Upgrade to Chess.com Premium!

A study on engines


  • 12 months ago · Quote · #1

    Haiku575

    I've decided to embark on an interesting experiment -- I shall set out to discover how much, exactly, the search depths of a chess engine affect its rating. To do this, I shall use the GUI Arena and a group of chess engines for testing . . .each engine shall suddenly be made into 20 different engines -- each one representing that engine at a specific search depth. I shall make them play enough games, hopefully, against enough opponents, that I shall obtain a semi-accurate ELO rating. Out of curiosity, I shall include not only some of the world's top chess engines, but also some of the worst. . . several of the worst chess programs do not have good pruning, so they never reach the kind of search depths that others can. This might even the playing field for the weaker chess engines, because they will analyze more moves for each search depth than their competitors... an intriguing thought.

    The rating changes between the search depths will then be graphed, after a time. The largest differences will be between the lowest search depths (1 to 10 moves).

    All engines shall use the same opening book, which I have decided will be a slightly modified version of the book that Arena provides.

    An alternative motive for this: watching the world's top chess engines play like idiots. XD

    If any of the games prove to be interesting, I shall post them.

    The engines I plan on using:

    Anmon 5.75

    Critter 1.4 (64)

    Deep Shredder 12 (64)

    Dragon 4.6

    Fruit 2.3.1

    Gull 1.2 (64)

    Hermann 2.6 (64)

    Houdini 1.5a (64)

    Komodo 3 (64)

    Nejmet 3.07

    Ruffian 1.0.5

    Rybka 2.2 (64)

    SOS 5.1

    Stockfish 2.2 (64)

    Toga II 1.4 (64)

    I'll keep myself interested by running tournaments between them, working out the statistics, etc. I'll try to post regular updates on this.

    Finally, I can add more chess engines to this experiment. If anyone would to suggest some, you are welcome to do so. However, my budget is severely limited -- note that only one commercial chess engine is being used. Please be nice and suggest free versions!

  • 12 months ago · Quote · #2

    Haiku575

                                          Assisted Suicide



  • 12 months ago · Quote · #3

    Haiku575

    Oh. I made a typo then. Sorry! It was 0-1 XD

  • 12 months ago · Quote · #4

    Dodger111

    Take up  jogging instead you will be much happier and get a lot more out of it.

  • 12 months ago · Quote · #5

    MSC157

    No Fritz?

  • 12 months ago · Quote · #6

    silentiarius

    You may find this thread interesting:

    http://talkchess.com/forum/viewtopic.php?t=43598

  • 12 months ago · Quote · #7

    Haiku575

    I did find it interesting... but that used different time controls, not different search depths!

  • 12 months ago · Quote · #8

    Lucidish_Lux

    Yes, Fritz should be included. It comes free with ChessBase Lite...not sure if you can extract it to run it in a different program, but if you can, that's another popular one. You'll get an old version though.

    I like the idea. I ran Fritz against Rybka (both free versions) once, and Fritz was sure it was winning the whole time, +2 to +3...Rybka thought it was equal...Rybka won. Fritz won material, but ruined his positon for it, and Rybka was able to take advantage of that. It was a very interesting game to watch.

  • 12 months ago · Quote · #9

    MSC157

    Lucidish_Lux wrote:

    Yes, Fritz should be included. It comes free with ChessBase Lite...not sure if you can extract it to run it in a different program, but if you can, that's another popular one. You'll get an old version though.

    I like the idea. I ran Fritz against Rybka (both free versions) once, and Fritz was sure it was winning the whole time, +2 to +3...Rybka thought it was equal...Rybka won. Fritz won material, but ruined his positon for it, and Rybka was able to take advantage of that. It was a very interesting game to watch.

    Woow, really? Which Fritz? Which Rybka? I have Fr6 and Ryb222

  • 12 months ago · Quote · #10

    Haiku575

    I'm running a double elimination tournament right now, with the maximum search depth at 5. It isn't finished yet, but so far I've awarded some titles to some of the competitors:

    Most Prepared Opponent: Deep Shredder 12 (depth = 5). I haven't managed to switch Deep Shredder to Arena, so it's playing on ChessBase with the opening book that came with it... while this won't make a difference at this search depth, it will be distinct advantage over its Titus-weilding opponents...

    Most Idiotic Opponent: Probably Nejmet (depth = 2). I posted its most recent game above, hahaha.

    Most Useless Opponent: Dragon 4.6 (depth = 1). For some reason, Dragon won't play at a search depth of 1, at all. I keep getting an error message. Further search depths are fine though, so it must be some programming thing.

    The program version that plays at the highest search depth is probably the strongest one for that program (there has been a debate about this somewhere in chess programming circles. It has to do with how selective searching might hinder programs at deep searches.). If we assume that the results of the highest search setting will represent the odds of that program winning this tournament, then it is now safe to say that SOS 5.1 for Arena will lose. It's the only program, at a depth of 5 moves, that has been knocked out of the first bracket. Its usurper was Critter (depth = 2), in a humiliating upset.

  • 12 months ago · Quote · #11

    Haiku575

    My ChessBase comes with Fritz 5.32. I'm not sure about his, but it would probably be the same. I'll add it into the next tournament :)

  • 12 months ago · Quote · #12

    MSC157

    Haiku575 wrote:

    ...

    Most Idiotic Opponent: Probably Nejmet (depth = 2). I posted its most recent game above, hahaha.

    Most Useless Opponent: Dragon 4.6 (depth = 1). For some reason, Dragon won't play at a search depth of 1, at all. I keep getting an error message. Further search depths are fine though, so it must be some programming thing.

    ...

    Errr, how does it look like? Depth one?! :)

  • 12 months ago · Quote · #13

    Lucidish_Lux

    It was Fritz 6, not sure which Rybka. I think we ran it to equal search depth of...7? It was the first one that took a long time (more than 10-15 seconds).

  • 12 months ago · Quote · #14

    Haiku575

    Strange. On my computer, pretty much all of my chess programs get to a search depth of 10 by the first second... Houdini, Critter and Komodo get a bit further and reach 14-16 in the first second, and I've actually seen Stockfish reach a depth of 20 within the first 3 seconds in an opening position. For such a strong, free engine, Rybka actually has a pretty small search depth in tournament time controls. It might actually shine in this experiment. We'll see.

  • 12 months ago · Quote · #15

    Haiku575

    MSC157 wrote:
    Haiku575 wrote:

    ...

    Most Idiotic Opponent: Probably Nejmet (depth = 2). I posted its most recent game above, hahaha.

    Most Useless Opponent: Dragon 4.6 (depth = 1). For some reason, Dragon won't play at a search depth of 1, at all. I keep getting an error message. Further search depths are fine though, so it must be some programming thing.

    ...

    Errr, how does it look like? Depth one?! :)

    For Dragon, it'll play out the opening moves from Titus, and then as soon as it is out of book Arena says: "1-0. Illegal move attempted" or something along those lines.

    As for the other chess programs, playing at a search depth of 1? Imagine a chess player with no positional knowledge at all, and you'll have a good idea what it looks like :)

  • 12 months ago · Quote · #16

    Haiku575

    And as for tactics, at a depth of 1 move? To show you an example of what happens, I'll pull Houdini out and play against it myself. Here's what happens when Houdini, the most powerful chess program in the world, plays at that depth without an openings book:



  • 12 months ago · Quote · #17

    Lucidish_Lux

    Perhaps I'm confusing seach depth in moves versus ply. I may just be misremembering the search depth too. It was awhile ago.

  • 12 months ago · Quote · #18

    MSC157

    Haiku575 wrote:

    And as for tactics, at a depth of 1 move? To show you an example of what happens, I'll pull Houdini out and play against it myself. Here's what happens when Houdini, the most powerful chess program in the world, plays at that depth without an openings book:

     



    Very very interesting! Thanks for kibitzes :)

  • 12 months ago · Quote · #19

    Haiku575

    You're welcome :), I try to keep it interesting for everone!

  • 12 months ago · Quote · #20

    billwall

    Other engines out there, mostly free, that I have tested in the past, include Arasan, Ares, Aristarch, Booot, Comet, Crafty, Cyrano, Daydreamer, Deep Frenzee, ECE, Firebird, Gibbon, Glass, Glaurug, Hamsters, Hiarcs, Homer, List, LittleThought, Movei, N2, Naum, Patzer, Pawny, Pharaon, Philou, Pro Deo, RobboLite, Simplex, SlowChess, Spike, Sungorus, Tao, The Baron, Tornado, Twister, Ufim, Umko, Wildcat, Zappa, bright, and spark. 


Back to Top

Post your reply: