End game position: Move unclear
omegaman
of Snowie (3.0). There was an end game holding position that I'm not
clear on when I rolled double ones. See position & match score below
along with my notes on the position. Does anyone have an answer? If
you have more recent versions of either Snowie or Gnubg maybe you can
run the position through and see if the equities are different.
--------------------------------------------------------------------
| Captain_America (X) vs. clay (O) |
| 7 point Match |
--------------------------------------------------------------------
Match to 7. Score X-O: 3-4
X to play (1 1)
+24-23-22-21-20-19-------18-17-16-15-14-13-+
| O O O O O O | | X |
| O O O O | | X |
| O |
| | S
| O |
| | n
| |
| | o
| |
| | w
| |
| | i
| |
| | e
| X X | |
O |
| X X X X X | | O
|
| X X X X X | | X O |
+-1--2--3--4--5--6----------7---8---9--10--11--12-+
Pipcount X: 78 O: 77 X-O: 3-4 of 7 (1)
CubeValue: 2, X owns Cube
1. 2 18/17(2) 5/4 2/1 Eq.: 0.446
0.0% 0.7% 47.4% 52.6% 0.6% 0.1%
2. 2 18/17(2) 7/6 5/4 Eq.: 0.445 (-0.000)
0.1% 0.9% 47.4% 52.6% 0.9% 0.2%
3. 2 18/17(2) 7/6 3/2 Eq.: 0.430 (-0.016)
0.1% 0.8% 46.9% 53.1% 0.7% 0.1%
4. 3 18/16(2) Eq.: 0.409 (-0.037)
0.1% 0.5% 46.4% 53.6% 0.6% 0.0%
Simplified 3-ply, 20%.
5. 2 18/17(2) 5/4 3/2 Eq.: 0.402 (-0.044)
0.0% 0.7% 46.3% 53.7% 0.6% 0.1%
6. 3 18/17(2) 7/6 2/1 Eq.: 0.355 (-0.090)
0.1% 0.7% 45.2% 54.8% 0.7% 0.1%
Simplified 3-ply, 20%.
7. 3 18/17(2) 7/5 Eq.: 0.347 (-0.099)
0.1% 0.7% 45.2% 54.8% 0.9% 0.1%
Simplified 3-ply, 20%.
8. 3 18/17(2) 3/1 Eq.: 0.338 (-0.107)
0.1% 0.6% 45.0% 55.0% 0.8% 0.1%
Simplified 3-ply, 20%.
9. 3 18/17(2) 5/4(2) Eq.: 0.336 (-0.110)
0.1% 0.6% 44.9% 55.1% 0.7% 0.1%
Simplified 3-ply, 20%.
10. 1 18/17(2) 5/3 Eq.: 0.273 (-0.173)
0.1% 0.8% 43.7% 56.3% 0.9% 0.1%
Why move one of the 1's to 5/4 as the better move? Wouldn't
it make more sense going 3/1 (position 8) with the two remaining
1's? I
understand only going to the 17pt because of the pip count.
My move of 18/17(2) 3/1 was the 8th move at -0.107 and is
considered a
blunder?
omegaman
the match output.
Match to 7. Score X-O: 3-4
X to play (1 1)
+24-23-22-21-20-19-------18-17-16-15-14-13-+
| O O O O O O | | X |
| O O O O | | X |
| O | | | S
| O | | | n
| | | | o
| |BAR| | w
| | | | i
| | | | e
| X X | | O |
| X X X X X | | O |
| X X X X X | | X O |
+-1--2--3--4--5--6--------7--8--9-10-11-12-+
Pipcount X: 78 O: 77 X-O: 3-4/7 (1)
Stick
GNU has 18/16(2) correct by .026 over the second best play, 18/17(2)
7/5. Any play after that is at least .05 back. (your play remains
back by .082) The latest version of Snowie also agrees with the top
two plays though it has a difference between them of only .016.
Raccoon
> 1. 2 18/17(2) 5/4 2/1 Eq.: 0.446
> 2. 2 18/17(2) 7/6 5/4 Eq.: 0.445 (-0.000)
> 3. 2 18/17(2) 7/6 3/2 Eq.: 0.430 (-0.016)
> 4. 3 18/16(2) Eq.: 0.409 (-0.037)
> Simplified 3-ply, 20%.
> 5. 2 18/17(2) 5/4 3/2 Eq.: 0.402 (-0.044)
> 6. 3 18/17(2) 7/6 2/1 Eq.: 0.355 (-0.090)
> Simplified 3-ply, 20%.
> 7. 3 18/17(2) 7/5 Eq.: 0.347 (-0.099)
> Simplified 3-ply, 20%.
> 8. 3 18/17(2) 3/1 Eq.: 0.338 (-0.107)
> Simplified 3-ply, 20%.
> Why move one of the 1's to 5/4 as the better move? Wouldn't
> it make more sense going 3/1 (position 8) with the two remaining
> 1's?
It doesn't make sense to me. If you hit, or are hit and hit back, or
hit and are hit and hit back, you'll want a board, not blots.
But I notice that the evaluation compares apples to oranges -- some
moves are analyzed 2-ply and some moves are analyzed 3-ply (20%). How
does your bot rank the plays when they're ALL analyzed on 3-ply
(preferably, 100%)?
omegaman
through. They basically came out in the same sequence as before with
slight variance in equities. I'm using cubeful rollouts so taking the
score into account I'm not too sure if moving both back men over to
the 16pt is correct either. However for a holding game I can't see
bringing home the outside checker on the 7pt as that could leave a
shot next roll if a six is thrown. Since both Raccoon and Stick are
using more recent bots the 18/16 must be the better move in the long
run.
omegaman
pretty much came up with the same sequence of suggested play with
slight variance in equity. I was using cubeful rollouts so taking the
match score into account I'm not sure if moving the two back men to
the 16pt is all that correct. However since this is a holding game I
don't see moving the checker on the 7pt to the home board either since
that could leave a shot next turn if I roll a six. Since both Raccoon
and Stick are running current versions of bots the 18/16 move must be
correct.