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

BGonline.org Forums

Foul up these aces

Posted By: Stick
Date: Thursday, 24 February 2011, at 3:30 p.m.

In Response To: Foul up these aces (ah_clem)

BTW, why not roll it out? Because you don't believe the bots can play this accurately?

From my experience bots on 3 ply can really foul up backgame positions, XG included, and I'll include some wtf plays at the end of this post. If I wanted to try to roll this out the first thing I would do is jack up my settings to 4 ply play and 4 ply cube and have a huge search space for the plays. Ideally I'd rather have an XGR+ rollout but the amount of time that 4 ply takes is already impossible, XGR+ would be brutal.

This position for example I would have to roll out 5 plays, already a lot. Now I'd have to roll them out on the hopped up settings expecting a result sometime next month if I'm lucky. This still wouldn't mean as much to me as getting feedback from a couple of the best players in the world because of how the bot can sometimes play out these positions. I put forth a few positions now with their 3 ply analysis and a short word after each position.

is Player 2

score: 0
pip: 147
Money session
Jacoby Beaver
pip: 212
score: 0

is Player 1
XGID=--a-aBB-B-B-c----cbcCbAAAA:0:0:1:42:0:0:3:0:10
to play 42

1.3 plyBar/23 20/16eq: -0.308
Player:
Opponent:		 45.84% (G:8.53% B:0.35%)
54.16% (G:25.39% B:2.08%)
2.3 plyBar/23 24/20eq: -0.362 (-0.054)
Player:
Opponent:		 43.87% (G:8.38% B:0.34%)
56.13% (G:24.62% B:1.53%)

XGR+ reverses this to having 23, 20 as the best play by a small amount. This to me is a standard play and the fact that the bot can think on 3 ply that my play is wrong by .05 is disgusting.

is Player 2

score: 0
pip: 116
Money session
Jacoby Beaver
pip: 179
score: 0

is Player 1
XGID=--aaABB-BBA-b-----adBbBbbA:0:0:1:33:0:0:3:0:10
to play 33

1.3 plyBar/22 20/17 10/4eq: +0.278
Player:
Opponent:		 61.02% (G:19.40% B:1.12%)
38.98% (G:22.11% B:0.83%)
2.3 plyBar/22 10/4 9/6eq: +0.177 (-0.100)
Player:
Opponent:		 57.07% (G:16.74% B:0.79%)
42.93% (G:19.74% B:0.77%)

Totally clueless. A play I'd expect MCG to make. XGR+ reverses this and makes its own 3 ply play a blunder.

is Player 2

score: 0
pip: 144
Money session
Jacoby Beaver
pip: 191
score: 0

is Player 1
XGID=-----BBcB-C-b--A-cbcBbAB--:0:0:1:62:0:0:3:0:10
to play 62

1.3 ply22/20 15/9eq: -0.463
Player:
Opponent:		 37.97% (G:6.58% B:0.14%)
62.03% (G:16.84% B:0.74%)
2.3 ply22/16 15/13eq: -0.477 (-0.014)
Player:
Opponent:		 39.12% (G:6.53% B:0.17%)
60.88% (G:19.81% B:1.01%)
3.3 ply22/14eq: -0.477 (-0.014)
Player:
Opponent:		 38.69% (G:6.46% B:0.16%)
61.31% (G:19.06% B:0.93%)

Now the difference in plays according to XGR+ isn't all that much, ~.050, but if the bot on 3 ply doesn't know to flood the outfield in this type of position how can I trust it to play out these games. This is a play I'd next.

is Player 2

score: 0
pip: 40
Money session
Jacoby Beaver
pip: 198
score: 0

is Player 1
XGID=-----BB-B-B-----A-bBBcbBh-:1:1:1:65:0:0:3:0:10
to play 65

1.3 ply19/14 19/13eq: -0.272
Player:
Opponent:		 47.66% (G:6.27% B:0.04%)
52.34% (G:44.03% B:0.38%)
2.3 ply16/5eq: -0.315 (-0.043)
Player:
Opponent:		 46.41% (G:6.09% B:0.10%)
53.59% (G:45.31% B:0.67%)
3.3 ply16/11 10/4eq: -0.328 (-0.057)
Player:
Opponent:		 45.95% (G:6.23% B:0.13%)
54.05% (G:45.72% B:0.85%)

Really? 14, 13? Get your head out of your ass.

is Player 2

score: 0
pip: 74
Money session
Jacoby Beaver
pip: 199
score: 0

is Player 1
XGID=----aBB-B-B----Aa-b-CeCdb-:2:-1:1:55:0:0:3:0:10
to play 55

1.3 ply22/12 20/15(2)eq: -0.003
Player:
Opponent:		 62.41% (G:11.40% B:0.56%)
37.59% (G:23.89% B:0.18%)
2.3 ply20/15(2) 20/10eq: -0.020 (-0.017)
Player:
Opponent:		 60.66% (G:11.72% B:0.70%)
39.34% (G:22.26% B:0.16%)
3.3 ply22/12 20/10eq: -0.033 (-0.030)
Player:
Opponent:		 59.85% (G:11.43% B:0.52%)
40.15% (G:21.09% B:0.14%)

Again the bot wants to give up one of the two anchors. Donkey.

is Player 2

score: 0
pip: 69
Money session
Jacoby Beaver
pip: 215
score: 0

is Player 1
XGID=-----BC--aB---b----ACeDdc-:3:1:1:51:0:0:3:0:10
to play 51

1.3 ply22/17 10/9* eq: -0.135
Player:
Opponent:		 48.53% (G:7.79% B:0.38%)
51.47% (G:34.55% B:0.53%)
2.3 ply10/5 10/9* eq: -0.155 (-0.020)
Player:
Opponent:		 48.72% (G:7.12% B:0.39%)
51.28% (G:36.25% B:0.70%)
3.3 ply22/17 20/19eq: -0.221 (-0.086)
Player:
Opponent:		 46.65% (G:5.58% B:0.39%)
53.35% (G:36.77% B:0.45%)

I'm sorry...my play is a blunder? XGR+ has the non hitting play best by ~.020.

is Player 2

score: 0
pip: 33
Money session
Jacoby Beaver
pip: 191
score: 0

is Player 1
XGID=----BBB-A---A----AA-CcfBc-:1:1:1:63:0:0:3:0:10
to play 63

1.3 ply17/8eq: -1.546
Player:
Opponent:		 12.18% (G:0.00% B:0.00%)
87.82% (G:78.26% B:4.87%)
2.3 ply18/15 17/11eq: -1.548 (-0.002)
Player:
Opponent:		 11.94% (G:0.00% B:0.00%)
88.06% (G:77.98% B:4.74%)

Now it doesn't matter much which play XG picked on 3 ply here I'll admit but when my play doesn't make it through the filter of candidates and it stands as a whopping .472 blunder on 2 ply I take offense. XGR+ has my play, 14, 15, as slightly best when forced to analyze it.

These type of positions and much worse, my favorite being plays that break the 6pt for no good reason, go on and on and on and make rolling out any type of major backgammon a wasted effort.

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.