BGonline.org Forums

Mochy vs Melzi @GG wednesday 6pm edt

Posted By: eXtreme Gammon
Date: Friday, 24 June 2011, at 2:55 a.m.

In Response To: Mochy vs Melzi @GG wednesday 6pm edt (Massimiliano Calusio)

This is an interesting case for discussing RO results.

---

Here is a 3-ply RO

		White is Joueur 2 score: 1 pip: 99 13 point match pip: 59 score: 10 Blue is Joueur 1
XGID=aBCCCCa--------A---dbbcb--:1:1:1:11:10:1:0:13:10
Blue to play 11

1.	Rollout1	15/13 5/4 2/1	eq: +0.998
	Player: Opponent: 92.24% (G:11.11% B:0.03%) 7.76% (G:0.20% B:0.00%) Conf.: ± 0.002 (+0.996...+1.001) - [100.0%] Duration: 46.6 seconds
2.	Rollout1	15/13 5/3	eq: +0.988 (-0.010)
	Player: Opponent: 92.11% (G:11.06% B:0.02%) 7.89% (G:0.21% B:0.00%) Conf.: ± 0.002 (+0.986...+0.991) - [0.0%] Duration: 43.1 seconds
3.	Rollout1	15/13 5/4 3/2	eq: +0.882 (-0.116)
	Player: Opponent: 91.54% (G:9.46% B:0.11%) 8.46% (G:0.20% B:0.00%) Conf.: ± 0.001 (+0.882...+0.883) - [0.0%] Duration: 38.8 seconds
4.	Rollout1	15/12 5/4	eq: +0.873 (-0.125)
	Player: Opponent: 90.23% (G:15.01% B:0.19%) 9.77% (G:0.24% B:0.01%) Conf.: ± 0.001 (+0.873...+0.874) - [0.0%] Duration: 15.0 seconds
1 5184 Games rolled with Variance Reduction. Moves and cube decisions: 3-ply

eXtreme Gammon Version: 2.00.170.beta, MET: Kazaross XG2

---

and now a 3-ply/XGR RO

		White is Joueur 2 score: 1 pip: 99 13 point match pip: 59 score: 10 Blue is Joueur 1
XGID=aBCCCCa--------A---dbbcb--:1:1:1:11:10:1:0:13:10
Blue to play 11

1.	Rollout1	15/13 5/4 2/1	eq: +0.907
	Player: Opponent: 91.34% (G:10.24% B:0.08%) 8.66% (G:0.18% B:0.00%) Conf.: ± 0.001 (+0.906...+0.907) - [70.3%] Duration: 31.8 seconds
2.	Rollout1	15/13 5/3	eq: +0.906 (0.000)
	Player: Opponent: 91.19% (G:9.97% B:0.07%) 8.81% (G:0.20% B:0.00%) Conf.: ± 0.001 (+0.906...+0.907) - [21.0%] Duration: 29.9 seconds
3.	Rollout1	15/13 5/4 3/2	eq: +0.906 (0.000)
	Player: Opponent: 90.98% (G:9.54% B:0.06%) 9.02% (G:0.20% B:0.00%) Conf.: ± 0.001 (+0.906...+0.907) - [8.7%] Duration: 32.1 seconds
4.	Rollout1	15/12 5/4	eq: +0.892 (-0.015)
	Player: Opponent: 91.05% (G:13.96% B:0.27%) 8.95% (G:0.23% B:0.01%) Conf.: ± 0.001 (+0.891...+0.893) - [0.0%] Duration: 1 minute 42 seconds
1 5184 Games rolled with Variance Reduction. Moves: 3-ply, cube decisions: XG Roller

eXtreme Gammon Version: 2.00.170.beta, MET: Kazaross XG2

---

That's widely different results, so let's try to figure why... let's only compare 15/13 5/4 3/2 and 15/13 5/4 2/1

As the only change is the cube level, it has to come for the next cube decision and very likely is in the line after the opponent fans.

Let's check the 2 positions

White is Joueur 2 score: 1 pip: 99 13 point match pip: 55 score: 10 Blue is Joueur 1 XGID=aCBCDBa------A-----dbbcb--:1:1:1:00:10:1:0:13:10 Blue on roll, cube action? eXtreme Gammon Version: 2.00.170.beta	White is Joueur 2 score: 1 pip: 99 13 point match pip: 55 score: 10 Blue is Joueur 1 XGID=aBDBDBa------A-----dbbcb--:1:1:1:00:10:1:0:13:10 Blue on roll, cube action? eXtreme Gammon Version: 2.00.170.beta
15/13 5/4 2/1 3ply: ND 0.963, DT 0.982 XGR: ND 0.972, DT 1.106 3-ply/XGR 1296 RO: ND: 0.980, DT: 1.129	15/13 5/4 3/2 3ply: ND 0.958, DT 0.946 XGR: ND 0.966, DT 1.094 3-ply/XGR 1296 RO: ND: 0.971, DT: 1.108

---

Here it happens that XGR get both cube right (DP), while 3-ply get DT and ND for 3/2 and 2/1.

As a result 2/1 Eval in 3-ply RO are inflated by 0.129 and 3/2 is decreased by 1.000-0.958=0.042. the rest of the error are probably from susquent play-fan-double error

Conclusion: 3-ply RO are not perfect. 3-ply/XGR Ro are better but are still not perfect!

The position shown here has a very narrow doubling window (ATS) and it makes the cube action very sharp/difficult and hence increases the risk that during RO the bot will err.

Messages In This Thread

• Mochy vs Melzi @GG wednesday 6pm edt
  Leonardo Jerkovic -- Tuesday, 21 June 2011, at 9:23 p.m.
  • Mochy vs Melzi @GG wednesday 6pm edt
    Bill Patterson -- Wednesday, 22 June 2011, at 3:19 a.m.
    • +1
      Klaus Evers -- Wednesday, 22 June 2011, at 6:31 a.m.
    • Mochy vs Melzi @GG wednesday 6pm edt
      Mochy -- Thursday, 23 June 2011, at 11:56 a.m.
      • Mochy vs Melzi @GG wednesday 6pm edt
        Mochy -- Thursday, 23 June 2011, at 11:57 a.m.
        • Melzi to play 1-1 (SPOILER)
          Carlo Melzi -- Thursday, 23 June 2011, at 12:38 p.m.
          • Melzi to play 1-1 (SPOILER)
            Carlo Melzi -- Thursday, 23 June 2011, at 12:42 p.m.
            • Melzi to play 1-1 (SPOILER)
              Leonardo Jerkovic -- Friday, 24 June 2011, at 1:00 a.m.
      • Mochy vs Melzi @GG wednesday 6pm edt
        Massimiliano Calusio -- Thursday, 23 June 2011, at 4:47 p.m.
        • XGR++ versus a full rollout
          Timothy Chow -- Friday, 24 June 2011, at 1:38 a.m.
        • Mochy vs Melzi @GG wednesday 6pm edt
          eXtreme Gammon -- Friday, 24 June 2011, at 2:55 a.m.
          • Mochy vs Melzi @GG wednesday 6pm edt
            Mochy -- Friday, 24 June 2011, at 5:34 a.m.
            • Mochy vs Melzi @GG wednesday 6pm edt
              eXtreme Gammon -- Friday, 24 June 2011, at 12:56 p.m.
              • How many time does 3-ply and XGR++ choice get overturned by RO
                eXtreme Gammon -- Saturday, 25 June 2011, at 2:15 p.m.
                • How many time does 3-ply and XGR++ choice get overturned by RO - Final version
                  eXtreme Gammon -- Monday, 27 June 2011, at 2:01 a.m.
          • XGR+ and XGR++ RO
            eXtreme Gammon -- Friday, 24 June 2011, at 2:09 p.m.
  • Mochy vs Melzi RESULT
    Leonardo Jerkovic -- Wednesday, 22 June 2011, at 11:54 p.m.