come under the gun?
Kit Woolsey
Peter Fankhauser (fank...@darmstadt.gmd.de) wrote:
: +24-23-22-21-20-19-+---+18-17-16-15-14-13-+
: | O X X | | X X X O |
: | O X X | | X O |
: | X | | X O |
: | | | O |
: | | | | +---+
: | | | | | 1 |
: | O | | | +---+
: | O | | |
: | O | | X |
: | X O | | O O X |
: | X O | | O O X |
: +-1--2--3--4--5--6-+---+-7--8--9-10-11-12-+
: O: 2-away, X: 5-away
: O to play 3-2:
: Options are:
: 6/3, 13/11
: 24/21, 13/11
: 13/8
: The correct play here may be automatic for experts, but it surprised me.
: O, one of my favorite opponents on fibs, did choose the correct play -
: at least for moneygame. I have done jellyfish level 6 rollouts, which I will post
: in a couple of days.
: What do you think? Split and come under the gun, or wait and see?
: Does the score have an influence here? Does X have a double after
: any of the plays?
: funk
Coming under the gun of four builders when the bad guy has the stronger
board looks pretty scary. However, I believe it is the correct play.
When your opponent has an advanced anchor and the timing is otherwise
about even, it is very important for you to attempt to get an advanced
anchor of your own or free one of your back men. If you sit on the 24
point, you eventually get crushed in the timing battle. O won't like it
if he is pointed on, but otherwise he has a fighting chance. If O doesn't
split, X has a lot of good rolls which make the four, five, or bar
points, and it will then be harder for O to do anything. After the
split, X's only really good rolls are those which point on the blot.
Note the critical conditions which must exist for such a play to be correct:
1) X must have an advanced anchor. If X's back men were back on O's ace
point, then splitting would be folly. That would just play into X's
strength -- the stronger board and the builders. Instead O should
concentrate on building up the offense and winning a priming battle.
2) The overall timing must be about even. If X were well ahead in the
race, O would have a better chance by staying back, building up the
offense, and hoping to pounce when X is forced off the anchor. In this
position, if O stays back the timing figures to go against him.
As for the cube: At the match score, X clearly has a HUGE double after
any play. I would double any play for money, although admittedly this
might be a bit premature. O has a trivial take for money, but at the
match score with no recube vig and gammons counting against but not for
him the take looks marginal at best -- I think it is a very close call.
Kit
Peter Fankhauser
Peter Fankhauser
Peter Fankhauser <fank...@darmstadt.gmd.de> wrote:
>
> +24-23-22-21-20-19-+---+18-17-16-15-14-13-+
> | O X X | | X X X O |
> | O X X | | X O |
> | X | | X O |
> | | | O |
> | | | | +---+
> | | | | | 1 |
> | O | | | +---+
> | O | | |
> | O | | X |
> | X O | | O O X |
> | X O | | O O X |
> +-1--2--3--4--5--6-+---+-7--8--9-10-11-12-+
>
>
> O: 2-away, X: 5-away
> O to play 3-2:
>
(Jellyfish 2.01, level 6 rollouts with Jellyfish 2.01,
432 games, sd approx. 0.12)
2ply Rollout
24/21, 13/11: -0.382 -0.354
13/11, 6/3 -0.429 -0.433
So O's and Kit's choice are clearly superior. Kit
(thanks a lot for the illuminating analysis) gave
as the two necessary conditions for the split, inspite
of all the gearing of X, (a) X's advanced anchor and (b)
X's only small lead in the race.
And indeed when X anchors only on the 21pt instead of the 20pt,
Jellyfish 2ply prefers building/slotting plays over the split.
7/4, 6/4: -0.266
24/21, 13/11: -0.285
(I have not tried it out, but I'm sure the building play 13/11,
13/10 would come out best when X would anchor on the 24pt)
As for the race, X's lead needs to be pretty much larger to render
the split wrong. I moved back one after the other of O's checkers,
and only when there were 2 of O's checkers on the midpoint instead
of the 6pt, and one of X's checkers on the 19pt instead of the
the 12pt - all in all this makes 21 pips more) the split
became inferior:
13/10, 13/11: -0.438
13/8: -0.443
24/21, 13/11: -0.499
Btw., I have found this "what if" method, i.e., varying a position
a little until another checkerplay becomes superior, fairly
helpful, when I can't explain a jellyfish preference otherwise.
funk
Kit Woolsey
Peter Fankhauser (fank...@darmstadt.gmd.de) wrote:
: Btw., I have found this "what if" method, i.e., varying a position
: a little until another checkerplay becomes superior, fairly
: helpful, when I can't explain a jellyfish preference otherwise.
: funk
Fairly helpful is quite an understatement. I think this approach is the
best way to learn about a position. When I come across a position with
two very different plays which have about equal equities in rollouts, I
know that I have some valuable information. I have a benchmark position
for the concepts involved. Now I take that position, make small
modifications, and see what the relevant factors which favor one play or
the other from the original position are. It is from analyses such as
these that I develop my weights and priorities which I use to choose my
moves in similar positions.
Kit