Gammonfever BG Quiz - Position #16
Micke Nilsson
Money session. Score X-O: 0-0
O on roll, cube action
+-1--2--3--4--5--6--------7--8--9-10-11-12-+
| X O O | | O O O X |
| O O | | O O O X |
| O | | O | S
| O | | | n
| | | | o
| |BAR| | w
| X | | | i
| X | | | e
| X | | |
| X X X | | O |
| X X X X | | X X O |
+24-23-22-21-20-19-------18-17-16-15-14-13-+
Pipcount X: 105 O: 124 X-O: 0-0/Money (1)
CubeValue: 1
paulde...@att.net
needs to be hit several times for 0 to catch up. So 0's main winning
strategy is to build a strong prime against X's 3 man.
However X can escape the back man when 0 moves forward.
Only strong market losers involve 0 making the 7 point (without losing
the 9 point ) [no, also 11 is strong].
When I began the analysis, I assumed that there were few such market
losers. However, I am now revising my opinion in mid-post because
there are more than I initially guessed. 0's probability of making the
7 point (without breaking another point of the prime to make it) is
22.2 %. Although 11 doesn't do this, it is just as strong -- so make
that 25%
So 25% significant market losers and it's quite difficult for X to
extract the back man.
Easy take of course. Double/take.
Paul Epstein
Philippe Michel
Double and Take.
Easy take. X has a good lead in the race and will have some oppurtunities
to escape. He will often build a 4-points board (or better) and have a few
indirect shots in O's outfield.
O has a few strong rolls (22, 33, 44, 55, 66, 64) and some other that
could be market losers if X doesn't roll well next (61, 41, 53, 63, 65). I
don't think 11 is that strong.
It is far from obvious to me that making a prime is the main game plan for
O. Given the choice, I prefer attacking (with 33 or 44 I would make the
3-point for instance, but probably not with 22). I think it is better for
the race and for impairing X's board building, and it wins some gammons
when it works.
Grunty
O rolls 65: 9/3* 8/3, X rolls 54: b/21/16*.
Almost *every* roll improves O's position, either making the bar or
4-point, or pointing on the blot. Most of them might push X into a
pass. So, double.
X should take. He's ahead in the race, and could still manage to jump
outside and avoid a shot.
paulde...@att.net
After Snowie has given its opinion on the cube action, I'd like to see
someone check how 33 or 44 should play. (Assume X possesses the cube
since all those in thread so far say double/take.)
Paul Epstein
Micke Nilsson
Money session. Score X-O: 0-0
O on roll, cube action
+-1--2--3--4--5--6--------7--8--9-10-11-12-+
| X O O | | O O O X |
| O O | | O O O X |
| O | | O | S
| O | | | n
| | | | o
| |BAR| | w
| X | | | i
| X | | | e
| X | | |
| X X X | | O |
| X X X X | | X X O |
+24-23-22-21-20-19-------18-17-16-15-14-13-+
Pipcount X: 105 O: 124 X-O: 0-0/Money (1)
CubeValue: 1
3-Ply Money equity: 0,385
0,4% 11,2% 66,8% 33,2% 6,4% 0,2%
1. No double 0,578
2. Double, take 0,528 (-0,050)
3. Double, pass 1,000 (+0,422)
Proper cube action: No double, take 11%
------------------------------ End ----------------------------------
3-ply full rollout:
Money session. Score X-O: 0-0
O on roll, cube action
+-1--2--3--4--5--6--------7--8--9-10-11-12-+
| X O O | | O O O X |
| O O | | O O O X |
| O | | O | S
| O | | | n
| | | | o
| |BAR| | w
| X | | | i
| X | | | e
| X | | |
| X X X | | O |
| X X X X | | X X O |
+24-23-22-21-20-19-------18-17-16-15-14-13-+
Pipcount X: 105 O: 124 X-O: 0-0/Money (1)
CubeValue: 1
Rollout Money equity: 0,442
0,4% 8,9% 69,8% 30,2% 4,7% 0,0%
95% confidence interval:
- money cubeless eq.: 0,442 ą0,022,
- live cube no double: 0,713 ą0,040,
- live cube double take: 0,701 ą0,056.
Rollout settings:
Full rollout,
324 games (equiv. 10451 games),
played 3-ply (standard), cube 3-ply,
settlement 0,550 at 16 pts,
seed 1, without race database.
Evaluations
1. No double 0,667
2. Double, take 0,663 (-0,004)
3. Double, pass 1,000 (+0,333)
Proper cube action: No double, take 1%
Live cube
1. No double 0,713
2. Double, take 0,701 (-0,012)
3. Double, pass 1,000 (+0,287)
Proper cube action: No double, take 4%
------------------------------ End ----------------------------------
Micke Nilsson
33 doubled
Money session. Score X-O: 0-0
O to play (3 3)
+-1--2--3--4--5--6--------7--8--9-10-11-12-+
| X O O | | O O O X |
| O O | | O O O X |
| O | | O | S
| O | | | n
| | | | o
| |BAR| | w
| X | | | i
| X | | | e
| X | | |
| X X X | | O |
| X X X X | | X X O |
+24-23-22-21-20-19-------18-17-16-15-14-13-+
Pipcount X: 105 O: 124 X-O: 0-0/Money (1)
CubeValue: 2, X owns Cube
1. 3 13/10(2) 6/3*(2) 0,609
2. 3 13/7(2) 0,525 (-0,084)
3. 3 13/10 8/5 6/3*(2) 0,519 (-0,090)
4. 3 13/7 6/3*(2) 0,518 (-0,091)
5. 3 11/5 6/3*(2) 0,510 (-0,099)
------------------------------ End ----------------------------------
33 not doubled
Money session. Score X-O: 0-0
O to play (3 3)
+-1--2--3--4--5--6--------7--8--9-10-11-12-+
| X O O | | O O O X |
| O O | | O O O X |
| O | | O | S
| O | | | n
| | | | o
| |BAR| | w
| X | | | i
| X | | | e
| X | | |
| X X X | | O |
| X X X X | | X X O |
+24-23-22-21-20-19-------18-17-16-15-14-13-+
Pipcount X: 105 O: 124 X-O: 0-0/Money (1)
CubeValue: 1
1. 3 13/10(2) 6/3*(2) 0,993
2. 3 13/7(2) 0,953 (-0,040)
3. 3 11/5 6/3*(2) 0,889 (-0,104)
4. 3 13/10 8/5 6/3*(2) 0,885 (-0,108)
5. 3 9/3* 8/5 6/3 0,862 (-0,131)
------------------------------ End ----------------------------------
44 doubled
Money session. Score X-O: 0-0
O to play (4 4)
+-1--2--3--4--5--6--------7--8--9-10-11-12-+
| X O O | | O O O X |
| O O | | O O O X |
| O | | O | S
| O | | | n
| | | | o
| |BAR| | w
| X | | | i
| X | | | e
| X | | |
| X X X | | O |
| X X X X | | X X O |
+24-23-22-21-20-19-------18-17-16-15-14-13-+
Pipcount X: 105 O: 124 X-O: 0-0/Money (1)
CubeValue: 2, X owns Cube
1. 3 11/3*(2) 0,535
2. 3 13/5 11/7(2) 0,497 (-0,038)
3. 3 11/7(2) 6/2(2) 0,459 (-0,076)
4. 3 13/9 11/7(2) 8/4 0,458 (-0,076)
5. 3 13/9 11/7(2) 6/2 0,421 (-0,114)
------------------------------ End ----------------------------------
44 not doubled
Money session. Score X-O: 0-0
O to play (4 4)
+-1--2--3--4--5--6--------7--8--9-10-11-12-+
| X O O | | O O O X |
| O O | | O O O X |
| O | | O | S
| O | | | n
| | | | o
| |BAR| | w
| X | | | i
| X | | | e
| X | | |
| X X X | | O |
| X X X X | | X X O |
+24-23-22-21-20-19-------18-17-16-15-14-13-+
Pipcount X: 105 O: 124 X-O: 0-0/Money (1)
CubeValue: 1
1. 3 11/3*(2) 0,920
2. 3 11/7(2) 6/2(2) 0,902 (-0,019)
3. 3 13/5 11/7(2) 0,871 (-0,050)
4. 3 13/9 11/7(2) 6/2 0,844 (-0,076)
5. 3 13/9 11/7(2) 8/4 0,736 (-0,184)
------------------------------ End ----------------------------------
<paulde...@att.net> skrev i meddelandet
news:1129024854.3...@g47g2000cwa.googlegroups.com...
Grunty
> 3-ply instant evaluation:
>
> Proper cube action: No double, take 11%
>
> 3-ply full rollout:
>
> Live cube
> Proper cube action: No double, take 4%
Snowie got crazy?!
Can anybody roll this one out on a truly good program, like GNUBG, at
maximum level?
Cat_in_awe
You don't know much about bots if you think it's unusual or the sign of a
bad program that the results of a full rollout and 3-ply evaluation can be
quite different.
--
GPC
Grunty
> Grunty wrote:
> >
> > Snowie got crazy?!
> > Can anybody roll this one out on a truly good program, like GNUBG, at
> > maximum level?
>
> You don't know much about bots if you think it's unusual or the sign of a
> bad program that the results of a full rollout and 3-ply evaluation can be
> quite different.
Hi Gregg,
I know you hold some friendly tie with Snowiegroup, don't be so
jealous. For a moment, I thought you were going to give your opinion on
the position...oh well.
I was just joking about the fact that *all* of the contestants assessed
the position as a double. Next time I'll add a smiley to my post, so
you can find it funny.
Raccoon
Micke Nilsson wrote:
> - live cube no double: 0,713 ±0,040,
> - live cube double take: 0,701 ±0,056.
With 95% confidence intervals so large (±0,040, ±0,056) 324 trials
doesn't really tell us much about decisions differing by only 0.012. If
you have time, perhaps you could extend this rollout to at least 1296
trials.
Shawn K. Quinn
from Grunty <grunti...@yahoo.com>
in message <1129056266.2...@o13g2000cwo.googlegroups.com>
posted at 2005-10-11T18:44
> Micke Nilsson wrote:
>> 3-ply instant evaluation:
>> 3-ply full rollout:
GNU Backgammon agrees, but I didn't crank it up to maximum level:
Cube analysis
Rollout cubeless equity +0.4054
Cubeful equities:
1. No double +0.6198
2. Double, pass +1.0000 ( +0.3802)
3. Double, take +0.5784 ( -0.0414)
Proper cube action: No double, take (9.8%)
Rollout details:
Centered 1-cube:
0.6800 0.0995 0.0037 - 0.3200 0.0570 0.0008 CL +0.4054 CF +0.6198
[0.0021 0.0015 0.0002 - 0.0021 0.0017 0.0001 CL 0.0054 CF 0.0110]
Player red X owns 2-cube:
0.6837 0.0932 0.0037 - 0.3163 0.0590 0.0010 CL +0.8084 CF +0.5784
[0.0023 0.0018 0.0002 - 0.0023 0.0018 0.0001 CL 0.0119 CF 0.0151]
Truncated cubeful rollout (depth 10) with var.redn.
360 games, Mersenne Twister dice gen. with seed -1916338358 and quasi-random dice
Stop when std.errs. are small enough: ratio 0.025 (min. 108 games)
Play: 0-ply cubeful prune [expert]
Cube: 0-ply cubeful prune [expert]
--
___ _ _____ |*|
/ __| |/ / _ \ |*| Shawn K. Quinn
\__ \ ' < (_) | |*| skq...@speakeasy.net
|___/_|\_\__\_\ |*| Houston, TX, USA
Jim Segrave
Currently I'm at 1018 games and the jsd's are about 0.26 after 8
hours. I'll leave it rolling out overnight, 1296 games is not going to
resolve this, I'll be surprised if there's an answer before Friday
Cube analysis
Rollout cubeless equity +0.430
Cubeful equities:
1. No double +0.725
2. Double, pass +1.000 ( +0.275)
3. Double, take +0.721 ( -0.004)
Proper cube action: No double, take (1.5%)
Rollout details:
Centered 1-cube:
0.701 0.100 0.004 - 0.299 0.075 0.001 CL +0.430 CF +0.725
[0.001 0.002 0.000 - 0.001 0.002 0.000 CL 0.003 CF 0.008]
Player Allie owns 2-cube:
0.710 0.104 0.004 - 0.290 0.077 0.002 CL +0.899 CF +0.721
[0.002 0.003 0.001 - 0.002 0.002 0.000 CL 0.011 CF 0.013]
Full cubeful rollout with var.redn.
1018 games, Mersenne Twister dice gen. quasi-random dice
Stop when best play is enough JSDs ahead: limit 2.006 (min. 648 games)
Play: supremo 2-ply cubeful prune [world class]
keep the first 0 0-ply moves and up to 16 more moves within equity 0.32
Skip pruning for 1-ply moves.
Cube: 2-ply cubeful prune [world class]
--
Jim Segrave (j...@jes-2.demon.nl)
Philippe Michel
>
> Micke Nilsson wrote:
>> - live cube no double: 0,713 ±0,040,
>> - live cube double take: 0,701 ±0,056.
>
> With 95% confidence intervals so large (±0,040, ±0,056) 324 trials
> doesn't really tell us much about decisions differing by only 0.012. If
> you have time, perhaps you could extend this rollout to at least 1296
> trials.
There is a real risk that it would be mostly a waste of time (going from
"it's twice more likely that it's not a double than a double" to "it's 3
times more likely..."). Big deal!
What seems more useful to me would be to test similar problems with
various positions for X (or possibly O). When is it a clear no-double ? A
clear double ? A pass ?
For what it's worth, there was a somewhat similar problem (with a more
advanced, post-blitz-like, forward position for the leader) in one of the
last _Inside Backgammon_ (quiz 149, Fall 1997 issue).
Original problem was :
GNU Backgammon Position ID: 020GAwQzOwcHAA
Match ID : UQkAAAAAAAAA
+13-14-15-16-17-18------19-20-21-22-23-24-+ O: White
| X O | | O O O O O | 0 points
| X O | | O O O O |
| X | | O |
| | | |
| | | |
v| |BAR| |
| | | |
| | | |
| X | | X |
| O X | | X X X X | On roll
| O X | | X X O X X | 0 points
+12-11-10--9--8--7-------6--5--4--3--2--1-+ X: Black (Cube: 2)
Cube action, then 5-4 to play by X.
Then, there were 16 derived positions with the same X position and
various board strengths and pipcounts for O's.
This is the last part that is really helpful, much more than the
discussion or rollout of one single position.
Raccoon
Philippe Michel wrote:
> There is a real risk that it would be mostly a waste of time (going from
> "it's twice more likely that it's not a double than a double" to "it's 3
> times more likely..."). Big deal!
Hmmm. Well, after 324 trials the results overlap so much that the
rollout has not revealed with much certainty at all what Snowie
'thinks' is the best decision. 1296 trials might be enough. If so, it's
not a waste of time. If they're not enough, at least we'd know that the
decision is (according to Snowie) so close that 1296 trials weren't
enough. And then you could decide whether to do more trials. It's
looking like about 5000 trials at 2-ply might be enough with GNUBG, but
I'm going to stop my own 'world class' rollout after 3888 trials (jsd
is 1.6 after 3300 trials) and not waste any more time :) It would be
nice to have a long Snowie rollout to compare the gnubg results to.
> Then, there were 16 derived positions with the same X position and
> various board strengths and pipcounts for O's.
> This is the last part that is really helpful, much more than the
> discussion or rollout of one single position.
That position looks interesting and the discussion sounds useful. I
wonder how much time it took to roll out all 17 positions?
paulde...@att.net
is close so X has to hit. There is only one hit that makes any sense
at all -- 13-4* This is a completely forced play for anyone who's not
a beginner. My guess is that this is a typo, and that you intend 5-4
to mean something else.
If there is an alternative play worth considering, it might be 8-4* 8-3
because it gives an extra builder for the 2 point but the disadvantage
of creating a blot on the 8 point hugely outweighs the extra builder.
Conclusion: probable typo. If not a typo, play 13-4* no matter where
the cube is.
X has got a powerhouse double. How can 0 win? The most likely winning
variations for 0 are where X fails to hit and 0 wins the race after X
misses yet again; or where X has to hit loose and suffers a return hit.
However, nothing is completely certain and even without these major
winning opportunities, 0 is bound to have other winning chances.
Probably 0 has more than 25% equity so 0 has a take. I don't feel like
justifying the double because it's so very clear -- X has a superb
attack on the blot and is 3 pips ahead before the roll [on second
thoughts, I have justified it.]
Another point (justifying the take) is that X does not have huge gammon
equity. Even in the dream/nightmare scenario where X closes 0 out
before 0 has a chance to move, X is still probably less than 40% likely
to get a gammon. (If 0's men on 17 were home, the probability is 27%
if I remember rightly.)
double, very marginal take.
Paul Epstein
Raccoon
paulde...@att.net wrote:
> I don't understand why playing 54 is a genuine problem. The pip count
> is close so X has to hit. There is only one hit that makes any sense
> at all -- 13-4* This is a completely forced play for anyone who's not
> a beginner. My guess is that this is a typo, and that you intend 5-4
> to mean something else.
I don't think so he did. The point is, I think, that proper cube and
play might not be so obvious in some of the 16 altered positions. If
13/4* is clear now, is is clear if O has a closed board? Is the double
clear if O is 10 pips ahead in the race? And so on -- analyzing the 17
similar positions and figuring out why proper play might differ in them
can be, as PM suggested, much more useful than just seeing a single
position and its rollout without noticing what the important features
of the position are.
paulde...@att.net
closed.
Same checker play problem but move 0's men on the 21 and 17 points to
close the board.
My play changes to 13/8, 6/2 but I don't know. Hitting loose still
gives the most wins but the return hits lose too many gammons for me to
stick with 13/4*
The only safe 4 is 6/2. I play 13/8 with the 5, grabbing an
opportunity to remove a checker from the midpoint which will be
difficult to clear. 0 will break his/her own midpoint so X will be
able to clear the midpoint checkers one at a time. 13/8 clears one.
Paul Epstein
Raccoon
> 3-ply full rollout:
> Money session. Score X-O: 0-0
> O on roll, cube action
> +-1--2--3--4--5--6--------7--8--9-10-11-12-+
> | X O O | | O O O X |
> | O O | | O O O X |
> | O | | O | S
> | O | | | n
> | | | | o
> | |BAR| | w
> | X | | | i
> | X | | | e
> | X | | |
> | X X X | | O |
> | X X X X | | X X O |
> +24-23-22-21-20-19-------18-17-16-15-14-13-+
> Pipcount X: 105 O: 124 X-O: 0-0/Money (1)
> CubeValue: 1
Cube analysis
Rollout cubeless equity +0.4327
Cubeful equities:
1. Double, take +0.7343
2. Double, pass +1.0000 ( +0.2657)
3. No double +0.7179 ( -0.0164)
Proper cube action: Double, take
Rollout details:
Centered 1-cube:
0.6989 0.1047 0.0044 - 0.3011 0.0731 0.0011 CL +0.4327 CF +0.7179
[0.0008 0.0010 0.0002 - 0.0008 0.0009 0.0001 CL 0.0019 CF 0.0044]
Player Blue owns 2-cube:
0.7098 0.1085 0.0056 - 0.2902 0.0750 0.0014 CL +0.9144 CF +0.7343
[0.0009 0.0013 0.0003 - 0.0009 0.0010 0.0001 CL 0.0044 CF 0.0062]
Full cubeful rollout with var.redn.
3888 games, Mersenne Twister dice gen. with seed 979790848 and
quasi-random dice
Play: world class 2-ply cubeful prune [world class]
keep the first 0 0-ply moves and up to 8 more moves within equity 0.16
Philippe Michel
That's the idea, although safety is more important than you seem to think.
According to gnubg rollouts, 13/4* is better by 0.08 in the original
position (8/4* 8/3 is a lot weaker), but just giving O a 5-points board by
advancing O's 3rd man from the 4pt to the 3pt make 13/8 6/2 better by
0.06.
In general, X should usually hit except if O has a good position (good
enough for a redouble/pass in case of a return hit maybe ?) and X is ahead
or equal in the race.
For the cube problem, it is more complicated since with varying O's front
positions, the situation can go from a race with a little contact (the
latter helping X more than O) when X has to play safely, to some kind of
blitz when O's board is weaker but improving, to playing against a busted,
unsalvageable inner board. The original position is a double/clear take
(equity goes from 2*0.74 to 2*0.82).
paulde...@att.net
original position or the closed-board modification?
Did I get both 54s right? I gather I was wrong to think there were no
alternatives in the original position.
Paul Epstein
paulde...@att.net
Grunty
>> Snowie got crazy?!
>> Can anybody roll this one out on a truly good program,
>> like GNUBG, at maximum level?
Raccoon wrote:
> 3888 games 2-ply cubeful prune [world class]
> Proper cube action: Double, take
Aha, finally, we got there!
I like this rollout much more than Micke's (Snowie 324 games at 3-ply)
and Shawn's (GNUBG 360 games at 0-ply cubeful prune [expert])
:-P <---- smiley for Gregg
Let's rest our case, folks.
Philippe Michel
> original position or the closed-board modification?
GNU Backgammon Position ID: 020GAwQzOwcHAA
Match ID : QgkWAAAAAAAA
+13-14-15-16-17-18------19-20-21-22-23-24-+ O: White (Cube: 4)
| X O | | O O O O O | 0 points
| X O | | O O O O |
| X | | O |
| | | |
| | | |
v| |BAR| |
| | | |
| | | |
| X | | X |
| O X | | X X X X | Rolled 45
| O X | | X X O X X | 0 points
+12-11-10--9--8--7-------6--5--4--3--2--1-+ X: Black
Pip counts: O 102, X 99
Rolled 45: 1. Rollout 13/4* Eq.: +0.292
0.646 0.187 0.003 - 0.354 0.084 0.001 CL +0.397 CF +0.292
[0.005 0.008 0.002 - 0.005 0.004 0.000 CL 0.011 CF 0.014]
2. Rollout 13/8 6/2 Eq.: +0.214 ( -0.078)
0.649 0.047 0.001 - 0.351 0.023 0.000 CL +0.322 CF +0.214
[0.002 0.005 0.000 - 0.002 0.003 0.000 CL 0.007 CF 0.008]
3. Rollout 8/3 6/2 Eq.: +0.185 ( -0.107)
0.640 0.045 0.001 - 0.360 0.029 0.001 CL +0.295 CF +0.185
[0.003 0.005 0.000 - 0.003 0.003 0.000 CL 0.007 CF 0.008]
4. Rollout 8/4* 8/3 Eq.: +0.121 ( -0.170)
0.598 0.173 0.003 - 0.402 0.147 0.003 CL +0.223 CF +0.121
[0.003 0.006 0.000 - 0.003 0.005 0.000 CL 0.009 CF 0.010]
Truncated cubeful rollout (depth 11) with var.redn.
72 games, Mersenne Twister dice gen. with seed 97798583 and quasi-random dice
Stop when best play is enough JSDs ahead: limit 1.96 (min. 72 games)
Play: world class 2-ply cubeful prune [world class]
keep the first 0 0-ply moves and up to 8 more moves within equity 0.16
Skip pruning for 1-ply moves.
Cube: 2-ply cubeful prune [world class]
Different evaluations after 7 plies:
Play: 0-ply cubeful prune [expert]
Cube: 0-ply cubeful prune [expert]
GNU Backgammon Position ID: s20GAwQzOwcHAA
Match ID : QgkWAAAAAAAA
+13-14-15-16-17-18------19-20-21-22-23-24-+ O: White (Cube: 4)
| X O | | O O O O O | 0 points
| X O | | O O O O O |
| X | | |
| | | |
| | | |
v| |BAR| |
| | | |
| | | |
| X | | X |
| O X | | X X X X | Rolled 45
| O X | | X X O X X | 0 points
+12-11-10--9--8--7-------6--5--4--3--2--1-+ X: Black
Pip counts: O 101, X 99
Rolled 45: 1. Rollout 13/8 6/2 Eq.: +0.158
0.633 0.028 0.000 - 0.367 0.018 0.000 CL +0.276 CF +0.158
[0.002 0.003 0.000 - 0.002 0.003 0.000 CL 0.007 CF 0.007]
2. Rollout 13/4* Eq.: +0.100 (-0.058)
0.585 0.132 0.002 - 0.415 0.107 0.001 CL +0.196 CF +0.100
[0.003 0.006 0.000 - 0.003 0.003 0.000 CL 0.008 CF 0.009]
GNU Backgammon Position ID: 020GAwQzOwcHAA
Match ID : UQkAAAAAAAAA
+13-14-15-16-17-18------19-20-21-22-23-24-+ O: White
| X O | | O O O O O | 0 points
| X O | | O O O O |
| X | | O |
| | | |
| | | |
v| |BAR| |
| | | |
| | | |
| X | | X |
| O X | | X X X X | On roll
| O X | | X X O X X | 0 points
+12-11-10--9--8--7-------6--5--4--3--2--1-+ X: Black (Cube: 2)
Pip counts: O 102, X 99
Cube analysis
Rollout cubeless equity +0.471
Cubeful equities:
1. Double, take +0.817
2. Double, pass +1.000 ( +0.183)
3. No double +0.737 ( -0.080)
Proper cube action: Redouble, take
Rollout details:
Player Black owns 2-cube:
0.704 0.094 0.001 - 0.296 0.031 0.000 CL +0.471 CF +0.737
[0.003 0.006 0.000 - 0.003 0.004 0.000 CL 0.006 CF 0.012]
Player White owns 4-cube:
0.710 0.100 0.001 - 0.290 0.027 0.000 CL +0.984 CF +0.817
[0.003 0.007 0.000 - 0.003 0.004 0.000 CL 0.016 CF 0.021]
Full cubeful rollout with var.redn.
72 games, Mersenne Twister dice gen. with seed 97798583 and quasi-random dice
Stop when best play is enough JSDs ahead: limit 1.96 (min. 72 games)
Play: world class 2-ply cubeful prune [world class]
keep the first 0 0-ply moves and up to 8 more moves within equity 0.16
Skip pruning for 1-ply moves.
Cube: 2-ply cubeful prune [world class]
Different evaluations after 7 plies:
Jim Segrave
I was off from work after my last posting, while gnubg ground on at
this in the office. Supremo settings, Snowie MET:
Cube analysis
Rollout cubeless equity +0.433
Cubeful equities:
1. Double, take +0.735
2. Double, pass +1.000 ( +0.265)
3. No double +0.723 ( -0.012)
Proper cube action: Double, take
Rollout details:
Centered 1-cube:
0.700 0.104 0.004 - 0.300 0.075 0.001 CL +0.433 CF +0.723
[0.001 0.001 0.000 - 0.001 0.001 0.000 CL 0.001 CF 0.003]
Player Allie owns 2-cube:
0.711 0.109 0.006 - 0.289 0.077 0.001 CL +0.918 CF +0.735
[0.001 0.001 0.000 - 0.001 0.001 0.000 CL 0.003 CF 0.005]
Full cubeful rollout with var.redn.
7067 games, Mersenne Twister dice gen. with seed 1065139071 and
quasi-random dice
Stop when best play is enough JSDs ahead: limit 2.006 (min. 648 games)
Play: supremo 2-ply cubeful prune [world class]
keep the first 0 0-ply moves and up to 16 more moves within equity 0.32
Skip pruning for 1-ply moves.
Cube: 2-ply cubeful prune [world class]
--
Jim Segrave (j...@jes-2.demon.nl)
Micke Nilsson
Money session. Score X-O: 0-0
O on roll, cube action
+-1--2--3--4--5--6--------7--8--9-10-11-12-+
| X O O | | O O O X |
| O O | | O O O X |
| O | | O | S
| O | | | n
| | | | o
| |BAR| | w
| X | | | i
| X | | | e
| X | | |
| X X X | | O |
| X X X X | | X X O |
+24-23-22-21-20-19-------18-17-16-15-14-13-+
Pipcount X: 105 O: 124 X-O: 0-0/Money (1)
CubeValue: 1
Rollout Money equity: 0,432
0,4% 9,1% 69,4% 30,6% 5,0% 0,0%
95% confidence interval:
- money cubeless eq.: 0,432 ±0,011,
- live cube no double: 0,682 ±0,021,
- live cube double take: 0,674 ±0,028.
Rollout settings:
Full rollout,
1296 games (equiv. 38750 games),
played 3-ply (standard), cube 3-ply,
settlement 0,550 at 16 pts,
random seed, without race database.
Evaluations
1. No double 0,653
2. Double, take 0,641 (-0,012)
3. Double, pass 1,000 (+0,347)
Proper cube action: No double, take 3%
Live cube
1. No double 0,682
2. Double, take 0,674 (-0,008)
3. Double, pass 1,000 (+0,318)
Proper cube action: No double, take 2%
------------------------------ End ----------------------------------
"Raccoon" <racg...@yahoo.com> skrev i meddelandet
news:1129086196.0...@g44g2000cwa.googlegroups.com...
Raccoon
7067-trial gnubg rollout, even 1296 trials isn't enough with Snowie for
this position. It's interesting that gnubg seems to play the Doubling
side a little better than Snowie, or the Taking side a little worse.