BGonline.org Forums

BOT Comparison some early data

Posted By: Michael Depreli
Date: Saturday, 17 October 2009, at 4:45 p.m.

In Response To: BOT Comparison some early data (Fabrice Liardet)

I have been acquainted to many GNU 2-ply's blitz takes turning out to be too optimistic by >0.1 when checked with a rollout. That effect could be aggravated by my habit of doing truncated rollouts. I am currently extensively rolling out one example on my old computer, it will take one more night.

Well if GNU 2-ply is taking optimitic cubes then all other bots on all plies are making the same mistake too. Your statement would indicate you don't understand how the positions are selected for the bot comparison.

ALL moves and cube decisions are compared to ALL bots and ALL plies so only if ALL are making the same error will the position be missed.

Once I'm done rolling out all the disagreements, I'll probably run through the whole session and roll out more checker plays and cube decisions within a certain equity where all bots agree.

In fact based on my current rollout parameters, I found a couple of plays where ALL bots and plies don't find the best play, but to be sure I'd like to roll them out a lot more. This happens when say bots chose plays A B and D and I roll out play C as well which turns out to be best.

Michael

Messages In This Thread

• BOT Comparison some early data
  Michael Depreli -- Friday, 16 October 2009, at 3:09 p.m.
  • BOT Comparison some early data
    neilkaz -- Friday, 16 October 2009, at 3:34 p.m.
    • BOT Comparison some early data
      Michael Depreli -- Friday, 16 October 2009, at 3:53 p.m.
      • BOT Comparison some early data
        neilkaz -- Friday, 16 October 2009, at 3:56 p.m.
  • BOT Comparison some early data
    Rich Munitz -- Friday, 16 October 2009, at 3:56 p.m.
    • BOT plies
      neilkaz -- Friday, 16 October 2009, at 4:06 p.m.
    • BOT Comparison some early data
      Michael Depreli -- Friday, 16 October 2009, at 4:29 p.m.
  • BOT Comparison some early data
    Matt Cohn-Geier -- Friday, 16 October 2009, at 5:28 p.m.
    • BOT Comparison some early data
      Timothy Chow -- Friday, 16 October 2009, at 5:44 p.m.
  • BOT Comparison some early data
    neilkaz -- Friday, 16 October 2009, at 5:33 p.m.
    • Reply to MCG re: headings above *NM*
      neilkaz -- Friday, 16 October 2009, at 5:47 p.m.
  • BOT Comparison some early data
    Fabrice Liardet -- Friday, 16 October 2009, at 7:44 p.m.
    • BOT Comparison some early data
      Michael Depreli -- Friday, 16 October 2009, at 8:48 p.m.
      • BOT Comparison some early data
        Fabrice Liardet -- Saturday, 17 October 2009, at 3:58 p.m.
        • BOT Comparison some early data
          Michael Depreli -- Saturday, 17 October 2009, at 4:45 p.m.
          • BOT Comparison some early data
            neilkaz -- Saturday, 17 October 2009, at 5:33 p.m.
          • BOT Comparison some early data
            Fabrice Liardet -- Saturday, 17 October 2009, at 6:56 p.m.
      • Example : 62S - 55 - Dance
        Fabrice Liardet -- Sunday, 18 October 2009, at 9:54 a.m.
        • Example : 62S - 55 - Dance
          Michael Depreli -- Sunday, 18 October 2009, at 10:42 a.m.
  • Rollout volunteer
    Mislav Radica -- Friday, 16 October 2009, at 8:25 p.m.
    • Rollout volunteer .. what to expect
      neilkaz -- Friday, 16 October 2009, at 10:25 p.m.
  • BOT Comparison some early data
    Christian Munk-Christensen -- Friday, 16 October 2009, at 9:06 p.m.
  • BOT Comparison some early data
    Ian Shaw -- Sunday, 18 October 2009, at 12:31 p.m.
    • BOT Comparison some early data
      Michael Depreli -- Sunday, 18 October 2009, at 12:38 p.m.