duplicate?

[tiredretired]

Here is a an interesting position I found on Mochy's blog.

The score (after 0 games) is: sayUncle 0, karush54 0 (match to 11
points)
Match Information:

Date: November 15, 2011
Move number 6: X to play 35

GNU Backgammon Position ID: bdsIAwT3fQAMAA
Match ID : UYl1AQAAAAAA
+13-14-15-16-17-18------19-20-21-22-23-24-+ O: sayUncle
| X O | | O O O O O O | 0 points
| X | | O O O O O |
| | | |
| | | |
| | | |
v| |BAR| | 11 point match
| | | X X |
| | | X X |
| | | X X X |
| O | | X X X | Rolled 35
| O | | O X X X | 0 points
+12-11-10--9--8--7-------6--5--4--3--2--1-+ X: karush54 (Cube: 2)
Pip counts: O 97, X 56

[badgolferman]

I play 14/6. That leaves only a combination of 2 that can get me.

[Bradley K. Sherman]

tiredretired <ttexas...@gmail.com> wrote:
> +13-14-15-16-17-18------19-20-21-22-23-24-+ O: sayUncle
> | X O | | O O O O O O | 0 points
> | X | | O O O O O |
> | | | |
> | | | |
> | | | |
>v| |BAR| | 11 point match
> | | | X X |
> | | | X X |
> | | | X X X |
> | O | | X X X | Rolled 35
> | O | | O X X X | 0 points
> +12-11-10--9--8--7-------6--5--4--3--2--1-+ X: karush54 (Cube: 2)
>Pip counts: O 97, X 56

14/6, based on duplication of 2's.

--bks

[Paul]

On Nov 15, 1:35 pm, b...@panix.com (Bradley K. Sherman) wrote:

Yes, this is the natural play over the board, minimizing shots. But
this can't be correct because it would then make a very boring blog,
and would therefore never have been posted on the blog. Even if there
were such a boring blogpost, it would not have made it onto rgb.

14-9, 14-11 gives more immediate shots but is much safer if O misses.
It's not that surprising to me if the enhanced safety-after-missing
compensates for the greater immediate risk.

With blatant QF, I'm going for 14-9, 14-11.

Paul Epstein

[Bradley K. Sherman]

In article <61eb6d3a-4a7a-4ffe...@w3g2000vbw.googlegroups.com>,

Paul <peps...@gmail.com> wrote:
>
>Yes, this is the natural play over the board, minimizing shots. But
>this can't be correct because it would then make a very boring blog,
>and would therefore never have been posted on the blog. Even if there
>were such a boring blogpost, it would not have made it onto rgb.
>

Right, that thought did jump into my head. But I find that
being humiliated in public is a boon to my education.

--bks

[Walt]

14/6 duplicates 2's and leaves any 2 (12 rolls) or 10 (three rolls) to
hit. That's 15 shots.

14/11 14/9 leaves 5's (15 rolls) and 9's (3 rolls) to hit. That's 18
shots.

It seems like duplication is the way to go, but QF says otherwise. Is it
worth leaving three extra shots to make the next roll slightly easier to
clean up? I'm not seeing it.

//Walt

--
//Walt

[Walt]

On 11/15/2011 8:57 AM, Bradley K. Sherman wrote:

> Right, that thought did jump into my head. But I find that
> being humiliated in public is a boon to my education.
>

Exactly. You only learns something from these problems when you get it
wrong. I'm proud to say that I do a lot of learning here. (c:


--
//Walt

[Tim Chow]

On Nov 15, 4:36 am, tiredretired <ttexashol...@gmail.com> wrote:
>  GNU Backgammon  Position ID: bdsIAwT3fQAMAA
>                  Match ID   : UYl1AQAAAAAA
>  +13-14-15-16-17-18------19-20-21-22-23-24-+     O: sayUncle
>  |    X     O       |   | O  O  O  O  O  O |     0 points
>  |    X             |   | O  O  O  O  O    |
>  |                  |   |                  |
>  |                  |   |                  |
>  |                  |   |                  |
> v|                  |BAR|                  |     11 point match
>  |                  |   |          X  X    |
>  |                  |   |          X  X    |
>  |                  |   |          X  X  X |
>  | O                |   |          X  X  X |     Rolled 35
>  | O                |   |       O  X  X  X |     0 points
>  +12-11-10--9--8--7-------6--5--4--3--2--1-+     X: karush54 (Cube: 2)
> Pip counts: O 97, X 56

The first thing to do, of course, is to count shots. I get 15 shots
after 14/6 (one double hit) and 19 shots after 14/11 14/9 (two double
hits). If we're hit then I think the difference between the two plays
is negligible.

This is only the first step, though. We have to assess the position
if O doesn't hit. Because of O's strong board and huge race deficit,
O will not be afraid to leave blots everywhere to maximize contact.
Let's assume that O, with her non-hitting roll, leaves the checkers on
X's side of the board untouched (this will not always be the case, but
it should cover many cases). What is the situation now on X's next
roll? After 14/6, X will clean up entirely with 66 62 55 53 44 (7
rolls), and leave only an indirect shot with 43 32 22 21 11 (8
rolls). The remaining 21 rolls will leave O a direct shot. After
14/11 14/9, X will clean up entirely with 66 64 55 53 42 31 22 21 11
(14 rolls), and leave only an indirect shot with 62 61 52 51 43 (10
rolls). The remaining 11 rolls will leave O a direct shot.

In effect, after 14/6, X is not just leaving 15 immediate shots; about
2/3 of the remaining 22 rolls are also effectively direct shots if you
look ahead to the next turn. So instead of getting hit 15/36 times, X
will be hit something like 20/36 times. After 14/11 14/9, the same
sort of calculation gives us about 21/36 hits. So it's looking pretty
close from this point of view, with maybe a slight edge to 14/6.

What other considerations are there? Some of the above shot-leaving
rolls actually hit O and put O on the bar. There are a few more of
these after 14/11 14/9 than after 14/6. These are good for X's gammon
chances. What about the chances of X getting gammoned? This will
usually happen only if both of X's blots get picked up. It's hard to
calculate this but I think it will happen more often after 14/6
because O will usually have more opportunities to pick up the second
blot (she'll hit X's checker on the 6pt first of course if she can).

The plays look close but I would try 14/11 14/9 because of the gammon
situation. At DMP I think I would try 14/6.

---
Tim Chow

[Tim Chow]

On Nov 15, 4:36 am, tiredretired <ttexashol...@gmail.com> wrote:

> Here is a an interesting position I found on Mochy's blog.

I doubt that any more rec.games.backgammon readers will comment at
this stage, so can you please tell us what Mochy and/or the computer
said?

---
Tim Chow

[tiredretired]

yes ofcourse, sorry :-)

He suggests that the cube action scenario following the play is key.

Here is the blog article link
http://bgmochy.blogspot.com/2011/02/very-rare-position.html

Here are a couple more YouTube video links to Mochy wisdom.
http://www.youtube.com/watch?v=8SH0GfgDjZY cube handling seminar
http://www.youtube.com/watch?v=lD_TzrsW2nM prime vs blitz lesson

[tiredretired]

my version of the default 0 ply setting for gnubg rollouts didnt catch
this play but if strengthened to 2 ply rollout (slow), gnubg does make
the 2 checkers down play.

[Paul]

On Nov 17, 12:25 am, tiredretired <ttexashol...@gmail.com> wrote:
> ....

> He suggests that the cube action scenario following the play is key.
>
> Here is the blog article linkhttp://bgmochy.blogspot.com/2011/02/very-rare-position.html

> ...

I see that the recommended play makes it easier to clean up safely if
the opponent misses, but at the cost of allowing more immediate
shots. Thus it doesn't seem implausible for the recommended play to
come out on top. However, my reaction is absolutely identical to
Walt's: "Is it worth leaving three extra shots to make the next roll
slightly easier to clean up? I'm not seeing it." [Actually, there are
four extra shots and not three but Walt's response seems spot-on and
is a great guide to the vast majority of similar positions].

The reference to the cube is interesting, and it took me a while to
get it. The point is that the upside of the recommended play (the
greater safety-after-missing) is magnified by a further turning of the
cube, but the downside of the recommended play (the greater number of
shots) does not get magnified by the cube. Thus the upside of the
recommended play outweighs the downside.

At DMP, 14-6 is probably correct by a big margin.

Paul Epstein

[Tim Chow]

On Nov 16, 7:25 pm, tiredretired <ttexashol...@gmail.com> wrote:
> He suggests that the cube action scenario following the play is key.
>
> Here is the blog article linkhttp://bgmochy.blogspot.com/2011/02/very-rare-position.html

Very interesting! The cube action does look like the crucial factor.
I tried rolling it out with the cube on the other side and 14/6 comes
out clearly on top. (Results not posted here.)

However, there is one point on which I believe Mochy is mistaken. He
says that 14/6 loses fewer gammons. By this, he presumably means that
14/6 loses fewer *cubeless* gammons, and he is basing this statement
on the pseudocubeless numbers reported by the bot. However, those
pseudocubeless numbers are obtained from a cubeful rollout and cannot
be trusted to give an accurate estimate of how many gammons would be
won or lost if the game were actually played without a cube.

I used eXtreme Gammon to perform a cubeless rollout (by setting the
maximum cube value to "cubeless," and additionally checking the
"cubeless" option in the rollout dialog box). The results are shown
below. According to this cubeless rollout, 14/6 loses more gammons
than 14/11 14/9. Variance reduction doesn't apply to gammon estimates
so even though this is a 10K rollout, I think the 95% confidence
interval would put the gammon losses for 14/6 in the approximate range
13.6% to 14.6%, and similarly for 14/11 14/9 the range would be 12.8%
to 13.8%. Still, the evidence seems pretty good that 14/6 loses more
gammons.

XGID=-CEEa-------b-B-a--bbbbba-:0:0:1:53:0:0:3:0:0

X:Player 1 O:Player 2
Score is X:0 O:0. Money session
+13-14-15-16-17-18------19-20-21-22-23-24-+

| X O | | O O O O O O |
| X | | O O O O O |
| | | |
| | | |
| | | |

| |BAR| |

| | | X X |
| | | X X |
| | | X X X |
| O | | X X X |

| O | | O X X X |

+12-11-10--9--8--7-------6--5--4--3--2--1-+
Pipcount X: 56 O: 97 X-O: 0-0
Cube : 1
X to play 53

1. Rollout: 14/6 eq:-0.158
Player : 42.10% (G:8.59% B:0.00%)
Opponent: 57.90% (G:14.12% B:0.22%)
Confidence: ±0.001 (-0.159<E<-0.157)

2. Rollout: 14/11 14/9 eq:-0.189 (-0.031)
Player : 40.53% (G:8.01% B:0.01%)
Opponent: 59.47% (G:13.27% B:0.18%)
Confidence: ±0.001 (-0.190<E<-0.188)

10368 Games rolled with Variance Reduction.
Dice Seed: 2
Cubeless
Moves: 3 ply

eXtreme Gammon Version: 1.21

---
Tim Chow