BGonline.org Forums

xG evals + rollouts + quirk

Posted By: Robert-Jan Veldhuizen (Zorba)
Date: Sunday, 5 July 2009, at 10:23 a.m.

In Response To: GNU rollout and eval (neilkaz)

Unfortunately xG is quirky when setting up a cube problem (not unlike GnuBG). If you don't use "play from position", but just analyze right away, it works, but won't show up in any exports.

If you "play from position" in a cube problem, you can't actually cube and it will let opponent roll. Using "take decision" instead of "doubling decision" might be the solution, but it's far from obvious.

Here's text output from xG (a bit messy and anti-clockwise despite my settings), using 3-,4- and 5-ply, xG roller (a short, fast truncated rollout) and then rollouts with increasingly higher settings.

 

XGID=--bb-BCCBB--bA--bb-cBb----:0:0:1:00:0:0:0:0:8 

 

Score is X:0 O:0. 

 +13-14-15-16-17-18------19-20-21-22-23-24-+ 

 | X        O  O    |   | O  X  O          | 

 |          O  O    |   | O  X  O          | 

 |                  |   | O                | 

 |                  |   |                  | 

 |                  |   |                  | 

 |                  |BAR|                  | 

 |                  |   |                  | 

 |                  |   |                  | 

 |                X |   | X                | 

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

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

 +12-11-10--9--8--7-------6--5--4--3--2--1-+ 

Pipcount  X: 136  O: 176 X-O: 0-0 

Cube  : 1 

X roll or double ? 

 

Analyzed in 3 ply 

Player Winning Chances:    72,08% (G: 24,78% B:  1,51%) 

Opponent Winning Chances:  27,92% (G:  3,71% B:  0,09%) 

 

Cubeless Equities: No Double=+0,667, Double=+1,333 

 

Cubefull Equities: 

      No Double:   +0,951 (-0,049) 

      Double/Take: +1,143 (+0,143) 

      Double/Drop: +1,000 

 

Best Cube action: Double / Drop 

 

Analyzed in 4 ply 

Player Winning Chances:    72,50% (G: 24,97% B:  1,61%) 

Opponent Winning Chances:  27,50% (G:  4,13% B:  0,11%) 

 

Cubeless Equities: No Double=+0,674, Double=+1,347 

 

Cubefull Equities: 

      No Double:   +0,948 (-0,052) 

      Double/Take: +1,160 (+0,160) 

      Double/Drop: +1,000 

 

Best Cube action: Double / Drop 

 

Analyzed in 5 ply 

Player Winning Chances:    72,23% (G: 24,68% B:  1,46%) 

Opponent Winning Chances:  27,77% (G:  3,67% B:  0,10%) 

 

Cubeless Equities: No Double=+0,668, Double=+1,337 

 

Cubefull Equities: 

      No Double:   +0,950 (-0,050) 

      Double/Take: +1,148 (+0,148) 

      Double/Drop: +1,000 

 

Best Cube action: Double / Drop 

 

Analyzed in XG roller 

Player Winning Chances:    72,13% (G: 24,64% B:  1,50%) 

Opponent Winning Chances:  27,87% (G:  3,73% B:  0,10%) 

 

Cubeless Equities: No Double=+0,666, Double=+1,331 

 

Cubefull Equities: 

      No Double:   +0,952 (-0,048) 

      Double/Take: +1,150 (+0,150) 

      Double/Drop: +1,000 

 

Best Cube action: Double / Drop 

 

ROLLOUT RESULTS

 

Analyzed in Rollout 

Player Winning Chances:    72,36% (G: 25,10% B:  1,54%) 

Opponent Winning Chances:  27,64% (G:  3,82% B:  0,13%) 

 

Cubeless Equities: No Double=+0,674, Double=+1,348 

 

Cubefull Equities: 

      No Double:   +0,978 (-0,022) 

      Double/Take: +1,195 (+0,195) 

      Double/Drop: +1,000 

 

Best Cube action: Double / Drop 

 

Rollout 

  3888 Games rolled with Variance Reduction. 

  Moves and cube decisions: 1 ply 

  Confidence No Double: ± 0,013 (+0,964  Confidence Double:    ± 0,024 (+1,171   

  Double Decision confidence: 99,9% 

  Take Decision confidence: 100,0% 

   

  Duration: 3 minutes 08 seconds 

 

Analyzed in Rollout 

Player Winning Chances:    72,52% (G: 25,21% B:  1,48%) 

Opponent Winning Chances:  27,48% (G:  3,88% B:  0,14%) 

 

Cubeless Equities: No Double=+0,677, Double=+1,354 

 

Cubefull Equities: 

      No Double:   +0,972 (-0,028) 

      Double/Take: +1,187 (+0,187) 

      Double/Drop: +1,000 

 

Best Cube action: Double / Drop 

 

Rollout 

  3888 Games rolled with Variance Reduction. 

  Moves: 1 ply, cube decisions: 3 ply Red 

  Confidence No Double: ± 0,012 (+0,960  Confidence Double:    ± 0,022 (+1,165   

  Double Decision confidence: 100,0% 

  Take Decision confidence: 100,0% 

   

  Duration: 5 minutes 05 seconds 

 

Analyzed in Rollout 

Player Winning Chances:    72,44% (G: 25,15% B:  1,48%) 

Opponent Winning Chances:  27,56% (G:  3,85% B:  0,14%) 

 

Cubeless Equities: No Double=+0,675, Double=+1,350 

 

Cubefull Equities: 

      No Double:   +0,967 (-0,033) 

      Double/Take: +1,188 (+0,188) 

      Double/Drop: +1,000 

 

Best Cube action: Double / Drop 

 

Rollout 

  3888 Games rolled with Variance Reduction. 

  Moves: 1 ply, cube decisions: 3 ply 

  Confidence No Double: ± 0,011 (+0,956  Confidence Double:    ± 0,023 (+1,165   

  Double Decision confidence: 100,0% 

  Take Decision confidence: 100,0% 

   

  Duration: 9 minutes 07 seconds 

 

Analyzed in Rollout 

Player Winning Chances:    71,84% (G: 24,54% B:  1,54%) 

Opponent Winning Chances:  28,16% (G:  3,74% B:  0,14%) 

 

Cubeless Equities: No Double=+0,659, Double=+1,318 

 

Cubefull Equities: 

      No Double:   +0,957 (-0,043) 

      Double/Take: +1,091 (+0,091) 

      Double/Drop: +1,000 

 

Best Cube action: Double / Drop 

 

Rollout 

  3888 Games rolled with Variance Reduction. 

  Moves: 2 ply, cube decisions: 3 ply Red 

  Confidence No Double: ± 0,012 (+0,945  Confidence Double:    ± 0,022 (+1,069   

  Double Decision confidence: 100,0% 

  Take Decision confidence: 100,0% 

   

  Duration: 10 minutes 40 seconds 

 

Analyzed in Rollout 

Player Winning Chances:    71,79% (G: 24,57% B:  1,54%) 

Opponent Winning Chances:  28,21% (G:  3,70% B:  0,14%) 

 

Cubeless Equities: No Double=+0,659, Double=+1,317 

 

Cubefull Equities: 

      No Double:   +0,958 (-0,042) 

      Double/Take: +1,088 (+0,088) 

      Double/Drop: +1,000 

 

Best Cube action: Double / Drop 

 

Rollout 

  3888 Games rolled with Variance Reduction. 

  Moves: 2 ply, cube decisions: 3 ply 

  Confidence No Double: ± 0,012 (+0,947  Confidence Double:    ± 0,022 (+1,066   

  Double Decision confidence: 100,0% 

  Take Decision confidence: 100,0% 

   

  Duration: 14 minutes 07 seconds 

 

Analyzed in Rollout 

Player Winning Chances:    72,91% (G: 25,57% B:  1,61%) 

Opponent Winning Chances:  27,09% (G:  3,91% B:  0,13%) 

 

Cubeless Equities: No Double=+0,690, Double=+1,379 

 

Cubefull Equities: 

      No Double:   +0,993 (-0,007) 

      Double/Take: +1,217 (+0,217) 

      Double/Drop: +1,000 

 

Best Cube action: Double / Drop 

 

Rollout 

  3888 Games rolled with Variance Reduction. 

  Moves and cube decisions: 3 ply Red 

  Confidence No Double: ± 0,009 (+0,984  Confidence Double:    ± 0,020 (+1,197   

  Double Decision confidence: 93,6% 

  Take Decision confidence: 100,0% 

   

  Duration: 1 hour 32 minutes 26 seconds 

 

Analyzed in Rollout 

Player Winning Chances:    73,07% (G: 25,49% B:  1,53%) 

Opponent Winning Chances:  26,93% (G:  3,70% B:  0,13%) 

 

Cubeless Equities: No Double=+0,693, Double=+1,387 

 

Cubefull Equities: 

      No Double:   +0,993 (-0,007) 

      Double/Take: +1,264 (+0,264) 

      Double/Drop: +1,000 

 

Best Cube action: Double / Drop 

 

Rollout 

  3888 Games rolled with Variance Reduction. 

  Moves and cube decisions: 3 ply 

  Confidence No Double: ± 0,009 (+0,984  Confidence Double:    ± 0,020 (+1,244   

  Double Decision confidence: 93,6% 

  Take Decision confidence: 100,0% 

   

  Duration: 2 hours 30 minutes 58 seconds 

Messages In This Thread

• A more complicated backgame position to test yourself and BOTs
  neilkaz -- Thursday, 2 July 2009, at 4:09 p.m.
  • A more complicated backgame position to test yourself and BOTs
    Dan Pelton -- Thursday, 2 July 2009, at 4:32 p.m.
  • A more complicated backgame position to test yourself and BOTs
    jdg -- Thursday, 2 July 2009, at 6:05 p.m.
  • A more complicated backgame position to test yourself and BOTs
    lenny -- Friday, 3 July 2009, at 12:52 a.m.
  • GNU rollout and eval
    neilkaz -- Friday, 3 July 2009, at 11:15 p.m.
    • Snowie eval
      rew -- Saturday, 4 July 2009, at 7:47 a.m.
    • BGB 3 ply eval
      Neil Robins -- Saturday, 4 July 2009, at 7:48 a.m.
      • BGB 3 ply eval, too pessimistic again
        neilkaz -- Saturday, 4 July 2009, at 8:54 p.m.
    • XBG eval
      nixo -- Saturday, 4 July 2009, at 12:31 p.m.
    • xG evals + rollouts + quirk
      Robert-Jan Veldhuizen (Zorba) -- Sunday, 5 July 2009, at 10:23 a.m.
      • overview + conclusion
        Robert-Jan Veldhuizen (Zorba) -- Sunday, 5 July 2009, at 11:27 a.m.