Game 3, move 13: skater 6-6

[diane]

Y9sCADPYtg0AFg:EQG7AAAAGAAA

[skater]

23-17*, 22-16, 22-10

[NIHILIST]

I realize that I'm on the other team, but this position piqued my interest.

I think it might be wrong to hit here. I think id have simply played the 2 checkers off the 22 point, thus allowing the herd to play its full roll.

From the herd's standpoint, we desperately need to slow down and preserve some timing for our fragile backgame. I think hitting does that. Granted we're favored to enter but there are also crushing numbers like 5-5, 4-4, 5-4, etc that make us less of a threat if we ever do hit one of his blots.

Maybe someone would do a rollout of this one.

Bob

[diane]

Clem...clem...clem...where are you clem??

[ah_clem]

Clem...clem...clem...where are you clem??

This move piqued my interest too.  I thought about eschewing the hit since white's runner isn't in any real danger. If white doesn't hit, blue's board will crunch and blue will have little chance of winning.  So I think declining to hit wins more games.

But Skater wants a gammon, and sending a 5th checker back makes that more likely.  So I'd hit here due to match score.  At DMP  I wouldn't.

Starting a rollout....

[ah_clem]

rollout

Spoiler

At score, hitting is best, but how one hits is important too.  gnubg 0-ply rollout likes the hit and then to leave a runner in his homeboard.  I'm not sure I see the logic behind this - I don't think white wants to pick up a sixth checker and face the possibility of three anchors, but the gammon rate definitely goes up with this approach, so maybe six men back is worth playing for.  Anyway the "obvious" moves that get all three runners out of there are about a .03 error.

Not hitting wins more single games than the bot move, but no more than Skater's play, so it may or may not be correct at DMP - we'd need to roll it out at that setting to know for sure, and we'd probably need better precision than 0-ply rolllout.

Interesting position....



    1. Rollout          23/17* 22/4                  Eq.:  +1.398
       0.896 0.413 0.025 - 0.104 0.008 0.000 CL  +1.398 CF  +1.398
      [0.002 0.003 0.001 - 0.002 0.001 0.000 CL   0.007 CF   0.007]

    2. Rollout          23/17*/5 22/16               Eq.:  +1.397 ( -0.001)
       0.896 0.413 0.024 - 0.104 0.008 0.000 CL  +1.397 CF  +1.397
      [0.002 0.003 0.001 - 0.002 0.001 0.000 CL   0.007 CF   0.007]

    3. Rollout          23/17*/11 22/10              Eq.:  +1.393 ( -0.005)
       0.893 0.416 0.023 - 0.107 0.008 0.000 CL  +1.393 CF  +1.393
      [0.002 0.003 0.001 - 0.002 0.001 0.000 CL   0.007 CF   0.007]

    4. Rollout          23/17*/11 22/16(2)           Eq.:  +1.368 ( -0.030)
       0.902 0.384 0.023 - 0.098 0.007 0.000 CL  +1.368 CF  +1.368
      [0.002 0.003 0.001 - 0.002 0.001 0.000 CL   0.006 CF   0.006]

    5. Rollout          23/17* 22/10 22/16           Eq.:  +1.367 ( -0.031)
       0.904 0.382 0.023 - 0.096 0.008 0.000 CL  +1.367 CF  +1.367
      [0.002 0.003 0.001 - 0.002 0.001 0.000 CL   0.006 CF   0.006]

    6. Rollout          22/10(2)                     Eq.:  +1.306 ( -0.092)
       0.903 0.348 0.015 - 0.097 0.013 0.000 CL  +1.306 CF  +1.306
      [0.002 0.003 0.001 - 0.002 0.002 0.000 CL   0.007 CF   0.007]
        Full cubeful rollout with var.redn.
        1296 games, Mersenne Twister dice gen. with seed 872503493 and quasi-random dice
        Play: 0-ply cubeful prune [expert]
        Cube: 0-ply cubeful prune [expert]

[close]

[diane]

How do you get reports of the rollout like that out of gnu...a 2 ply rollout seems to be taking 10 mins on this PC, but I dont know where the report goes..it does it..I click ok and it all vanishes..it does change what I see on 'hint' though.

I could offer help with these now...if I knew how to get the nice reports..

[NIHILIST]

A play that leaves a checker in the home board, hit or not, does a couple things; it can slow down skaters advance when the blot is hit and put on the bar.....hes only playing part of his next rolls......and it gives the opportunity to hit more checkers, increasing the gammon vig.

Bob

[ah_clem]

2-ply rollout

Spoiler

This just confirms the 0-ply rollout.

Note that not hitting wins the most games when rolled out at this score
- this implies that not hitting is correct at DMP, but we'd need a
rollout at that score to be "sure".  Since this 2-ply rollout took a
couple of days on my ancient home PC, I'll leave that to someone else.




1. Rollout          23/17*/11 22/10   Eq.:  +1.403
       0.889 0.423 0.029 - 0.111 0.006 0.000 CL  +1.403 CF  +1.403
      [0.002 0.003 0.001 - 0.002 0.002 0.000 CL   0.006 CF   0.006]

    2. Rollout          23/17* 22/4       Eq.:  +1.398 ( -0.005)
       0.886 0.422 0.029 - 0.114 0.007 0.000 CL  +1.398 CF  +1.398
      [0.002 0.003 0.001 - 0.002 0.002 0.000 CL   0.006 CF   0.006]

    3. Rollout          23/17*/5 22/16    Eq.:  +1.397 ( -0.005)
       0.888 0.420 0.028 - 0.112 0.006 0.000 CL  +1.397 CF  +1.397
      [0.002 0.003 0.001 - 0.002 0.002 0.000 CL   0.006 CF   0.006]

    4. Rollout          23/17* 22/10 22/16  Eq.:  +1.374 ( -0.029)
       0.898 0.394 0.023 - 0.102 0.006 0.000 CL  +1.374 CF  +1.374
      [0.002 0.003 0.001 - 0.002 0.001 0.000 CL   0.005 CF   0.005]

    5. Rollout          23/17*/11 22/16(2)  Eq.:  +1.373 ( -0.030)
       0.900 0.390 0.024 - 0.100 0.006 0.000 CL  +1.373 CF  +1.373
      [0.002 0.003 0.001 - 0.002 0.001 0.000 CL   0.005 CF   0.005]

    6. Rollout          22/10(2)        Eq.:      +1.329 ( -0.074)
       0.908 0.356 0.016 - 0.092 0.008 0.000 CL  +1.329 CF  +1.329
      [0.002 0.003 0.001 - 0.002 0.001 0.000 CL   0.005 CF   0.005]
        Full cubeful rollout with var.redn.
        1296 games, Mersenne Twister dice gen. with seed 872503493 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]