[ View Thread ] [ Post Response ] [ Return to Index ] [ Read Prev Msg ] [ Read Next Msg ]

BGonline.org Forums

Perry's Lecture from Chicago (long)

Posted By: Stick
Date: Saturday, 30 May 2009, at 3:36 a.m.

The entire sequence without rollouts can still be viewed here but the following are both GNU & SW rollouts with the SW rollouts being provided by Perry. Also, the first couple of moves I'm posting from my site rollouts which will be money rollouts whereas Perry's quiz was actually a match to 25, this will make no difference. Perry's quiz was also technically -24 -25 but again, for my GNU ROs, I used -25 -25, no matter. I will also be omitting the moves I saw as clear nexts.

gnubg167


2O ' ' ' '5X '3X ' ' '5O

2X ' ' ' '5O '3O ' ' '5X

user167

Position ID: 4HPwATDgc/ABMA Match ID: MAELAAAAAAAA

# Ply Move Equity
1 R 24/18 13/11 +0.0193
0.50680.13020.0064-0.49320.13600.0055 +0.0087 +0.0193
0.00050.00050.0002-0.00050.00050.0002 0.0014 0.0035
Full cubeful rollout with var.redn.
15552 games, Mersenne Twister dice gen. with seed 988951872 and quasi-random dice
Play: 2-ply cubeful prune [world class]
keep the first 0 0-ply moves and up to 12 more moves within equity 0.16
Skip pruning for 1-ply moves.
Cube: 2-ply cubeful prune [world class]
2 R 24/16 -0.0060 ( -0.0253)
0.50280.11910.0053-0.49720.13680.0056 -0.0124 -0.0060
0.00050.00050.0001-0.00050.00060.0002 0.0013 0.0035
Full cubeful rollout with var.redn.
15552 games, Mersenne Twister dice gen. with seed 988951872 and quasi-random dice
Play: 2-ply cubeful prune [world class]
keep the first 0 0-ply moves and up to 12 more moves within equity 0.16
Skip pruning for 1-ply moves.
Cube: 2-ply cubeful prune [world class]
3 R 13/5 -0.0296 ( -0.0489)
0.49140.14400.0068-0.50860.14070.0076 -0.0148 -0.0296
0.00090.00090.0002-0.00090.00100.0005 0.0023 0.0062
Full cubeful rollout with var.redn.
5184 games, Mersenne Twister dice gen. with seed 988951872 and quasi-random dice
Play: 2-ply cubeful prune [world class]
keep the first 0 0-ply moves and up to 12 more moves within equity 0.16
Skip pruning for 1-ply moves.
Cube: 2-ply cubeful prune [world class]


gnubg159


2O ' ' ' '5X '3X ' '1X5O

1X ' ' ' '5O1X3O ' ' '4X

user167

Position ID: 4HPkQSDgc/ABMA Match ID: cAkFAAAAAAAA

# Ply Move Equity
1 R 13/11 8/7* -0.1024
0.47650.13310.0062-0.52350.15330.0097 -0.0707 -0.1024
0.00110.00090.0003-0.00110.00110.0004 0.0032 0.0079
Full cubeful rollout with var.redn.
5184 games, Mersenne Twister dice gen. with seed 1018937379 and quasi-random dice
Play: 2-ply cubeful prune [world class]
keep the first 0 0-ply moves and up to 12 more moves within equity 0.16
Skip pruning for 1-ply moves.
Cube: 2-ply cubeful prune [world class]
2 R 24/22 8/7* -0.1343 ( -0.0320)
0.47030.12390.0051-0.52970.14810.0074 -0.0860 -0.1343
0.00090.00090.0003-0.00090.00100.0003 0.0026 0.0065
Full cubeful rollout with var.redn.
5184 games, Mersenne Twister dice gen. with seed 1018937379 and quasi-random dice
Play: 2-ply cubeful prune [world class]
keep the first 0 0-ply moves and up to 12 more moves within equity 0.16
Skip pruning for 1-ply moves.
Cube: 2-ply cubeful prune [world class]


Snowie4 Professional Edition Version 4.7.1 HTML Export

Match to 25. Score Green-White: 0-1




Pip: 182
Game 1
25 point match

Green-White:
Score 0-1
Pip: 156
#PlyMoveEquity
1Rbar/20 8/7*-0.357
95% confidence interval:
- money cubeless eq.: -0.264 ±0.014.
Rollout settings:
Full rollout,
1944 games (equiv. 37580 games),
played 3-ply (precise),
random seed, without race database.
2Rbar/20 6/5-0.402 (-0.046)
95% confidence interval:
- money cubeless eq.: -0.287 ±0.012.
Rollout settings:
Full rollout,
1944 games (equiv. 45395 games),
played 3-ply (precise),
random seed, without race database.
3Rbar/20 24/23-0.435 (-0.078)
95% confidence interval:
- money cubeless eq.: -0.301 ±0.012.
Rollout settings:
Full rollout,
1944 games (equiv. 43752 games),
played 3-ply (precise),
random seed, without race database.
4Rbar/20 11/10-0.443 (-0.086)
95% confidence interval:
- money cubeless eq.: -0.309 ±0.014.
Rollout settings:
Full rollout,
1296 games (equiv. 31379 games),
played 3-ply (precise),
random seed, without race database.
5Rbar/24 13/8-0.503 (-0.146)
95% confidence interval:
- money cubeless eq.: -0.348 ±0.019.
Rollout settings:
Full rollout,
648 games (equiv. 17622 games),
played 3-ply (precise),
random seed, without race database.


This is the only problem UBK originally missed and one of the slew that I missed. As you'll note, the GNU rollout is not statistically significant, twice as many trials, and cubeful whereas the Snowie rollout is cubeless. I will not be extending any of the GNU rollouts myself, I pounded out enough variance to get within' a degree of certainty where the plays may be closer than Snowie thinks and I'm ok with that. The rollouts take too long since this game often is long and complicated. I still feel bar/21 13/11 is best and am somewhat relieved to see that GNU doesn't likely think it's the .04+ error Snowie did and it could even still be right according to one bot.

gnubg176


 ' ' '1O '5X1X1X ' '1X4O

2X ' ' '1X5O '3O ' '1O4X

Stick163

Position ID: 4CvygDDgc+QBQg Match ID: cAkxAwAAAAAA

# Ply Move Equity
1 R bar/23 21/17* +0.3175
0.56040.20510.0168-0.43960.10620.0047 +0.2327 +0.3175
0.00150.00150.0010-0.00150.00100.0003 0.0045 0.0104
Full cubeful rollout with var.redn.
5184 games, Mersenne Twister dice gen. with seed 859549127 and quasi-random dice
Play: 2-ply cubeful prune [world class]
keep the first 0 0-ply moves and up to 12 more moves within equity 0.16
Skip pruning for 1-ply moves.
Cube: 2-ply cubeful prune [world class]
2 R bar/21 13/11 +0.3050 ( -0.0124)
0.57100.16000.0098-0.42900.10010.0038 +0.2087 +0.3050
0.00100.00100.0004-0.00100.00080.0002 0.0026 0.0063
Full cubeful rollout with var.redn.
5184 games, Mersenne Twister dice gen. with seed 859549127 and quasi-random dice
Play: 2-ply cubeful prune [world class]
keep the first 0 0-ply moves and up to 12 more moves within equity 0.16
Skip pruning for 1-ply moves.
Cube: 2-ply cubeful prune [world class]

Snowie4 Professional Edition Version 4.7.1 HTML Export

Match to 25. Score Green-White: 0-1




Pip: 176
Game 1
25 point match

Green-White:
Score 0-1
Pip: 163
#PlyMoveEquity
1Rbar/23 21/17* 0.346
95% confidence interval:
- money cubeless eq.: 0.250 ±0.012.
Rollout settings:
Full rollout,
2352 games (equiv. 45548 games),
played 3-ply (precise),
random seed, without race database.
2Rbar/21 13/11 0.303 (-0.043)
95% confidence interval:
- money cubeless eq.: 0.221 ±0.010.
Rollout settings:
Full rollout,
2352 games (equiv. 58159 games),
played 3-ply (precise),
random seed, without race database.


There are two GNU rollouts posted below because the one with ~900 trials froze up on me and wouldn't extend. I'm posting it to not only lend more weight to my play and add a few trials, but to demonstrate why stat. sig. is important. The hitting play was often ahead during these rollouts and could still come out on top. Reaching stat. sig. with these is too much work but still interesting to compare to Snowie after a solid number of trials. As I told Perry after the quiz, I'm not sure what play is theoretically correct against an equal opponent but I know that I hit against the entire field.

gnubg193


 '1O ' ' '5X1X1O ' '1X4O

2X ' ' '1X5O '3O ' '1O4X

Stick157

Position ID: 4HPkIRDgC3lAWA Match ID: MAEtAwAAAAAA

# Ply Move Equity
1 R bar/23 11/8* -0.3206
0.43960.09890.0044-0.56040.21530.0170 -0.2511 -0.3206
0.00370.00320.0004-0.00370.00400.0021 0.0121 0.0292
Full cubeful rollout with var.redn.
862 games, Mersenne Twister dice gen. with seed 859589846 and quasi-random dice
Play: 2-ply cubeful prune [world class]
keep the first 0 0-ply moves and up to 12 more moves within equity 0.16
Skip pruning for 1-ply moves.
Cube: 2-ply cubeful prune [world class]
2 R bar/20 -0.3329 ( -0.0122)
0.43500.09950.0035-0.56500.19770.0100 -0.2358 -0.3329
0.00300.00220.0013-0.00300.00310.0009 0.0083 0.0170
Full cubeful rollout with var.redn.
866 games, Mersenne Twister dice gen. with seed 859589846 and quasi-random dice
Play: 2-ply cubeful prune [world class]
keep the first 0 0-ply moves and up to 12 more moves within equity 0.16
Skip pruning for 1-ply moves.
Cube: 2-ply cubeful prune [world class]

gnubg193


 '1O ' ' '5X1X1O ' '1X4O

2X ' ' '1X5O '3O ' '1O4X

Stick157

Position ID: 4HPkIRDgC3lAWA Match ID: MAEtAwAAAAAA

# Ply Move Equity
1 R bar/20 -0.3222
0.43820.09950.0042-0.56180.20040.0138 -0.2352 -0.3222
0.00160.00120.0005-0.00160.00160.0012 0.0053 0.0115
Full cubeful rollout with var.redn.
4543 games, Mersenne Twister dice gen. with seed 859096851 and quasi-random dice
Play: 2-ply cubeful prune [world class]
keep the first 0 0-ply moves and up to 12 more moves within equity 0.16
Skip pruning for 1-ply moves.
Cube: 2-ply cubeful prune [world class]
2 R bar/23 11/8* -0.3420 ( -0.0198)
0.43920.10210.0040-0.56080.22450.0204 -0.2619 -0.3420
0.00190.00120.0004-0.00190.00200.0014 0.0063 0.0138
Full cubeful rollout with var.redn.
4543 games, Mersenne Twister dice gen. with seed 859096851 and quasi-random dice
Play: 2-ply cubeful prune [world class]
keep the first 0 0-ply moves and up to 12 more moves within equity 0.16
Skip pruning for 1-ply moves.
Cube: 2-ply cubeful prune [world class]

Snowie4 Professional Edition Version 4.7.1 HTML Export

Match to 25. Score Green-White: 0-1




Pip: 193
Game 1
25 point match

Green-White:
Score 0-1
Pip: 157
#PlyMoveEquity
1Rbar/20-0.315
95% confidence interval:
- money cubeless eq.: -0.241 ±0.017.
Rollout settings:
Full rollout,
1296 games (equiv. 24792 games),
played 3-ply (precise),
random seed, without race database.
2Rbar/23 11/8*-0.388 (-0.073)
95% confidence interval:
- money cubeless eq.: -0.288 ±0.018.
Rollout settings:
Full rollout,
1296 games (equiv. 22282 games),
played 3-ply (precise),
random seed, without race database.


Snowie4 Professional Edition Version 4.7.1 HTML Export

Match to 25. Score Green-White: 0-1




Pip: 188
Game 1
25 point match

Green-White:
Score 0-1
Pip: 157
#PlyMoveEquity
1R17/14*/13 0.327
95% confidence interval:
- money cubeless eq.: 0.247 ±0.013.
Rollout settings:
Full rollout,
1944 games (equiv. 40391 games),
played 3-ply (precise),
random seed, without race database.
2R23/22 17/14* 0.308 (-0.019)
95% confidence interval:
- money cubeless eq.: 0.233 ±0.014.
Rollout settings:
Full rollout,
1944 games (equiv. 37133 games),
played 3-ply (precise),
random seed, without race database.
3R17/14* 11/10 0.287 (-0.040)
95% confidence interval:
- money cubeless eq.: 0.217 ±0.014.
Rollout settings:
Full rollout,
1944 games (equiv. 37072 games),
played 3-ply (precise),
random seed, without race database.
4R13/10 11/10 0.260 (-0.068)
95% confidence interval:
- money cubeless eq.: 0.189 ±0.049.
Rollout settings:
Full rollout,
108 games (equiv. 2894 games),
played 3-ply (precise),
seed 1, without race database.
5R17/14* 8/7 0.257 (-0.071)
95% confidence interval:
- money cubeless eq.: 0.191 ±0.058.
Rollout settings:
Full rollout,
108 games (equiv. 2198 games),
played 3-ply (precise),
seed 1, without race database.


Here we see an astounding difference of opinion and once again, I'm relieved. No real stat. sig but again, enough variance pounded out of it that it's higggggggggggghly unlikely to be the near blunder Snowie says it is. I'm happy because I hit and if I went passively into the night taking Snowie's rollouts as the final word I'd have done for shit on this quiz. As it is, I may yet recoup my money from UBK :)

gnubg202


 '1O ' ' '5X1X ' ' ' '5O

2X ' ' '2X5O '3O ' '1O4X

user153

Position ID: 4HPkAxDgCzxgWA Match ID: MIE4AwAAAAAA

# Ply Move Equity
1 R bar/24 20/14* -0.3428
0.44330.09920.0045-0.55670.23550.0243 -0.2709 -0.3428
0.00200.00120.0004-0.00200.00190.0013 0.0062 0.0135
Full cubeful rollout with var.redn.
4212 games, Mersenne Twister dice gen. with seed 859181552 and quasi-random dice
Play: 2-ply cubeful prune [world class]
keep the first 0 0-ply moves and up to 12 more moves within equity 0.16
Skip pruning for 1-ply moves.
Cube: 2-ply cubeful prune [world class]
2 R bar/24 13/7 -0.3527 ( -0.0100)
0.42670.09620.0035-0.57330.19110.0119 -0.2510 -0.3527
0.00130.00100.0002-0.00130.00130.0005 0.0035 0.0084
Full cubeful rollout with var.redn.
4212 games, Mersenne Twister dice gen. with seed 859181552 and quasi-random dice
Play: 2-ply cubeful prune [world class]
keep the first 0 0-ply moves and up to 12 more moves within equity 0.16
Skip pruning for 1-ply moves.
Cube: 2-ply cubeful prune [world class]

Snowie4 Professional Edition Version 4.7.1 HTML Export

Match to 25. Score Green-White: 0-1




Pip: 202
Game 1
25 point match

Green-White:
Score 0-1
Pip: 153
#PlyMoveEquity
1Rbar/24 13/7-0.347
95% confidence interval:
- money cubeless eq.: -0.261 ±0.015.
Rollout settings:
Full rollout,
1296 games (equiv. 29399 games),
played 3-ply (precise),
random seed, without race database.
2Rbar/18-0.392 (-0.044)
95% confidence interval:
- money cubeless eq.: -0.299 ±0.016.
Rollout settings:
Full rollout,
1296 games (equiv. 25280 games),
played 3-ply (precise),
random seed, without race database.
3Rbar/24 20/14*-0.407 (-0.060)
95% confidence interval:
- money cubeless eq.: -0.304 ±0.018.
Rollout settings:
Full rollout,
1296 games (equiv. 20331 games),
played 3-ply (precise),
random seed, without race database.


I would have had more trials on this also but the first rollout also crashed. The first rollout had basically the same figures as are shown here in this 2k trial rollout. I instinctively ran on the quiz but upon seeing the other play set up preferred it. That was one of a couple hitches in the format, time press where I didn't want to hold everyone up and you're staring at a screen, you can't actually set up the plays. (I didn't have my board with me at the time and I didn't think ppl would think it's fair if I was setting it up on my computer lol) This may be a reason UBK did so well, cheating bastard had a board in front of him and set up the plays.

gnubg195


 '1O ' ' '5X2X ' ' ' '5O

3X ' ' '2X5O '3O ' '1O3X

user153

Position ID: 4Bs4YDjgc+QDEA Match ID: cIkxAwAAAAAA

# Ply Move Equity
1 R 13/9 11/8 +0.2818
0.55610.18690.0120-0.44390.10660.0042 +0.2012 +0.2818
0.00160.00150.0006-0.00160.00130.0005 0.0045 0.0107
Full cubeful rollout with var.redn.
2592 games, Mersenne Twister dice gen. with seed 859181552 and quasi-random dice
Play: 2-ply cubeful prune [world class]
keep the first 0 0-ply moves and up to 12 more moves within equity 0.16
Skip pruning for 1-ply moves.
Cube: 2-ply cubeful prune [world class]
2 R 23/16 +0.2631 ( -0.0187)
0.55870.16580.0110-0.44130.11360.0042 +0.1768 +0.2631
0.00150.00150.0007-0.00150.00130.0003 0.0043 0.0102
Full cubeful rollout with var.redn.
2592 games, Mersenne Twister dice gen. with seed 859181552 and quasi-random dice
Play: 2-ply cubeful prune [world class]
keep the first 0 0-ply moves and up to 12 more moves within equity 0.16
Skip pruning for 1-ply moves.
Cube: 2-ply cubeful prune [world class]

Snowie4 Professional Edition Version 4.7.1 HTML Export

Match to 25. Score Green-White: 0-1




Pip: 195
Game 1
25 point match

Green-White:
Score 0-1
Pip: 153
#PlyMoveEquity
1R13/9 11/8 0.293
95% confidence interval:
- money cubeless eq.: 0.220 ±0.012.
Rollout settings:
Full rollout,
1944 games (equiv. 44008 games),
played 3-ply (precise),
random seed, without race database.
2R11/4 0.269 (-0.024)
95% confidence interval:
- money cubeless eq.: 0.200 ±0.012.
Rollout settings:
Full rollout,
1944 games (equiv. 45316 games),
played 3-ply (precise),
random seed, without race database.
3R23/16 0.266 (-0.027)
95% confidence interval:
- money cubeless eq.: 0.195 ±0.012.
Rollout settings:
Full rollout,
1944 games (equiv. 45277 games),
played 3-ply (precise),
random seed, without race database.

Any individual plays you'd like to discuss I suggest making a separate posting in this thread and entitle it "61 from quiz" and repost the position. I didn't want to get into analyzing the positions since this post is already so long.

Stick

Messages In This Thread

 

Post Response

Your Name:
Your E-Mail Address:
Subject:
Message:

If necessary, enter your password below:

Password:

 

 

[ View Thread ] [ Post Response ] [ Return to Index ] [ Read Prev Msg ] [ Read Next Msg ]

BGonline.org Forums is maintained by Stick with WebBBS 5.12.