Lessons from the board -- #2
Michael J. Zehr
+24-23-22-21-20-19-------18-17-16-15-14-13-+
| O O O O | | X X |
| O O O O | | X |
| O O O | | | S
| O | | | n
| 6 | X | | o
| |BAR| | w
| | | | i
| | | | e
| X | | |
| X X X | | X X |
| X O X X | | X X |
+-1--2--3--4--5--6--------7--8--9-10-11-12-+
Pipcount X: 133 O: 55 X-O: 0-1/5 (1)
CubeValue: 2, O owns Cube
After bar/19, X has two choices for the 3: 19-16 or 17-14. (13-10
leaves a direct shot and should be discarded.)
The normal guideline is to stay back as far as possible with 17-14.
Here that is the wrong approach, however. The more precise strategy is
to try to aim for a spot 11 pips away from O's back checker. The
closest X can come to that is with 19-16.
One can also do some quick counting: 17-14 is worse after a 64, better
after a 66, and worse after a 65. While 17-14 brings that checker into
direct range of the 10, 19-16 also brings a checker into direct range of
the 10 (plus giving better coverage of the 11 and 12).
Containing a back checker after a hit is difficult and requires a lot of
focus and concentration. While this might look like a play where it
doesn't make a lot of difference, that's deceptive. The wrong play
loses 4% more games than the right one!
There are two keys to this position -- realizing what the right move is
and why, and realizing that X should be thinking carefully about every
play.
bar/16 Eq.: -0.136/-.039
bar/19 17/14 Eq.: -0.254/-.124
Rollouts by both neural nets comfirms a roughly .100 equity difference
between the two plays.
-Michael J. Zehr
bshe...@my-dejanews.com
"Michael J. Zehr" <mich...@michaelz.com> wrote:
> +24-23-22-21-20-19-------18-17-16-15-14-13-+
> | O O O O | | X X |
> | O O O O | | X |
> | O O O | | | S
> | O | | | n
> | 6 | X | | o
> | |BAR| | w
> | | | | i
> | | | | e
> | X | | |
> | X X X | | X X |
> | X O X X | | X X |
> +-1--2--3--4--5--6--------7--8--9-10-11-12-+
> Pipcount X: 133 O: 55 X-O: 0-1/5 (1)
> CubeValue: 2, O owns Cube
> X to play (6 3)
>
> After bar/19, X has two choices for the 3: 19-16 or 17-14. (13-10
> leaves a direct shot and should be discarded.)
>
> The normal guideline is to stay back as far as possible with 17-14.
> Here that is the wrong approach, however. The more precise strategy is
> to try to aim for a spot 11 pips away from O's back checker. The
> closest X can come to that is with 19-16.
Excellent analysis, Michael.
I have found two principles extremely helpful in containment positions.
Applying these principles will usually produce the correct moves.
The most important priority in a containment position is to hit the enemy
man. This priority works towards the goal of reducing the number of
opportunities the opponent has of rolling a 6-6 to win the game outright. You
may have a judgment call if the opponent cannot be hit safely, but the
principle of "when in doubt, hit" applies here; you should hit unless you are
certain that hitting is wrong.
If you cannot hit an enemy man, then your goal is to prevent escape. There are
two tactical goals in this strategy. You should make points close to the enemy
man, and keep blots far away (12 or more pips). Nearby points directly impede
the enemy's progress, but nearby blots do not. The opponent is more likely to
hit or skip over a nearby blot than he is to offer a direct shot to it. Nearby
blots should be moved into attacking position, so that you can attack on
subsequent turns if the opponent does not escape this turn.
These principles cover the majority of containment situations. However, there
are plenty of times when priorities conflict. In those situations you have to
fall back on careful calculation of variations, as Michael pointed out in his
original post.
Warm regards,
Brian Sheppard
(BTW: the neural networks I am familiar with do not play according to these
principles unless they search at least 3 ply ahead. You can think of these
principles as encapsulating the understanding of a 3-ply lookahead.)
-----== Posted via Deja News, The Leader in Internet Discussion ==-----
http://www.dejanews.com/rg_mkgrp.xp Create Your Own Free Member Forum