Page 1 of 3

Do SF still have troubles with 2nd repetitions ?

Posted: Mon Mar 23, 2015 2:10 pm
by Vinvin
I though this issue was solve some months ago.
The solution was to treat differently the first move of the PV.

In analyzing this game :

[Event "z"]
[Site "?"]
[Date "2012.01.01"]
[White "x"]
[Black "y"]
[Result "1-0"]
[PlyCount "80"]

1. e4 d6 2. d4 Nf6 3. Nc3 e5 4. Nf3 Nbd7 5. Bc4 Be7 6. O-O O-O 7. a4 c6 8.
a5 exd4 9. Nxd4 Nxe4 10. Bxf7+ Rxf7 11. Ne6 Qe8 12. Nxe4 d5 13. Nc7 Qf8 14.
Ne6 Qe8 15. Nc7 Qf8 16. Nxa8 dxe4 17. Be3 Ne5 18. Qd4 Ng4 19. h3 c5 20.
Qxe4 Bf5 21. Qxb7 Nf6 22. Qxa7 Bd6 23. Qa6 Bb8 24. Rfd1 Qe8 25. Qc4 Be4 26.
Nb6 Bc7 27. Qxc5 Bxg2 28. Kxg2 Qe4+ 29. Kg1 Qe6 30. Bf4 Ne4 31. Qc4 Qg6+
32. Kf1 Qf5 33. Rd7 Qxh3+ 34. Ke1 Qh1+ 35. Qf1 Qh5 36. Rxf7 Qxf7 37. Bxc7
Qe8 38. Qc4+ Kh8 39. Be5 Nd6 40. Qc5 Ne4 1-0

After move 14..Qe8, we got this position :
[D]r1b1q1k1/pp1nbrpp/2p1N3/P2p4/4N3/8/1PP2PPP/R1BQ1RK1 w - - 4 15
Here, Nc7 is clearly winning because it's only the second repetition but SF gives 0.00 :?

Re: Do SF still have troubles with 2nd repetitions ?

Posted: Mon Mar 23, 2015 3:13 pm
by Hai
Yes it's bad.
When i use IDeA (opening book) i got the same position with 0.00 into the book after the 2-fold rep but it is a white win.
Now i have a wrong 0.00 in my book.

Re: Do SF still have troubles with 2nd repetitions ?

Posted: Mon Mar 23, 2015 4:21 pm
by Evert
I think the last time this came up the conclusion was that fixing this doesn't gain any Elo and so isn't worth it.

Maybe worth it to broach the subject again, or otherwise for someone to fork stockfish with this fix applied. The problem with such forks is keeping them current with the main development line...

Re: Do SF still have troubles with 2nd repetitions ?

Posted: Mon Mar 23, 2015 5:20 pm
by nabildanial
A pull request of the fix got rejected in the past despite being favored by Gary and Marco. https://github.com/official-stockfish/Stockfish/pull/80

Re: Do SF still have troubles with 2nd repetitions ?

Posted: Mon Mar 23, 2015 5:33 pm
by Vinvin
nabildanial wrote:A pull request of the fix got rejected in the past despite being favored by Gary and Marco. https://github.com/official-stockfish/Stockfish/pull/80
What a pity !
99% of chess players would think SF is really dumb to not play Nc7 in this position ...

Re: Do SF still have troubles with 2nd repetitions ?

Posted: Mon Mar 23, 2015 5:44 pm
by nabildanial
Vinvin wrote:
nabildanial wrote:A pull request of the fix got rejected in the past despite being favored by Gary and Marco. https://github.com/official-stockfish/Stockfish/pull/80
What a pity !
99% of chess players would think SF is really dumb to not play Nc7 in this position ...
Tell Joona that lol.

Re: Do SF still have troubles with 2nd repetitions ?

Posted: Mon Mar 23, 2015 5:53 pm
by Vinvin
nabildanial wrote:
Vinvin wrote:
nabildanial wrote:A pull request of the fix got rejected in the past despite being favored by Gary and Marco. https://github.com/official-stockfish/Stockfish/pull/80
What a pity !
99% of chess players would think SF is really dumb to not play Nc7 in this position ...
Tell Joona that lol.
I already see the answer "SF is designed to play chess". But SF masters don't know there are 1000 more people who use SF to analyze games than people who use SF to play games ...

Re: Do SF still have troubles with 2nd repetitions ?

Posted: Mon Mar 23, 2015 6:05 pm
by vincenegri
Evert wrote:I think the last time this came up the conclusion was that fixing this doesn't gain any Elo and so isn't worth it.

Maybe worth it to broach the subject again, or otherwise for someone to fork stockfish with this fix applied. The problem with such forks is keeping them current with the main development line...
That's not really a problem. The SF sources are maintained on Github; keeping a forked version with a small patch synced with master is very simple with any git client.

P.S. If you are using SF as an analysis tool rather than to play engine matches, you may want to look at the "mate finder" fork by Joseph Ellis. It sacrifices OTB strength in return for better tactical line detection, rather like Houdini's tactical mode.

Re: Do SF still have troubles with 2nd repetitions ?

Posted: Mon Mar 23, 2015 6:33 pm
by Joerg Oster
Vinvin wrote:
nabildanial wrote:
Vinvin wrote:
nabildanial wrote:A pull request of the fix got rejected in the past despite being favored by Gary and Marco. https://github.com/official-stockfish/Stockfish/pull/80
What a pity !
99% of chess players would think SF is really dumb to not play Nc7 in this position ...
Tell Joona that lol.
I already see the answer "SF is designed to play chess". But SF masters don't know there are 1000 more people who use SF to analyze games than people who use SF to play games ...
Here https://github.com/joergoster/Stockfish ... ae4cce5b7d you can find the necessary patch.
Insert it into the actual sources, compile, and you're done!

Re: Do SF still have troubles with 2nd repetitions ?

Posted: Mon Mar 23, 2015 8:27 pm
by BBauer
Joerg Oster wrote
Insert it into the actual sources, compile, and you're done!
Youre done for the moment.
Stockfish is work in development. After the next change you have to do the same again.

Perhaps the developers will reconsider.
Kind regards
Bernhard