BGonline.org Forums

* * * Rollout, Surprise and Mystery... * * *

Posted By: Klaus Evers
Date: Tuesday, 15 January 2008, at 9:46 p.m.

In Response To: Do you think your bot can handle this roll? (Klaus Evers)

The score (after 5 games) is: Snowie 0, You 7

Message at bottom!

Move number 77: You to play 32

Pip counts: Snowie 30, You 33
Position ID: HwAAoO02AAAAAA Match ID: UYkJAAAAOAAA

• You moves 5/3 5/2

Alert: doubtful move ( -0,073)

#		Ply	Move	Equity
	1	2	5/off	+0,254
			0,515 0,000 0,000 - 0,485 0,000 0,000
			2-ply cubeful
	2	2	3/off 2/off	+0,233 ( -0,020)
			0,523 0,000 0,000 - 0,477 0,000 0,000
			2-ply cubeful
•	3	2	5/3 5/2	+0,180 ( -0,073)
			0,510 0,000 0,000 - 0,490 0,000 0,000
			2-ply cubeful

Alert: doubtful move ( -0,031)

#		Ply	Move	Equity
	1	R	5/off	+0,307
			0,519 0,000 0,000 - 0,481 0,000 0,000 +0,038 +0,307 0,001 0,000 0,000 - 0,001 0,000 0,000 0,001 0,002
			Full cubeful rollout with var.redn.
			1296 games, Mersenne Twister dice gen. with seed 1050869504 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]
•	2	R	5/3 5/2	+0,277 ( -0,031)
			0,520 0,000 0,000 - 0,480 0,000 0,000 +0,040 +0,277 0,001 0,000 0,000 - 0,001 0,000 0,000 0,001 0,003
			Full cubeful rollout with var.redn.
			1296 games, Mersenne Twister dice gen. with seed 1050869504 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]
	3	R	3/off 2/off	+0,270 ( -0,037)
			0,525 0,000 0,000 - 0,475 0,000 0,000 +0,051 +0,270 0,001 0,000 0,000 - 0,001 0,000 0,000 0,001 0,003
			Full cubeful rollout with var.redn.
			1296 games, Mersenne Twister dice gen. with seed 1050869504 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]

First of all, from my experience I knew that GnuBG has some problems with this kind of position.
The equities on 2-ply can not be trusted in these bearoff positions, so that made me create a rollout.

Move positions 2 and 3 change places in RO, and the equity difference among the top3 moves change
a lot compared to 2-ply.

But now, take a close look at the numbers:

How can it be, that in this simple positions, with all gammons gone, the play that wins the most games (52.5%) is only ranked as #3 play?

At DMP it is in the order that you would expect it:

DMP:

#		Ply	Move	Equity
	1	R	3/off 2/off	+0,050
			0,525 0,000 0,000 - 0,475 0,000 0,000 +0,050 +0,050 0,001 0,000 0,000 - 0,001 0,000 0,000 0,001 0,001
			Full cubeful rollout with var.redn.
			1296 games, Mersenne Twister dice gen. with seed 1050776256 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.
	2	R	5/3 5/2	+0,040 ( -0,010)
			0,520 0,000 0,000 - 0,480 0,000 0,000 +0,040 +0,040 0,001 0,000 0,000 - 0,001 0,000 0,000 0,001 0,001
			Full cubeful rollout with var.redn.
			1296 games, Mersenne Twister dice gen. with seed 1050776256 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.
	3	R	5/off	+0,039 ( -0,011)
			0,520 0,000 0,000 - 0,480 0,000 0,000 +0,039 +0,039 0,001 0,000 0,000 - 0,001 0,000 0,000 0,001 0,001
			Full cubeful rollout with var.redn.
			1296 games, Mersenne Twister dice gen. with seed 1050776256 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.

Who can shed some light on this mystery?
Thanks.

Cheers, Klaus

Side note:

You get very similar results with Snowie 4.6, but I am absolutely not familiar with SW export and html, so I left this out.

Messages In This Thread

• Do you think your bot can handle this roll?
  Klaus Evers -- Sunday, 13 January 2008, at 10:29 a.m.
  • Do you think your bot can handle this roll?
    Havard Raddum -- Sunday, 13 January 2008, at 8:49 p.m.
    • Do you think your bot can handle this roll?
      scuba -- Monday, 14 January 2008, at 7:24 p.m.
      • Do you think your bot can handle this roll?
        Klaus Evers -- Monday, 14 January 2008, at 8:20 p.m.
  • Do you think your bot can handle this roll?
    Caryl Stephanson -- Monday, 14 January 2008, at 3:48 a.m.
    • Do you think your bot can handle this roll?
      neilkaz -- Monday, 14 January 2008, at 4:37 a.m.
      • Do you think your bot can handle this roll?
        Caryl Stephanson -- Tuesday, 15 January 2008, at 6:27 p.m.
  • Do you think your bot can handle this roll?
    Chuck Bower -- Monday, 14 January 2008, at 2:09 p.m.
  • Do you think your bot can handle this roll?
    Keene -- Monday, 14 January 2008, at 8:24 p.m.
  • * * * Rollout, Surprise and Mystery... * * *
    Klaus Evers -- Tuesday, 15 January 2008, at 9:46 p.m.
    • * * * Rollout, Surprise and Mystery... * * *
      Stick -- Wednesday, 16 January 2008, at 1:28 a.m.
      • * * * Rollout, Surprise and Mystery... * * *
        Klaus Evers -- Wednesday, 16 January 2008, at 5:28 p.m.
    • look at Neil's post for the explanation (nt) *NM*
      Peter G. -- Wednesday, 16 January 2008, at 9:09 a.m.
    • * * * Rollout, Surprise and Mystery... * * *
      Daniel Murphy -- Wednesday, 16 January 2008, at 11:49 a.m.
      • * * * Rollout, Surprise and Mystery... * * *
        Klaus Evers -- Wednesday, 16 January 2008, at 9:15 p.m.
        • * * * Rollout, Surprise and Mystery... * * *
          Daniel Murphy -- Thursday, 17 January 2008, at 3:11 a.m.