SF question: Aronian/Mamedyarov game [Candidates 2014]

[Dan Astrachan]

I was browsing the Candidates games today when something drew my attention in the Aronian Vs Mamedyarov game.

Here is the complete game:

[pgn]1. d4 Nf6 2. c4 e6 3. Nf3 d5 4. Nc3 Bb4 5. Bg5 Nbd7 6. cxd5 exd5 7. Nd2 c6 8. e3 Nf8 9. Bd3 Ng6 10. O-O O-O 11. f4 h6 12. Bxf6 Qxf6 13. f5 Ne7 14. Nde4 dxe4 15. Nxe4 Qh4 16. g3 Qh3 17. Nf2 Qxf1+ 18. Kxf1 Nxf5 19. Qf3 Nd6 20. e4 f6 21. Bc2 Be6 22. Nd3 Nc4 23. Kg1 Bd6 24. Nf4 Bxf4 25. gxf4 Rad8 26. f5 Bf7 27. Qc3 Rfe8 28. Bd3 Nb6 29. a4 a6 30. a5 Nc8 31. e5 Ne7 32. e6 Bh5 33. Be4 Nd5 34. Qh3 Be2 35. Kf2 Bb5 36. Rg1 Kh7 37. Qa3 Bc4 38. Rg4 Bb5 39. Rg1 Bc4 40. Rc1 Bb5 41. Bf3 Nf4 42. Rd1 Kh8 43. d5 Nxd5 44. Bh5 1-0[/pgn]


In Chessbomb, SF marked move 39.Rg1 as red, indicating a bad move, with an eval of 0.00. This is obviously wrong as this is not a mistake by White, which ended up winning the game.

I tested it on my computer and once 39. Rg1 is played, SF does report an eval of 0.00 (with Bc4 being next for Black). However, once 39...Bc4 is played, the eval becomes really high again.

I tested the position with Critter 1.6a, and it does not have this issue. Does anyone have an idea why SF evaluates this move incorrectly?

[Vinvin]

It's a well known problem with some chess engines : they eval draw when there's 2 times the same position in the game (as the rules state one needs 3 times the same position to make a draw) .

[Dan Astrachan]

Thanks Vincent. I was pretty sure it had to do with a repetition of position. Still, an odd thing for such a strong engine

[Evert]

Thanks Vincent. I was pretty sure it had to do with a repetition of position. Still, an odd thing for such a strong engine

It doesn't affect strength, just analysis (for which it would make sense to only treat the third repetition as a draw).