BGonline.org Forums

41$ 11? (longish)

Posted By: neilkaz
Date: Friday, 31 July 2009, at 1:55 p.m.

In Response To: Replies to opening 41 where oppenent slots, 13/9 6/5 (ectonomo)

Your concerns about the best way to play 11 after 41 slot are well founded. The Snowie cubeful result is over 5000 games and has statistical significance (Snowie's +/- figures are reported as the "standard" near 2 sd interval) and shows hitting is best.

GNU however, has a huge 46,000+ game rollout which has hitting best cubeless, but has priming slightly (not enough jsd to be sure) best cubefully. This is do to the cubeful effects of winning more total games. This cubeful effect has been shown to be very strong in many GNU rollouts and I am often surprised how much difference it can make.

So we have a cubefully Snowie rollout with stat sig showing hit is best vs a much longer GNU rollout implying the cubefully priming is likely best... how to add more data to this quandry? Well..someone could extend the initial Snowie rollout to 10K+ and see if it changes substantially. Or...someone could run a long GNU rollout with a different (hopefully better playing strength setting).

I think it was about a year ago, that MCG did some rollouts(starting from game beginning) that showed GNU 3 ply play was somewhat better than 2 ply play, but that overall 3 ply cube wasn't better than 2 ply cube. Therefore it would seems that if I rolled out this play using 3 ply play and 2 ply cube I'd have a somewhat stronger setting the 2 ply for both. One thing is for sure, is that we can expect GNU to make some different plays with 3 ply play vs 2 ply play and overall we cannot expect it to play weaker. Different..YES ..somewhat... Stronger.. maybe overall.. Weaker overall..very doubtful.

But 3 ply play will take a long time..won't it? Well..GNU takes advantage of hyper threading and multiple cores and my comp on the trading floor looks like it has 8 cores and GNU rocks on it. I installed the latest GNU download for windows (comp is XP) and am getting a calib. eval of about 350,000/sec.

I slightly expanded the 2 ply search space and somewhat expanded the 3 ply but my 3/2 rollout is still blazing. I expect no problems getting 15,552 rollouts done by Monday morning when I can reaccess that computer. So far the rollout is clearly favoring hitting and priming is winning fewer total games using 3 ply play than it did in the posted 2 ply rollout. Perhaps 3 ply play better fights the priming plan? Or perhaps I don't have enough games rolled out yet. Anyhow, by the end of the work day I expect to have close to 2000 games done and will post those results. In any event, we'll soon have some new data here from a very strong bot rollout.

I'll also note that unless I am mistaken, rollouts with other bots, BGB,XG,JF also favor hitting.

I'll also test my 3 ply setting vs the 2 ply setting used in the rollout just to make sure it does play the checkers a bit better (DMP 52S is about as close to an even game as one can get it seems)

.. neilkaz ..

Messages In This Thread

• Replies to opening 41 where oppenent slots, 13/9 6/5
  ectonomo -- Tuesday, 28 July 2009, at 4:27 p.m.
  • Replies to opening 41 where oppenent slots, 13/9 6/5
    Jason Lee -- Tuesday, 28 July 2009, at 4:45 p.m.
  • Replies to opening 41 where oppenent slots, 13/9 6/5
    Keene -- Tuesday, 28 July 2009, at 5:20 p.m.
    • Replies to opening 41 where oppenent slots, 13/9 6/5
      ectonomo -- Tuesday, 28 July 2009, at 5:32 p.m.
      • Replies to opening 41 where oppenent slots, 13/9 6/5
        Matt Cohn-Geier -- Tuesday, 28 July 2009, at 6:03 p.m.
        • and then there was...ExtremeGammon
          rummer1973 -- Tuesday, 28 July 2009, at 8:24 p.m.
          • PS: Phil Simborg
            rummer1973 -- Tuesday, 28 July 2009, at 8:26 p.m.
  • 41$ 11? (longish)
    neilkaz -- Friday, 31 July 2009, at 1:55 p.m.
    • 41$ 11? (longish)
      Stick -- Friday, 31 July 2009, at 3:01 p.m.
      • 41$ 11? (longish)
        neilkaz -- Friday, 31 July 2009, at 4:14 p.m.
        • 3 ply vs. 2 ply again
          Matt Cohn-Geier -- Friday, 31 July 2009, at 4:31 p.m.
          • 3 ply vs. 2 ply again
            neilkaz -- Friday, 31 July 2009, at 4:50 p.m.
    • 41$ 11? (longish)
      Rich Munitz -- Friday, 31 July 2009, at 4:21 p.m.
      • 41$ 11? (longish)
        neilkaz -- Friday, 31 July 2009, at 4:37 p.m.
    • 1000 rollouts 3 ply play=very clear
      neilkaz -- Friday, 31 July 2009, at 4:22 p.m.
      • Oops didn't realize had truncated turned on..LOL
        neilkaz -- Friday, 31 July 2009, at 4:29 p.m.
    • 41$ 11? (longish)
      Michael Depreli -- Thursday, 27 August 2009, at 2:39 p.m.
  • See this article
    Paul Vianna -- Friday, 31 July 2009, at 3:08 p.m.