Game 1 Move 9 The Herd to play 1-1

[ah_clem]

Other plays cheerfully added upon request.

Now WAKE UP! we've got some thin'in to do here.



Position ID: ne0GECB02wYARg
Match ID: cImkAAAAAAAA

[ah_clem]

Spoiler

I don't think we want to hit a second checker - as they say, "attack a single checker, prime two".  We'd like to have him stuck behind the prime while his board crunches, and his board will never crunch if he's on the bar. If we hit, it seems our board will be the one to crunch not his.

I like covering the blot on the three - making the two point allows him to enter  on the three and hop the prime too easily.

So it's down to playing 8/7(2) or 8/6.  Normally, I'd go for 8/6 to maximize point making chances, but we're going to play 5s and 6s to escape the backmen, so I don't see the upside to the risk of a fly shot.  Thus, bar/24* 4/3 8/7(2).

[close]

[PersianLord]

Spoiler

bar/24* 4/3 8/6

[close]

[diane]

mmmmmm...coffeeeeeee...yum...not so keen on this play though 

[ah_clem]

Rollout

Spoiler

An interesting result. The two vote-getters are clearly the top two moves by a fair margin.  I would have thought that there wasn't much difference between the two, and when gnu 2-ply analysis had the Herd's choice as best by .01, I just nodded my head.

However, rollout shows a different story.  8/7(2) is a .1 favorite over 8/6 - enough that the Herd's choice is marked "bad".  What gives here?  My reason for voting for 8/7(2) was to save us from the fly shot, but that can't be the reason behind the difference - surely 2-ply analysis would incorporate the fly shot into the calculations.

The reason must be deeper, and what I come up with post-hoc is that we'll play 5s and 6's  to escape, 2's to move up the lone straggler, leaving 1s 3s and 4s to play from the prime.  8/7(2) gives us an extra 4 thus we'll hold the prime longer.  Any other ideas?

    1. Rollout          bar/24* 8/7(2) 4/3           Eq.:  +0.992
       0.690 0.256 0.016 - 0.310 0.100 0.004 CL  +0.601 CF  +0.992
      [0.002 0.002 0.001 - 0.002 0.002 0.001 CL   0.007 CF   0.015]

    2. Rollout          bar/24* 8/6 4/3              Eq.:  +0.883 ( -0.108)
       0.680 0.269 0.017 - 0.320 0.119 0.005 CL  +0.539 CF  +0.883
      [0.001 0.001 0.001 - 0.001 0.001 0.001 CL   0.004 CF   0.011]

    3. Rollout          bar/24* 4/2 3/2              Eq.:  +0.754 ( -0.238)
       0.644 0.242 0.014 - 0.356 0.119 0.007 CL  +0.475 CF  +0.754
      [0.002 0.003 0.001 - 0.002 0.002 0.001 CL   0.007 CF   0.016]

    4. Rollout          bar/24* 7/6(2) 4/3           Eq.:  +0.575 ( -0.416)
       0.595 0.263 0.019 - 0.405 0.138 0.009 CL  +0.375 CF  +0.575
      [0.002 0.002 0.002 - 0.002 0.001 0.001 CL   0.007 CF   0.016]

    5. Rollout          bar/24* 7/5 4/3              Eq.:  +0.573 ( -0.418)
       0.602 0.251 0.015 - 0.398 0.156 0.012 CL  +0.352 CF  +0.573
      [0.002 0.002 0.001 - 0.002 0.001 0.001 CL   0.007 CF   0.016]

    6. Rollout          bar/24* 4/1*                 Eq.:  +0.224 ( -0.768)
       0.555 0.210 0.013 - 0.445 0.213 0.021 CL  +0.117 CF  +0.224
      [0.002 0.003 0.001 - 0.002 0.002 0.002 CL   0.008 CF   0.019]
        Full cubeful rollout with var.redn.
        1296 games, Mersenne Twister dice gen. with seed 899112970 and quasi-random dice
        Play: 0-ply cubeful prune [expert]
        Cube: 0-ply cubeful prune [expert]

[close]

[ah_clem]

gnu 2-ply rollout

Spoiler

The 0-ply rollout seemed suspicious - could there really be a .1 equity difference between the plays?  So I rollled out in 2-ply, and 2-ply rollout agrees with the 2-ply analysis - 8/6 is slightly better, but not outside the margin of error.

XG 3-Ply  and Snowie 2-Ply are in the same ballpark - they're slpit on which play is better, but the difference is only .003  - basically a tie.

So the 0-ply rolout appears to be way out of whack.  The moral of the story here is to take bot results with a grain of salt, and if 2-ply analysis disagrees strongly with 0-ply rollout, you should do a longer rollout or try another bot.



   1. Rollout          bar/24* 8/6 4/3              Eq.:  +0.878
      0.684 0.271 0.016 - 0.316 0.115 0.006 CL  +0.549 CF  +0.878
     [0.001 0.001 0.001 - 0.001 0.001 0.000 CL   0.004 CF   0.010]

   2. Rollout          bar/24* 8/7(2) 4/3           Eq.:  +0.867 ( -0.010)
      0.686 0.258 0.015 - 0.314 0.107 0.004 CL  +0.549 CF  +0.867
     [0.001 0.001 0.000 - 0.001 0.001 0.001 CL   0.004 CF   0.008]
       Full cubeful rollout with var.redn.
       1296 games, Mersenne Twister dice gen. with seed 899112970 and quasi-random dice
       Play:  2-ply cubeful prune [world class]
       keep the first 0 0-ply moves and up to 8 more moves within equity 0.1
       Skip pruning for 1-ply moves.
       Cube: 2-ply cubeful prune [world class]

[close]