settings giving him some advantage - yet 2 games finished draw.
Anyone else having similar expierence?
What could be the cause ? Games are short, bullet or blitz, but same settings for both machines

I played at least 10 games - no wins for SF 17, and at least 2 wins for Sf 15.1. I even included opening book in SF 17 settings giving him some advantage - yet 2 games finished draw. Anyone else having similar expierence? What could be the cause ? Games are short, bullet or blitz, but same settings for both machines

@zilez2003 said in #1:

I played at least 10 games - no wins for SF 17, and at least 2 wins for Sf 15.1.
I even included opening book in SF 17 settings giving him some advantage - yet 2 games finished draw.
Anyone else having similar expierence?
What could be the cause ? Games are short, bullet or blitz, but same settings for both machines

There's only about 10 rating points between the two engines. It could just be random luck. You need to play lots more than just 10 games. (Try hundreds or even thousands.)
Also, make sure all settings are the same. Same opening book, etc.

https://computerchess.org.uk/ccrl/404/rating_list_all.html

@zilez2003 said in #1: > I played at least 10 games - no wins for SF 17, and at least 2 wins for Sf 15.1. > I even included opening book in SF 17 settings giving him some advantage - yet 2 games finished draw. > Anyone else having similar expierence? > What could be the cause ? Games are short, bullet or blitz, but same settings for both machines There's only about 10 rating points between the two engines. It could just be random luck. You need to play lots more than just 10 games. (Try hundreds or even thousands.) Also, make sure *all* settings are the same. Same opening book, etc. https://computerchess.org.uk/ccrl/404/rating_list_all.html

@NeuralGnat said in #3:

There's only about 10 rating points between the two engines.
Where is the source of this information?

@NeuralGnat said in #3: > There's only about 10 rating points between the two engines. Where is the source of this information?
<Comment deleted by user>

@zilez2003 said in #1:

I played at least 10 games - no wins for SF 17, and at least 2 wins for Sf 15.1.
I even included opening book in SF 17 settings giving him some advantage - yet 2 games finished draw.
Anyone else having similar expierence?
What could be the cause ? Games are short, bullet or blitz, but same settings for both machines

What? I never tried that. I just replace my old stockfish to the latest one for my reference. I also use ShashChess 38 for analyzing game. Stockfish was in the top of all engine after I made some engine tournaments. Hmm if you're not sure, just do stockfish 16 vs other engine and stockfish 15 vs other engine. How's the result?

@zilez2003 said in #1: > I played at least 10 games - no wins for SF 17, and at least 2 wins for Sf 15.1. > I even included opening book in SF 17 settings giving him some advantage - yet 2 games finished draw. > Anyone else having similar expierence? > What could be the cause ? Games are short, bullet or blitz, but same settings for both machines What? I never tried that. I just replace my old stockfish to the latest one for my reference. I also use ShashChess 38 for analyzing game. Stockfish was in the top of all engine after I made some engine tournaments. Hmm if you're not sure, just do stockfish 16 vs other engine and stockfish 15 vs other engine. How's the result?
  1. As others have said, you should play many more games.
  2. State what opening book you used; even better post all 10 games.
  3. For each opening, you should play a pair of games (each engine should play an opening with both colours), and see check if SF 15.1 won more game pairs.
  4. State your time control precisely (not just "at least 10 games of bullet or blitz").
  5. Start each engine on command line and run a command called 'bench' and let us know what nodes per second (NPS) it reached. (Better run the command a few times and check the average NPS reached.) An engine using hand crafted evaluation (HCE) may do a lot more NPS than the same engine with neural network (NN). This may make some difference for extremely fast time controls. Check if SF 15.1 is using HCE or NN. Stockfish 17 has only NN.
  6. Which executables did you use? Did you compile them (if so did you compile them on the same machine). or you downloaded them from somewhere?
  7. According to SF release notes https://github.com/official-stockfish/Stockfish/releases for SF 16 "In self-play against Stockfish 15, this new release gains up to 50 Elo and wins up to 12 times more game pairs than it loses."
1. As others have said, you should play many more games. 2. State what opening book you used; even better post all 10 games. 3. For each opening, you should play a pair of games (each engine should play an opening with both colours), and see check if SF 15.1 won more game pairs. 4. State your time control precisely (not just "at least 10 games of bullet or blitz"). 5. Start each engine on command line and run a command called 'bench' and let us know what nodes per second (NPS) it reached. (Better run the command a few times and check the average NPS reached.) An engine using hand crafted evaluation (HCE) may do a lot more NPS than the same engine with neural network (NN). This may make some difference for extremely fast time controls. Check if SF 15.1 is using HCE or NN. Stockfish 17 has only NN. 6. Which executables did you use? Did you compile them (if so did you compile them on the same machine). or you downloaded them from somewhere? 6. According to SF release notes https://github.com/official-stockfish/Stockfish/releases for SF 16 "In self-play against Stockfish 15, this new release gains up to 50 Elo and wins up to 12 times more game pairs than it loses."

@kajalmaya said in #8:

  1. As others have said, you should play many more games.
  2. State what opening book you used; even better post all 10 games.
  3. For each opening, you should play a pair of games (each engine should play an opening with both colours), and see check if SF 15.1 won more game pairs.
  4. State your time control precisely (not just "at least 10 games of bullet or blitz").
  5. Start each engine on command line and run a command called 'bench' and let us know what nodes per second (NPS) it reached. (Better run the command a few times and check the average NPS reached.) An engine using hand crafted evaluation (HCE) may do a lot more NPS than the same engine with neural network (NN). This may make some difference for extremely fast time controls. Check if SF 15.1 is using HCE or NN. Stockfish 17 has only NN.
  6. Which executables did you use? Did you compile them (if so did you compile them on the same machine). or you downloaded them from somewhere?
  7. According to SF release notes github.com/official-stockfish/Stockfish/releases for SF 16 "In self-play against Stockfish 15, this new release gains up to 50 Elo and wins up to 12 times more game pairs than it loses."

To many tasks for me :)
I was using Arena
Executables downloaded from official web site.
Time controls - 1 pin per side, 2 mins per side, 3 mins per side, 2 mins + 1 second bonus per move ...
I really dont have wish to test all of these things, just was curious if this happened to some other people.
Anyway, thanks for your good will to try to help

@kajalmaya said in #8: > 1. As others have said, you should play many more games. > 2. State what opening book you used; even better post all 10 games. > 3. For each opening, you should play a pair of games (each engine should play an opening with both colours), and see check if SF 15.1 won more game pairs. > 4. State your time control precisely (not just "at least 10 games of bullet or blitz"). > 5. Start each engine on command line and run a command called 'bench' and let us know what nodes per second (NPS) it reached. (Better run the command a few times and check the average NPS reached.) An engine using hand crafted evaluation (HCE) may do a lot more NPS than the same engine with neural network (NN). This may make some difference for extremely fast time controls. Check if SF 15.1 is using HCE or NN. Stockfish 17 has only NN. > 6. Which executables did you use? Did you compile them (if so did you compile them on the same machine). or you downloaded them from somewhere? > 6. According to SF release notes github.com/official-stockfish/Stockfish/releases for SF 16 "In self-play against Stockfish 15, this new release gains up to 50 Elo and wins up to 12 times more game pairs than it loses." To many tasks for me :) I was using Arena Executables downloaded from official web site. Time controls - 1 pin per side, 2 mins per side, 3 mins per side, 2 mins + 1 second bonus per move ... I really dont have wish to test all of these things, just was curious if this happened to some other people. Anyway, thanks for your good will to try to help

This topic has been archived and can no longer be replied to.