Game 4 Move 4 Forum 6-2..and..

[roadkillbooks]

Blitzxz took and Forum rolled 62 which I am nexting...with bar/23, 21/15*..stayed tuned for his roll

[roadkillbooks]

Blitzxz danced with 5-3

[roadkillbooks]

Ok forum rolled another 62 and Im through up a poll. Anything missing let me know.

[stiefnu]

Anything missing let me know.

Along with a diagram, we used to get info on what colour we're playing, the pip count and the match score - all of which, to varying degrees, I found rather helpful.  Any chance of restoring these useful features?  TIA, Steve

[roadkillbooks]

Forum is white. Blitzxz is brown.  The score is blitzxz 2- 1 Forum.  My pictures dont show pip count, while Im doing this it might be a good exercise to count the pips. Ill remeber to put the match score up on every roll.

thx rodkillbooks

[dorbel]

Spoiler

13/11*, 8/2* is the automatic play for me. It's hard to believe that any other play is getting votes! We should hit here even if we knew for certain that brown was about to roll a 2, just to stop him from anchoring! If he can anchor, he's in the game until the end and we will have all the problems associated with bearing in and bearing off against that anchor. If we can close him out with three on the roof he is dead.

[close]

[ah_clem]

Spoiler

I'm too late for the vote, but whacking both checkers seems to be the right approach.  

<ramones>

  Hey! Ho!

  Let's GO!

  Hey! Ho!

  Let's GO!

  They're forming in a straight line
  They're going through a tight wind
  The kids are losing their minds
  The Blitzkrieg Bop

  They're piling in the back seat
  They're generating steam heat
  Pulsating to the back beat
  The Blitzkrieg Bop.

  Hey ho, let's go
  Shoot'em in the back now
  What they want, I don't know
They're all reved up and ready to go
</ramones>

In some situations, hitting a third checker is unnecessary, but in this case it makes sense since we're actually doing something constructive with each hit.  Hitting loose on the 2 helps prevent the anchor and hitting on the 11 brings down another builder.  If there was no blot on the 11 I might still play 13/11 to continue the blitz theme.

We have a three point board against a one point board.  If this isn't blitzing territory, nothing is.

[close]

[diane]

Sure is, I couldn't work out why others were voting differently either - would any of them like to say why?

[roadkillbooks]

I voted for 13/7, 13/11* but i didnt have time to talk about it.  Simply i think it over plays the position unecessarily.  We have very few pieces ïn the zone... if we are hit with a 2 we cant carry out the blitz so we may as well bring pieces down and if blitzxz anchors we can play positionally cause we werent gonna be able to blitz well anyway.

[ah_clem]

rollout:

Spoiler

gnubg prefers to hit only the checker on the 11 point.  Hitting both is not a blunder, but is still sub-optimal. However, hitting both is superior to playing passively with the 5 (ie 21/15 or 22/16).  The "best" play 15/9 13/11*  garnered no votes.

So, why hit only one?  A post-hoc explanation would go something like this: 
1) It's not usually necessary to send a third checker back, the race gain is often outweighed by the strategic liability of giving your opponent another man behind your lines. 

2)Holding the eight point which is 6 points away from the opponent's checker forms a fine block, so abandoning it is not favored. 

3)The three plays gnubg favors all bring another builder into range - the builder on the 11 with our play is too far from the action.

Still, I don't feel bad about the double-hit.  It does win the most gammons, and gammons are particularly valuable for us.



    1. Rollout          15/9 13/11*                  Eq.:  +0.839
       0.693 0.298 0.060 - 0.307 0.076 0.010 CL  +0.904 CF  +0.839
      [0.002 0.003 0.004 - 0.002 0.003 0.002 CL   0.008 CF   0.009]

    2. Rollout          13/11* 13/7                  Eq.:  +0.833 ( -0.006)
       0.688 0.303 0.065 - 0.312 0.081 0.015 CL  +0.898 CF  +0.833
      [0.002 0.003 0.004 - 0.002 0.003 0.002 CL   0.009 CF   0.009]

    3. Rollout          13/11*/5                     Eq.:  +0.803 ( -0.035)
       0.682 0.294 0.057 - 0.318 0.080 0.013 CL  +0.866 CF  +0.803
      [0.002 0.003 0.004 - 0.002 0.002 0.002 CL   0.009 CF   0.009]

    4. Rollout          13/11* 8/2*                  Eq.:  +0.789 ( -0.049)
       0.660 0.321 0.060 - 0.340 0.094 0.012 CL  +0.869 CF  +0.789
      [0.002 0.003 0.004 - 0.002 0.003 0.002 CL   0.009 CF   0.009]

    5. Rollout          21/15 13/11*                 Eq.:  +0.766 ( -0.072)
       0.689 0.269 0.050 - 0.311 0.074 0.009 CL  +0.843 CF  +0.766
      [0.002 0.003 0.004 - 0.002 0.002 0.001 CL   0.008 CF   0.009]

    6. Rollout          22/16 13/11*                 Eq.:  +0.745 ( -0.094)
       0.688 0.260 0.058 - 0.312 0.078 0.009 CL  +0.824 CF  +0.745
      [0.002 0.003 0.004 - 0.002 0.003 0.002 CL   0.008 CF   0.008]
        Full cubeful rollout with var.redn.
        1296 games, Mersenne Twister dice gen. with seed 843463219 and quasi-random dice
        Play: 0-ply cubeful prune [expert]
        Cube: 0-ply cubeful prune [expert]

[close]

[blitzxz]

I have to say that playing blitz is very difficult when there is lots of checkers far behind. Forum has the difficult task of keeping the balance between continuing attack and bringing enough checkers to cover the blots and at some point it might be time to change for more positional priming/outfield control type of game. So I'm glad that I took and I can just watch how much match winning chances forum loses.  

[dorbel]

Fair enough, I suppose seven blots is a lot of weight to carry! With hindsight, one can see that the more modest 15/9, 13/11* is quite attractive, aiming at the important 4pt and keeping the 8pt, which is part of prime field at present. This allows for switching into a priming game or perhaps continuing the blitz as the dice dictate. Nonetheless, I am not entirely convinced by the gnu rollout, 0-ply might not be good enough for this and I will try one of my own.

[ah_clem]

I make no claim that my 0-ply rollouts are authoritative so I'd welcome a more precise simulation.  0-ply is what my aging home PC can crank out in a reasonable amount of time (~90 minutes), so that's what I do.

Please post the results when your rollout is complete.

[dorbel]

A 2-py rollout, cubeless, 1296 games, gives this result.
13/11*, 13/7 wins 70%, including 30 gammons. emg 0.934 or 58.03% MWC.
15/9, 13/11* wins 69.6% including 29.3 gammons, emg 0.910 or 57.53% MWC.
13/11*, 8/2* wins 66.7% including 32.1 gammons,  emg 0.897 or 57.27% MWC.

For those not familiar with these terms, emg means equivalent to money game. This is not exactly the same equity as you would have in a money game, but is adjusted to allow for the score. MWC is match winning chances.

So if this rollout is accurate, which it probably is, then we can learn several things.  The number of blots doesn't matter, as 13/11*, 13/7 leaves 7 blots, the same as the double hit.
Bringing a checker to the 7 pt is more valuable than one on the 9pt. Why? Because on the 7pt it has a dual function, acting both as a builder and as a slot for the 7pt which will be an important part of the prime field if that is the way that the game goes. Not only that, it bears on every point in the inner board. By contrast a checker on the 9pt is not part of the prime field (it can't form a prime including the 3pt) and doesn't bear on the 1 and 2pts.
You can't say that any of these plays is a bad play and I can't see any way of choosing between these over the board except by instinct. However, we have learned something here and remembering what we have found out in this position can only help if we come across something similar in the future.

[ah_clem]

Interesting.  The 0-ply rolllout had the double hit as a .035 error, while your 2-ply rollout has it as a .037 error.  That's pretty close.

More interesting how the 2-ply rollout ranks the top two moves.  0-ply has them as a statistical tie (.006 difference with a .009 confidence interval).  2-ply teases out that moving to the bar point is the better move.