Results of Computer Olympiad ?? Amazons forum

11 replies. Last post: 2010-10-09

Reply to this topic Return to forum

Results of Computer Olympiad ??
  • Ingo Althofer at 2010-10-01

    Does someone here know how the Amazons competition

    in the Computer Olympiad 2010 ended?

    Ingo.

  • Ingo Althofer at 2010-10-05

    Finally I found a website with the results:

    http://ticc.uvt.nl/icga/cg2010results/Amazons.html

    Both Invader (by Richard Lorentz + team) and

    8QP (by Johan de Koning) finished with 10 points

    out of 12 games. Invader won the playoff by 3-1.

    So Gold medal to the Invader team (congrats!)

    Silver to Johan de Koning,

    Bronze to Julien Kloetzer!

    Ingo.

  • FatPhil at 2010-10-05

    Many of the links on that page don't seem to point anywhere useful.

    Why wasn't gamazons represented? Does anyone know how well it would have fared compared to the other competitors?

  • wanderer_bot at 2010-10-05

    Unfortunately, there seem to be too many different sites posting various things related to the Olympiad(s). The site mentioned above, as far as I know, is only pointed to by the ICGA main site:

    http://ticc.uvt.nl/icga/

    JAIST, the host of the Kanazawa Olympiads, has a site for the Olympiads here (which has a pointer to the ICGA main site):

    http://www.jaist.ac.jp/ICGA-events-2010/index-e.html

    The ICGA site for Amazons at all the Olympiads is here (which you will notice appears to be hosted at a different place from the ICGA site):

    http://www.grappa.univ-lille3.fr/icga/game.php?id=15

    I hope this helps. :)

    -Richard (Lorentz, who still is not trustworthy enough to post under his own name)

  • Ingo Althofer at 2010-10-05

    Hi Richard,

    congratulations again for winning the gold medal

    in Amazons!

    What was your impression on the progress, especially:

    - How much stronger has Invader become?

    - How much stronger haas 8QP become?

    Are there now new insights, how much advantage (if any)

    the first player in Amazons has?

    > -Richard (Lorentz, who still is not trustworthy

    > enough to post under his own name)

    That is (a strange) part of the Richard I (Malaschitz) system.

    Perhaps he does not like another Richard besides him?!

    See it positively: You may become part of game

    programming history as “Richard the Wanderer”!

    Ingo (swimming in brackets).

  • wanderer_bot at 2010-10-06

    Thanks for the congrats. It was a real squeaker this time.

    Invader made a big leap 2 years ago when it switched from a purely mini-max/alpha-beta approach to a hybrid MCTS approach. “Hybrid” because even though it was/is MCTS it still made heavy use of the evaluation function from the mini-max version. Since then there has not been much progress. Some tuning of parameters. A foray into multi-threading that seems to have been of some help. I haven't actually tested the Beijing version against the Kanazawa version, but I would guess no more than 75 Elo points improvement.

    Johan claims to have not improved 8QP recently. Since we play so few games at these tournaments it is almost impossible to tell with any degree of confidence whether a program has really improved or not, but looking at the games I have the sneaky suspicion that he may also have done a little tuning. Again, with so few games it's hard to really tell, but this is the first time that both Invader and 8QP won all of their games against the other opponents.

    Does White, the first player, have an advantage? I have never believed this to be the case at the current level of play. Here is the data from Kanazawa. 28 games were played. Among the games where Invader was not playing 8QP, exactly half were won by White. Among the 8 games played between 8QP and Invader, 8 were won by White.

    -Richard (the Wanderer?)

  • wanderer_bot at 2010-10-06

    Oops. That last sentence should have ended: “…5 were won by white.”

  • Ingo Althofer at 2010-10-07

    Thx for the details, Richard.

    > Does White, the first player, have an advantage? I have never

    > believed this to be the case at the current level of play.

    > Here is the data from Kanazawa. 28 games were played. Among

    > the games where Invader was not playing 8QP, exactly half

    > were won by White.

    > Among the 8 games played between 8QP and Invader, 5 were won by White.

    Which sorts of opening books were used?

    How large and deep were they?

    Ingo.

  • Ingo Althofer at 2010-10-07

    Tournament Title Participants Score Games

    Amazons-2010-Kanazawa 7

    Amazons, 2009, Pamplona Gold medal 5 9.0 12

    Amazons, 2008, Beijing Gold medal 4 10.0 12

    Amazons, 2005, Taipei Silver medal 3 3.0 8

    Amazons, 2003, Graz Silver medal 5 5.0 8

    Amazons, 2002, Maastricht Bronze medal 6 7.0 10

    Amazons, 2001, Maastricht Bronze medal 4 5.0 12

  • Ingo Althofer at 2010-10-07

    Sorry for the previous semi-post. I want(ed) to give an

    overview of Richard's Amazons medals at Computer Olympiads

    and pressed a wrong button.

    Here is what I wanted to show:

    Tournament Title Participants

    2010-Kanazawa Gold 7

    2009-Pamplona Gold 5

    2008-Beijing Gold 4

    2005-Taipei Silver 3

    2003-Graz Silver 5

    2002-Maastricht Bronze 6

    2001-Maastricht Bronze 4

    AND: He got two medals in Havannah - a golden one in 2009,

    a silver one in 2010…

    plus a bronze in 2008 in “Dots and Boxes”

    (hopefully I did not forget any)

    Ingo.

  • wanderer_bot at 2010-10-09

    Ingo, thanks for the above. But to answer your question about opening books…

    Only Campya and Invader had opening books. Both were built automatically, though Campya's might have had a bit more hand tuning.

    Invader's book was designed not to go any deeper than 6 plies. The odds of getting hits beyond that are so low I thought it better to focus on breadth rather than depth. The only hand tuning I did was to take the top 4 or 5 moves at ply 1 and allow Invader to make any of them as the first move rather than insisting that it always make the one it thinks is best. (I believe the book has about 10 possible first moves.) This allows for more variety and, in fact, when playing 8QP we agreed that we would ensure that the first move was different in every game.

    Thinking back to the games played, the results were something like this. When playing White Invader was always in book for the first move :), maybe in about half the games also got a hit for his second move (ply 3), and never got a hit beyond that. When playing Black, I believe 1 game was out of book immediately (after White's first move), and the remaining games were split about 50 - 50 between getting 1 or 2 moves out of the book, except for the last game played against Arrow where Invader played 3 book moves – all the way down to ply 6.

    As a side note, I believe the opening book is of some benefit (I'm not sure how much) but Johan believes just the opposite. In fact, 8QP uses much less time for his first 4 moves before he hunkers down and starts really thinking after that.

Return to forum

Reply to this topic