Friday, January 14, 2011

The Growing Strength of Engines

Writing in the March 2005 edition of Chess Life magazine FIDE Master Alex Dunne gave a game from the ICCF 50th Jubilee World Championship (the game was started in 2001) between former world CC champion Grigory Sanakoev (who finished last with +0 -6 =2) and Mikhail Umansky who won the event (+6 -0 =2).
While leafing through the magazine what caught my eye was the notes. In the game’s introduction, Dunne wrote, “Umansky shows his tactical strength in this game, unleashing a combination …” and then of the move itself comments, “Umansky’s combination gives up a R and two pieces for the Q but leaves Black with a winning long range attack.”



Out of curiosity I put the position in Fritz just to see what the engines came up with. The first move Houdini considered was 14…Nfxg4 (evaluated at 9.33 in Black’s favor) which is what Umansky actually played. Fritz 12 first looked at 14…h4, but quickly found the correct move (which it evaluated at 9.20 in Black’s favor). Crafty 23.01 and Stockfish 1.06 also found the correct move in a matter of seconds.


I was ready to blast Dunne for trying to pretend that Umansky found 14…Nfxg4 on his own until I remembered that these two players didn’t have these engines available when this game was played…or at least not when 14…Nfxg4 was played. So, I set up the position and analyzed it with Fritz 6.0; in the initial position it suggested 14…h6 and after 14…Nfxg4 evaluated the position as very slightly in White’s favor. Of course this is all not very scientific because I only let Fritz 6 run for about 10 minutes.

As I’ve explained in several previous posts, in order to be a top level CC player these days you still have to be a very strong player with a tremendous understanding of chess and I think this little experiment proves the point. There is more to top level play than just choosing an engine’s move as evidenced by the fact that Umansky didn’t lose any games in this event while Sanakoev didn’t win any. In the above position, clearly Umansky demonstrated skill that was beyond the understanding of the engines of the day.

1 comment:

  1. Have you played with the personality settings at all. I found some settings that gave Stockfish a different feel to it! Actually so far a game ahead of the default settings in my tournament (but still early)

    [Stockfish-191-32-ja TAHOMA]]
    UCI_Chess960=false
    Minimum Thinking Time=20
    Emergency Move Time=70
    Emergency Base Time=200
    Emergency Move Horizon=40
    MultiPV=1
    Ponder=true
    Threads=1
    Maximum Number of Threads per Split Point=5
    Minimum Split Depth=4
    Pawn Endgame Extension (non-PV nodes)=2
    Pawn Endgame Extension (PV nodes)=2
    Passed Pawn Extension (non-PV nodes)=0
    Passed Pawn Extension (PV nodes)=1
    Pawn Push to 7th Extension (non-PV nodes)=1
    Pawn Push to 7th Extension (PV nodes)=1
    Mate Threat Extension (non-PV nodes)=2
    Mate Threat Extension (PV nodes)=2
    Single Evasion Extension (non-PV nodes)=2
    Single Evasion Extension (PV nodes)=2
    Check Extension (non-PV nodes)=1
    Check Extension (PV nodes)=2
    Cowardice=103
    Aggressiveness=97
    Space=117
    Passed Pawns (Endgame)=98
    Passed Pawns (Middle Game)=103
    Pawn Structure (Endgame)=97
    Pawn Structure (Middle Game)=97
    Mobility (Endgame)=95
    Mobility (Middle Game)=126
    Best Book Move=false
    Book File=book.bin
    Search Log Filename=SearchLog.txt
    Use Search Log=false

    ReplyDelete