Friday, March 18, 2011

What's the Best Chess Engine?

      That’s hard to say because it depends on what you are looking for. As reader Tommyg pointed out, for human-like play, Junior and Hiarcs are good. On the other hand, we saw in the post on Engine Positional Evaluations and Comparisons, some engines don’t make especially good positional evaluations under certain circumstances.
      An engine that is good in blitz play may not be the one you want to use in CC play or for analyzing games. My experience using engines in CC play is at Lechenicher SchachServer where engine us is allowed. My first tournaments there saw me using an old desk top computer and Fritz 5 and the results were not too good. In a recent (unfinished tournament) using Houdini I have much better results. So far I’m in first place with +2 -0 =2, but there’s still a long ways to go.
      The engines that rank the highest in engine tournaments are not necessarily the best engines to use for analysis or CC and the best way of determining which engines are best for CC analysis, or any analysis for that matter, is the infinite analysis function. When analyzing a position, I don’t want to let it run for hours. While analyzing any game, whether it’s when I’m playing over games from a book or analyzing a game on LSS, letting an engine run for a few minutes is sufficient for my purposes. I want an engine that will provide a reasonable evaluation in a short amount of time.
      One test I saw where free engines were tested in order to see which ones found the best move in the fastest amount of time were: Rybka 4, Critter, Stockfish, Houdini, Hannibal, Gull, Umko, Toga II 1.34, Toga II 1.4.5c, and Komodo. The tester tried Crafty but gave up on it because it was so slow in producing a result.
      It was a different story when it came to evaluations though. Houdini and Crafty (even though the tester gave up on waiting for it to make a final evaluation) were the most optimistic while Komodo, Stockfish, Protector were in the middle and the remaining engines were more conservative. The difference in evaluations was about 1.25 P’s which is actually quite considerable.
      So far in all the blitz tests I’ve conducted, Houdini has come out best and it’s the one I rely on. In a 6-game blitz match against Komodo, Houdini won +2 -1 =3. The most interesting game was one of the draws which reached the following position:
Houdini vs. Komodo
      I stopped the game here at move 81 because according to the Shredder endgame database the position is drawn. Houdini gave White a 1.02 P advantage though. In fact Houdini maintained it had about a one Pawn advantage throughout most of the game, but it was never able to force the win.  So, based on my experience, Houdini does tend to be overly optimistic in it evaluations.  Fritz 12 is more conservative but so far it has not done well in blitz matches against Houdini.  I’m really not sure what to make of this, but Houdini is my engine of choice for analysis and LSS play these days because it selects good moves fairly quickly despite its somewhat overoptimistic outlook.

1 comment:

  1. I have heard that Houdini is amazing for analysis but I just can't get myself to use it. Based on everything I have read it does seem that Houdini was created from less then ethical origins!

    And whenever I have asked the creator or Houdini outright on a forum he always evades the answer. He gives a very legally opaque answer just to give himself some legal room. I find his evasiveness very telling.

    ReplyDelete