BGonline.org Forums

Getting the bots to tell the whole truth: a concrete proposal

Posted By: Daniel Murphy
Date: Tuesday, 9 August 2011, at 5:25 a.m.

In Response To: Getting the bots to tell the whole truth: a concrete proposal (Phil Simborg)

I always want to see the gammons won and lost, even at DMP, as that saves me time from having to change the score to see what the answer is at other scores. ... In any event, I don't see the harm in knowing the gammon percentage, and I do see negatives of not seeing it.

But you won't be seeing what the results would be at other scores. The reason you shouldn't want to see meaningless gammon statistics is that the gammons won/lost ATS cannot be assumed to be anything close to what they would be at a nullo score or other score.

Again, there are two reasons why this happens:

(1) When the game result is still in doubt, the bot does not consider gammons won and/or lost in making its play decisions. Consequently play decisions and gammons won and/or lost may be quite different from what they would be at a nullo score.

(2) When the game is no longer in doubt, the bot makes no effort to save meaningless gammons/backgammons that would matter at another score.

Perhaps you are assuming that the gammon/backgammon results are always close enough to be meaningful. Let me give just one counterexample that may surprise you:

Opening 2-1 slot.

In a money game rollout, this play wins about 14.3% gammons and loses about 14.6% gammons ("cubeless"). Backgammon wins/losses are less than 1%.

In a rollout at DMP, this play wins about 19% meaningless gammons and 3.5% meaningless backgammons, and loses about 22% meaningless gammons and 5% meaningless gammons.

At gammon-go, this play wins about 15% legitimate gammons and loses 22.5% meaningless gammons. At gammon-save, this play wins about 20% meaningless gammons and loses about 13.5% legitimate gammons. Meaningless backgammon wins and losses vary from 2% to 5%.

Does this example help explain why I'd rather not see meaningless ATS rollout gammon/backgammon statistics at all, than see them and try to guess when they might be relevant to other scores?

Messages In This Thread

• Getting the bots to tell the whole truth: a concrete proposal
  Timothy Chow -- Saturday, 16 July 2011, at 9:35 p.m.
  • +1 (I think)
    Casper van der Tak -- Saturday, 16 July 2011, at 9:41 p.m.
    • +1 (I think)
      Casper van der Tak -- Sunday, 17 July 2011, at 9:39 a.m.
  • Getting the bots to tell the whole truth: a concrete proposal
    Chuck Bower -- Sunday, 17 July 2011, at 12:00 a.m.
    • Getting the bots to tell the whole truth: a concrete proposal
      Michael Petch -- Sunday, 17 July 2011, at 12:31 a.m.
      • Getting the bots to tell the whole truth: a concrete proposal
        Timothy Chow -- Sunday, 17 July 2011, at 12:39 a.m.
  • Getting the bots to tell the whole truth: a concrete proposal
    Daniel Murphy -- Monday, 8 August 2011, at 11:14 p.m.
    • Getting the bots to tell the whole truth: a concrete proposal
      Phil Simborg -- Tuesday, 9 August 2011, at 4:54 a.m.
      • Getting the bots to tell the whole truth: a concrete proposal
        Daniel Murphy -- Tuesday, 9 August 2011, at 5:25 a.m.
        • Getting the bots to tell the whole truth: a concrete proposal
          Joe Russell -- Tuesday, 9 August 2011, at 12:35 p.m.
          • Getting the bots to tell the whole truth: a concrete proposal
            Daniel Murphy -- Tuesday, 9 August 2011, at 2:59 p.m.
    • Getting the bots to tell the whole truth: a concrete proposal
      Timothy Chow -- Tuesday, 9 August 2011, at 5:08 p.m.