Game 4 Move 9 Forum 2-6

[roadkillbooks]

OK after our double hit..
blitzxz rolled 62 bringing in one and hitting us with bar/23*

Forum then rolled its own 62. We enter with bar/23.. and now a 6 to move

[ah_clem]

Spoiler

Make the 15 point.

We don't need the 21 point, and this position gives him an awful lot of awkward rolls if he enters.

This is one of the weirdest positions I've seen in awhile. It's like we're playing a backgame and a blitz at the same time.

[close]

[diane]

Spoiler

Make the 15 point.

We don't need the 21 point, and this position gives him an awful lot of awkward rolls if he enters.

This is one of the weirdest positions I've seen in awhile. It's like we're playing a backgame and a blitz at the same time.

[close]

  and agree with him, all at the same time!!

[roadkillbooks]

8-0 and 20 hours seems to be a threshold..moving it along..

[ah_clem]

rollout

Spoiler

Of the four legal moves, three are pretty reasonable. Only the over aggressive 13/7 is a blunder - it leaves a  double shot on the seven and two other blots - but the others play pretty well.

gnubg agrees with our idea of making the 15 point, but somewhat surprisingly 20/14 is almost within the margin of error.  Why?  Perhaps the additional flexibility balances the double shot, and we still retain two anchors.  If we end up playing a backgame, holding the 2 and 4 points is better than the 2 and 5 ( I think this is true, but I didn't look it up - 1 and 5 is unplayable, 2 and 5 isn't as bad, but I seem to recall that 2 and 4 is better)

I thought bringing down a builder with 15/9 was reasonable but not as good ass making the 15.  Gnubg bears this out, but it's closer than I expected.



    1. Rollout          bar/23 21/15                 Eq.:  +0.151
       0.530 0.151 0.019 - 0.470 0.145 0.012 CL  +0.255 CF  +0.151
      [0.002 0.003 0.002 - 0.002 0.004 0.002 CL   0.008 CF   0.009]

    2. Rollout          bar/23 20/14                 Eq.:  +0.140 ( -0.012)
       0.522 0.160 0.026 - 0.478 0.156 0.023 CL  +0.240 CF  +0.140
      [0.003 0.003 0.003 - 0.003 0.004 0.003 CL   0.009 CF   0.010]

    3. Rollout          bar/23 15/9                  Eq.:  +0.134 ( -0.017)
       0.521 0.155 0.023 - 0.479 0.149 0.015 CL  +0.239 CF  +0.134
      [0.003 0.003 0.003 - 0.003 0.004 0.002 CL   0.009 CF   0.010]

    4. Rollout          bar/23 13/7                  Eq.:  +0.033 ( -0.118)
       0.491 0.143 0.019 - 0.509 0.181 0.018 CL  +0.136 CF  +0.033
      [0.003 0.003 0.002 - 0.003 0.004 0.002 CL   0.010 CF   0.010]
        Full cubeful rollout with var.redn.
        1296 games, Mersenne Twister dice gen. with seed 845045210 and quasi-random dice
        Play: 0-ply cubeful prune [expert]
        Cube: 0-ply cubeful prune [expert]

[close]

[dorbel]

Yes, 21 and 23pts is a viable backgame plan, 20 and 23 isn't, although slightly superior to 20 and 24 as ah clem points out. 20/14 and 15/9 do well because they slightly increase our coverage of the outfield where we are desperate to make a point. Playing with no outfield points is extraordinarily difficult as you have no safe place(s) to land as you bring checkers around. Making the 15pt is interesting, as it frees up the checkers on the 13pt to play. In effect the 15pt becomes a sort of pseudo midpoint!

[ah_clem]

Good point about how 13/7 leaves us with no outfield point.  Now that you mention it, I'm sure that this strategic consideration is why it fares so poorly in gnubg's estimation.  Without an outfield point, our army is split into two islands at opposite ends of the board.  Not good.

I have to pick a nit about the use of the word slightly in "20 and 23 isn't [a viable backgame plan], although slightly superior to 20 and 24"   In Walter Trice's book he gives examples of different backgames using snowie rollouts to evaluate.  In his examples, 20 & 24 (or 5 & 1 as he and I like to call it) is much poorer than 20 & 23 (5 & 2):

20 & 24 ( 5 & 1)
Wbg   Wg     W              L      Lg     Lbg
1.9   26.2   75.6           24.4   0.9    0.0

20 & 23 (5 & 2)
Wbg   Wg     W              L      Lg     Lbg
0.4   23.7   61.4           38.6   4.4    0.1

That's more than a "slight" difference - it's double/pass vs no double.   20 & 23 (5 & 2) is actually quite playable.

More info and diagrams  here.. Unfortunately, no consideration of the 21 & 23 (4 & 2) backgame.  Anybody know where this ranks?  I'd assume it's better than  5 & 2, but I don't know for sure

[dorbel]

Very interesting, thank you. The difference between the two positions is of course that the author has transposed the X and O checkers on the 23 and 24 points. Superficially this appears to let us compare a 24/20 game with a 23/20 game, but I am not sure how valid this methodology is. Every position is different and without being able to put values on the difference in the race and in the timing we can't IMO draw conclusions with the degree of certainty assigned to them here. However, I accept the main thrust of your argument, which is that "slightly" is not accurate and "considerably" is more valid. The fact that X has to clear the 8 and 7 pts against both anchors is important, as well as the lower gammon rates when O is reduced to a 23pt anchor, which he often is.
Walter's guideline that one doubles a well timed backgame when on the verge of clearing the third point in front of the backgame is good. Easy though to construct such a position against 23 and 20 which is a very strong double, although they will usually be takes with enough timing.

[ah_clem]

If it's any consolation, I also thought 5-2 was fairly close to 5-1 until someone set me straight recently.  

Here's a link to a comparison of the relative merits of each structure, but it doesn't give much detail as to how they arrived at the numbers, other than to say that they were calculated by Snowie.

4-2 is better than 5-2, but not by all that much. 5-1 is much worse than any other.

The relavent chart:

Anchor Avg. pts lost  proper cube action   

2-1      1.03         Double       Pass
3-1      0.80         No double    Take
4-1      0.94         Double       Take
5-1      1.24         Double       Pass
3-2      0.58         No double    Take
4-2      0.71         No double    Take
4-3      0.56         No double    Take
5-2      0.80         No double    Take
5-3      0.86         Double       Pass
5-4      0.88         Double       Pass