[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Bug-gnubg] Use of Two-Sided DB in matches
|
From: |
Neil Robins |
|
Subject: |
Re: [Bug-gnubg] Use of Two-Sided DB in matches |
|
Date: |
Mon, 7 Sep 2009 17:07:51 +0100 |
OK I've looked at quite a few positions now where there would be an
automatic recube for the match (and take) after a double by the leader. I
hope this final one that I post demonstrates, what I am already quite
certain of:, that there is a fundamental bug resulting in inconsistent use
of the Two-Sided DB when exact matchplay figures are obtainable from it.
GNU Backgammon Position ID: qQIAgKoAAAAAAA
Match ID : cAngAEAAKAAA
+13-14-15-16-17-18------19-20-21-22-23-24-+ O: gnubg
| | | O O O O O | OO 4 points
| | | | OO
| | | | OO
| | | | OO
| | | | OO
v| |BAR| | 7 point match (Cube: 1)
| | | | XX
| | | | XX
| | | | XX
| | | | XX On roll
| | | X X X X X | XX 5 points
+12-11-10--9--8--7-------6--5--4--3--2--1-+ X: Neil
Two sided DB gives:
Player Opponent
Position 385 315
Cubeless equity : +0.3588
Owned cube : +0.5654
Centered cube : +0.5229
Opponent owns cube : +0.3030
From which it is directly calculable that if the leader doubles and it is
taken, then opponent recubes for the match.
Leader's MWC = (1 + 0.3588)/2 x 100 = 67.94%
GNU gets this one right for the D/T figure on 1-ply and then goes bonkers on
higher plys. I also see it as pretty serious that again a GNU rollout can't
be trusted, just as in the first position I posted. The correct XG one (well
correct for the D/T figure with absolute certainty) is under the GNU rollout
figures.
Cube analysis
0-ply cubeless MWC 67.04% (Money: +0.359)
0.679 0.000 0.000 - 0.321 0.000 0.000
Cubeful equities:
1. Double, take 74.60%
2. Double, pass 75.08% ( 0.48%)
3. No double 70.02% ( -4.58%)
Proper cube action: Double, take
Cube analysis
1-ply cubeless MWC 67.04% (Money: +0.359)
0.679 0.000 0.000 - 0.321 0.000 0.000
Cubeful equities:
1. No double 69.68%
2. Double, pass 75.08% ( 5.40%)
3. Double, take 67.94% ( -1.73%)
Proper cube action: No double, take (24.3%)
Cube analysis
2-ply cubeless MWC 67.04% (Money: +0.359)
0.680 0.000 0.000 - 0.320 0.000 0.000
Cubeful equities:
1. No double 69.61%
2. Double, pass 75.08% ( 5.46%)
3. Double, take 67.97% ( -1.64%)
Proper cube action: No double, take (23.1%)
Cube analysis
3-ply cubeless MWC 66.72% (Money: +0.334)
0.667 0.000 0.000 - 0.333 0.000 0.000
Cubeful equities:
1. No double 68.72%
2. Double, pass 75.08% ( 6.36%)
3. Double, take 66.69% ( -2.04%)
Proper cube action: No double, take (24.3%)
Cube analysis
4-ply cubeless MWC 66.82% (Money: +0.341)
0.671 0.000 0.000 - 0.329 0.000 0.000
Cubeful equities:
1. No double 68.87%
2. Double, pass 75.08% ( 6.21%)
3. Double, take 67.06% ( -1.81%)
Proper cube action: No double, take (22.6%)
Cube analysis
7-ply cubeless MWC 66.79% (Money: +0.339)
0.669 0.000 0.000 - 0.331 0.000 0.000
Cubeful equities:
1. No double 68.82%
2. Double, pass 75.08% ( 6.25%)
3. Double, take 66.95% ( -1.88%)
Proper cube action: No double, take (23.1%)
Rollout details:
Centered 1-cube:
0.673 0.000 0.000 - 0.327 0.000 0.000 CL 66.87% CF 69.00%
[0.000 0.000 0.000 - 0.000 0.000 0.000 CL 0.00% CF 0.00%]
Player gnubg owns 2-cube:
0.678 0.000 0.000 - 0.322 0.000 0.000 CL 78.04% CF 67.78%
[0.000 0.000 0.000 - 0.000 0.000 0.000 CL 0.00% CF 0.00%]
Full cubeful rollout with var.redn.
129600 games, Mersenne Twister dice gen. with seed 864259327 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
Skip pruning for 1-ply moves.
Cube: 2-ply cubeful prune [world class]
or in EMG
Rollout details:
Centered 1-cube:
0.673 0.000 0.000 - 0.327 0.000 0.000 CL +0.346 CF +0.516
[0.000 0.000 0.000 - 0.000 0.000 0.000 CL 0.000 CF 0.000]
Player gnubg owns 2-cube:
0.678 0.000 0.000 - 0.322 0.000 0.000 CL +1.237 CF +0.418
[0.000 0.000 0.000 - 0.000 0.000 0.000 CL 0.000 CF 0.000]
Full cubeful rollout with var.redn.
129600 games, Mersenne Twister dice gen. with seed 864259327 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
Skip pruning for 1-ply moves.
Cube: 2-ply cubeful prune [world class]
XG using the same MET gets:
Analyzed in Rollout
Player Winning Chances: 67.94% (G: 0.00% B: 0.00%)
Opponent Winning Chances: 32.06% (G: 0.00% B: 0.00%)
Cubeless Equities: No Double=+0.359, Double=+1.245
Cubeful Equities:
No Double: +0.460
Double/Take: +0.431 (-0.029)
Double/Drop: +1.000 (+0.540)
Best Cube action: No Double / Take
Rollout
129600 Games rolled with Variance Reduction.
Moves and cube decisions: 3 ply
Confidence No Double: ± 0.002 (+0.459<E<+0.462)
Confidence Double: ± 0.000 (+0.431<E<+0.431)
Double Decision confidence: 100.0%
Take Decision confidence: 100.0%
Duration: 5.5 seconds
----- Original Message -----
From: "Neil Robins" <address@hidden>
To: "Michael Petch" <address@hidden>; <address@hidden>
Sent: Monday, September 07, 2009 6:52 AM
Subject: Re: [Bug-gnubg] Use of Two-Sided DB in matches
I have another position here that GNU seems to fundamentally get almost
right and I don't really know why it calculates differently than from in my
initial position.
GNU Backgammon Position ID: kQcAgEoDAAAAAA
Match ID : UQngABAAIAAA
+13-14-15-16-17-18------19-20-21-22-23-24-+ O: gnubg
| | | O O O | OO 1 point
| | | O | OO
| | | O | OO
| | | O | OO
| | | | O
v| |BAR| | 7 point match
| | | | X
| | | | XX
| | | | XX
| | | X | XX On roll
| | | X X X X X | XX 4 points
+12-11-10--9--8--7-------6--5--4--3--2--1-+ X: Neil (Cube: 2)
Two sided DB gives:
Player Opponent
Position 641 707
Cubeless equity : +0.7554
Owned cube : +0.9339
Centered cube : +0.9303
Opponent owns cube : +0.7441
From which it is directly calculable that if the leader redoubles and it
is taken, then opponent recubes for the match ( the take of which is never
in the slightest doubt):
Leader's MWC = (1 + 0.7554)/2 x 100 = 87.77%
I think 3-ply and 4-ply really should give the same result for D/T, but
they are out by only 0.01%.
Cube analysis
0-ply cubeless MWC 85.15% (Money: +0.755)
0.878 0.000 0.000 - 0.122 0.000 0.000
Cubeful equities:
1. Double, pass 89.02%
2. Double, take 91.76% ( 2.74%)
3. No double 87.65% ( -1.37%)
Proper cube action: Redouble, pass
Cube analysis
1-ply cubeless MWC 85.15% (Money: +0.755)
0.878 0.000 0.000 - 0.122 0.000 0.000
Cubeful equities:
1. Double, take 87.77%
2. Double, pass 89.02% ( 1.26%)
3. No double 87.28% ( -0.48%)
Proper cube action: Redouble, take
Cube analysis
2-ply cubeless MWC 85.15% (Money: +0.755)
0.878 0.000 0.000 - 0.122 0.000 0.000
Cubeful equities:
1. Double, take 87.77%
2. Double, pass 89.02% ( 1.26%)
3. No double 87.61% ( -0.16%)
Proper cube action: Redouble, take
Cube analysis
3-ply cubeless MWC 85.15% (Money: +0.755)
0.878 0.000 0.000 - 0.122 0.000 0.000
Cubeful equities:
1. Double, take 87.76%
2. Double, pass 89.02% ( 1.26%)
3. No double 87.08% ( -0.68%)
Proper cube action: Redouble, take
Cube analysis
4-ply cubeless MWC 85.15% (Money: +0.755)
0.878 0.000 0.000 - 0.122 0.000 0.000
Cubeful equities:
1. Double, take 87.76%
2. Double, pass 89.02% ( 1.26%)
3. No double 87.17% ( -0.60%)
Proper cube action: Redouble, take
Rollout details:
Player Neil owns 2-cube:
0.878 0.000 0.000 - 0.122 0.000 0.000 CL 85.15% CF 86.94%
[0.000 0.000 0.000 - 0.000 0.000 0.000 CL 0.00% CF 0.01%]
Player gnubg owns 4-cube:
0.878 0.000 0.000 - 0.122 0.000 0.000 CL 92.67% CF 87.77%
[0.000 0.000 0.000 - 0.000 0.000 0.000 CL 0.00% CF 0.00%]
Full cubeful rollout with var.redn.
12960 games, Mersenne Twister dice gen. with seed 864628208 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
Skip pruning for 1-ply moves.
Cube: 2-ply cubeful prune [world class]
- Re: [Bug-gnubg] Use of Two-Sided DB in matches, (continued)
- Re: [Bug-gnubg] Use of Two-Sided DB in matches, Neil Robins, 2009/09/05
- Re: [Bug-gnubg] Use of Two-Sided DB in matches - I agree, this looks suspicious., Michael Petch, 2009/09/05
- Re: [Bug-gnubg] Use of Two-Sided DB in matches - I agree, this looks suspicious., Michael Petch, 2009/09/05
- Re: [Bug-gnubg] Use of Two-Sided DB in matches - I agree, this looks suspicious., Neil Robins, 2009/09/05
- Re: [Bug-gnubg] Use of Two-Sided DB in matches, Neil Robins, 2009/09/07
- Re: [Bug-gnubg] Use of Two-Sided DB in matches, Neil Robins, 2009/09/05
- Re: [Bug-gnubg] Use of Two-Sided DB in matches, Christian Anthon, 2009/09/05
- Re: [Bug-gnubg] Use of Two-Sided DB in matches, Neil Robins, 2009/09/05
Re: [Bug-gnubg] Use of Two-Sided DB in matches,
Neil Robins <=