obvious is not always right - Page 2 - Backgammon Forum - Discuss Backgammon Strategy, Clubs and Books

That big stack on the 20 point just looks like it's desperate to release a checker. Unfortunately, you have to step into a triple shot and make a concession on the home front to move one. I kind of think it's worth it though. So I'd play 20/15 7/6. The two blots in White's board are key. If his board was even a little more together, I'd just play 18/12.

There is some duplication here too. After 20/15 7/6, white hits with 6, 4, and 1; and covers with 6, 5, 4, and 1.

Quote

12-11-2014 , 01:18 PM

#27

peachpie

centurion

Join Date: Aug 2014 Posts: 182

Quote:

Originally Posted by yogiman

[CENTER]White - Pips 60
With a GNU untruncated rollout of 400+ at worldclass, the verdict is: NO REDOUBLE. Doubling and wisely taking would cost black 0.28.

My doubling needs some work, what else is new.

Quote

12-13-2014 , 11:47 AM

#28

yogiman

adept

Join Date: Jun 2012 Posts: 784

Here is another interesting situation

White - Pips 111

Black - Pips 142
Black to Play 4-3
Created with www.BGdiagram.com

Quote

12-13-2014 , 12:06 PM

#29

_Z_

adept

Join Date: Jul 2004 Posts: 731

Sort of a lot of options, but I would go with 10/6 24/21 over the board. Keeps most assets in place, start mobilizing the back men and it's not particularly dangerous. Quiz Factor -- Maybe we're not supposed to give up the longer term safety of the 24 point though? Maybe 9/6 9/5?? Actually, that play leads to a good cube on the dances which mine does not. That could be key.

Quote

12-13-2014 , 06:30 PM

#30

yogiman

adept

Join Date: Jun 2012 Posts: 784

According to an extensive rollout not giving up the 18 point is a big blunder, and 18/11 is best by far.

Quote

12-13-2014 , 11:25 PM

#31

Taper_Mike

adept

Join Date: Sep 2009 Posts: 1,030

Quote:

Originally Posted by yogiman

According to an extensive rollout not giving up the 18 point is a big blunder, and 18/11 is best by far.

This is another nice position from the Yogiman!

I was just about to suggest (with some reservations) that 18/11 would be my play when I saw this.

Breaking the 24pt must be wrong. Look how stripped White is outside. His only spares are inside, where he would love the chance to use them in an attack. Splitting the anchor is the play White wants Black to make.

Down by 31 pips, Black needs to make his 5pt before he can turn the corner on this game. So long as it is open, White is just one or two rolls away from escaping with the lead in his pocket. 18/11 brings another builder to the scene of the action. It must be better than plays that break the 9pt in order to create a new builder. Risks are minimal. From the bar, White will have 8 hitting numbers: 65, 61, 51, 55, and 11. Most of leave them return shots, sometimes many.

Mike

Quote

12-20-2014 , 09:51 AM

#32

yogiman

adept

Join Date: Jun 2012 Posts: 784

Playing with the bot you might get this one still wrong.

White - Pips 94. Match Score 0/64

Black - Pips 109. Match Score 1/64
Black to Play 5-2
Created with www.BGdiagram.com

Quote

12-20-2014 , 06:04 PM

#33

yogiman

adept

Join Date: Jun 2012 Posts: 784

when heroism is required:

White - Pips 92. Match Score 1/64

Black - Pips 123. Match Score 1/64
Black to Play 6-3
Created with www.BGdiagram.com

Quote

12-21-2014 , 06:37 AM

#34

yogiman

adept

Join Date: Jun 2012 Posts: 784

Actually this was a late one. Ofcourse it belongs in my other thread.

Quote

12-22-2014 , 06:36 AM

#35

Taper_Mike

adept

Join Date: Sep 2009 Posts: 1,030

White - Pips 92. Match Score 1/64

Black - Pips 123. Match Score 1/64
Black to Play 6-3

I like jumping out to the 16pt. That leaves a direct shot (and an indirect shot) outside, but avoids a triple shot on the 22pt. If White fails to hit, Black will get often get a direct shot of his own.

Mike

Quote

12-26-2014 , 05:22 AM

#36

yogiman

adept

Join Date: Jun 2012 Posts: 784

White - Pips 51. Match Score 0/64

Black - Pips 122. Match Score 0/64
Black to Play 4-3
Created with www.BGdiagram.com

Quote

12-27-2014 , 07:04 AM

#37

yogiman

adept

Join Date: Jun 2012 Posts: 784

Code:

    1.5/2 5/1*                     Eq.:  +0,959
       0,794 0,000 0,000 - 0,206 0,025 0,012 CL  +0,550 CF  +0,959
     
    2.17/14 11/7                   Eq.:  +0,906 ( -0,053)
       0,774 0,000 0,000 - 0,226 0,026 0,008 CL  +0,514 CF  +0,906
     
    3.10/6 7/4                     Eq.:  +0,881 ( -0,078)
       0,773 0,000 0,000 - 0,227 0,026 0,008 CL  +0,511 CF  +0,881

        Truncated cubeful rollout (depth 10) with var.redn.
        147 games, Mersenne Twister dice gen. with seed 766752762 and quasi-random dice
        Stop when std.errs. are small enough: ratio 0,1 (min. 144 games)
        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]

I included this problem, because the cubeful 0-ply showed a much greater difference. But the message is still clear. White shouldn't be allowed to make that 1-point, because black likes to be hit, plus he could spoil black's party by a last hit and/or some big doubles.

Quote

12-28-2014 , 09:10 AM

#38

yogiman

adept

Join Date: Jun 2012 Posts: 784

You guys were right not to respond to post #32. Though 0-ply noted 20/15 13/11 superior, a rollout brought to light that 13/6 is better with 0.1. You have remarkable QF-noses.

Quote

12-28-2014 , 11:52 AM

#39

yogiman

adept

Join Date: Jun 2012 Posts: 784

There are only two possibilities.

White - Pips 177. Match Score 0/64

Black - Pips 95. Match Score 0/64
Black to Play 6-1
Created with www.BGdiagram.com

Quote

12-28-2014 , 05:08 PM

#40

yogiman

adept

Join Date: Jun 2012 Posts: 784

Code:

    1.bar/18                       Eq.:  -0,085
       0,478 0,247 0,039 - 0,522 0,112 0,004 CL  +0,127 CF  -0,085
      
    2.bar/24 11/5                  Eq.:  -0,208 ( -0,124)
       0,436 0,226 0,030 - 0,564 0,095 0,003 CL  +0,029 CF  -0,208
      
        Truncated cubeful rollout (depth 10) with var.redn.
        147 games, Mersenne Twister dice gen. with seed 766568734 and quasi-random dice
        Stop when std.errs. are small enough: ratio 0,1 (min. 144 games)
        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]

Quote

12-28-2014 , 05:56 PM

#41

Taper_Mike

adept

Join Date: Sep 2009 Posts: 1,030

GnuBgID for Problem 10
2GYMgjhu7oYAQA:cAkHAAAAAAAE

This ID is not quite the one in Yogiman's problem 10 (post #39). Instead of setting match length to 64, I turned off the Jacoby Rule. This is an unlimited game, no Jacoby, no beavers.

Settings > Options > Cube > Use the Jacoby rule > Remove check mark
Settings > Options > Cube > Maximum number of beavers > 0

Mike

Quote

12-29-2014 , 06:10 AM

#42

yogiman

adept

Join Date: Jun 2012 Posts: 784

Okay, I put your settings. But whatever settings, whether there is no cube, or whether black has the cube, the outcome will be the same. And the reason is that the safe play has a lot of board crumbling potential.

That lonely checker will get stuck behind a growing prime, and white has two weakness that can be exploited:
-he has three checkers semi-stuck on black's one point
-he only has a 3-point homeboard with no extension

Black needs flexibility, and needs to hit a blot as he has the better board. If he plays bar/18 he not only serves his flexibility. In most cases white will leave a blot on the 18-point, with happy combinations 16, 25, and 34. If white will be able to move the blot on his 9-point, there is still the hitting loose chance for white's 21-point blot. All those blots give a good gammon potential if one of them is hit.

As a matter of fact black is not harmed much if he is hit on his 9-point, also not if both black checkers are thrown back in white's inferior 3-point board, because 2 checkers give more flexibility to move out of white's board than one checker, and there is a good chance that white will get into timing problems. However, black will not be glad when white is going to build his prime, but even in this case the gammon danger is less than white's.

Quote

12-29-2014 , 06:34 AM

#43

7NTXX

newbie

Join Date: Jan 2010 Posts: 49

Quote:

Originally Posted by yogiman

Right, with an equity difference of 0.9

Here is another more clear leave of the 5 point which I just met in a game:

White - Pips 60. Match Score 0/30

Black - Pips 122. Match Score 2/30
Black to Play 6-5
Created with www.BGdiagram.com

I see 24/13 as the right play by 0.075 in Xgammon. Am I missing something?

Quote

12-29-2014 , 07:18 AM

#44

yogiman

adept

Join Date: Jun 2012 Posts: 784

Code:

    1.24/18*/13                    Eq.:  +0,632
       0,623 0,180 0,004 - 0,377 0,105 0,002 CL  +0,323 CF  +0,632
      
    2.24/18* 20/15                 Eq.:  +0,573 ( -0,059)
       0,631 0,178 0,004 - 0,369 0,121 0,004 CL  +0,317 CF  +0,573

        Truncated cubeful rollout (depth 10) with var.redn.
        144 games, Mersenne Twister dice gen. with seed 764017137 and quasi-random dice
        Stop when std.errs. are small enough: ratio 0,1 (min. 144 games)
        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]

Correct, apparently I forgot to do a rollout, and 0-ply in Gnu turns out to be highly unreliable. It is inferior clearly because of lost gammons. If white is 2-away or 1-away, than 20/15 is still best. Though a lot depends on our skill with the doubling cube, which is quite random for most of us in these kind of situations.

Last edited by yogiman; 12-29-2014 at 07:31 AM.

Quote

01-01-2015 , 05:16 AM

#45

yogiman

adept

Join Date: Jun 2012 Posts: 784

White - Pips 121

Black - Pips 113
Black to Play 4-1
Created with www.BGdiagram.com

Quote

01-02-2015 , 12:00 AM

#46

Taper_Mike

adept

Join Date: Sep 2009 Posts: 1,030

Position 11 GnuBgID: bJ2DAwTbOJTBAA:cAkGAAAAAAAE

Nice position. This was not obvious to me.

Mike

Quote

01-02-2015 , 04:18 AM

#47

yogiman

adept

Join Date: Jun 2012 Posts: 784

I suppose you can enter this ID to get the board position in GNU. In the right down corner I see a small box, however I can't fill it in.

The intention is that other minds here come with an analysis. But as this doesn't happen I will give it a try.

Code:

    1. 13/9 10/9                    Eq.:  -0,025 
       0,495 0,158 0,002 - 0,505 0,144 0,006 CL  -0,001 CF  -0,025 
      
    2. 13/9 11/10                   Eq.:  -0,108 ( -0,082) 
       0,471 0,165 0,002 - 0,529 0,177 0,006 CL  -0,075 CF  -0,108 
    
    3. 11/6                         Eq.:  -0,129 ( -0,103) 
       0,460 0,126 0,001 - 0,540 0,147 0,005 CL  -0,105 CF  -0,129 

    4. 10/5                         Eq.:  -0,140 ( -0,115)
       0,461 0,142 0,002 - 0,539 0,176 0,006 CL  -0,117 CF  -0,140
     
        Truncated cubeful rollout (depth 10) with var.redn. 
        145 games, Mersenne Twister dice gen. with seed 763368501 and quasi-random dice 
        Stop when std.errs. are small enough: ratio 0,1 (min. 144 games) 
        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]

Let's review the different possibilities:
10/5 is the move with a lot of QF, because I seem to be fond of slotting in the homeboard. However, black is putting all energy into making the 5-point, but doesn't point on the white blot. Instead black will become a target himself when he likely will have to move from his 13-point next turn.

11/6 is the only move that leaves one blot, but if white hits, he has made a big step out of the danger zone while black is struggling from the bar and the same holds true for 13/9 11/10. We should also consider that black has a 3-point board, and that a 4-point board might make this the best move.

13/9 10/9 points on the white blot. If he throws a hitting 1 (some duplication) white will still be captivated within black's range.

Last edited by yogiman; 01-02-2015 at 04:26 AM.

Quote

01-02-2015 , 04:34 AM

#48

yogiman

adept

Join Date: Jun 2012 Posts: 784

White - Pips 139

Black - Pips 109
Black to Play 6-5
Created with www.BGdiagram.com

Quote

01-02-2015 , 06:19 AM

#49

Taper_Mike

adept

Join Date: Sep 2009 Posts: 1,030

Problem 12
There is a danger that Black could get trapped behind White’s block. He needs a 5 or 41 to escape, and since he has a 5 now, perhaps he should run. To my eye, 23/12 seems the best way to do that. Black should have an easy time picking up the third blot.

Another danger is that Black might not be able to hit an entering White blot. 23/18, 10/4 is an alternative that both escapes and brings another builder within direct range of the open 3pt.

Mike

Quote

01-02-2015 , 06:27 AM

#50

Taper_Mike

adept

Join Date: Sep 2009 Posts: 1,030

White - Pips 121

Black - Pips 113
Problem 11. Black to Play 4-1
XGID=-BBBa-C---AAcB---cBcabb---:0:0:1:41:0:0:0:0:10

Sorry, but my wrong-headed play was to make the 10pt, and then kill a checker by moving 6/2. It rolls out as a 3rd-best whopper in my GnuBg trunc. A full XG rollout yields the same result. In retrospect it is obvious that Black will need all his checkers working together if he is going to bring them home safely. Otherwise, his only game plan is to roll a couple of doublets.

Somehow, I overlooked making the 9pt, but notice how it duplicates the ace White needs to cover the blot on his 5pt. In addition, if Black is hit, he will usually have plenty of return shots, often on both sides of the board. If missed, his position will be very flexible, so improvement should be possible.

You can copy and paste a GnuBgID using commands on the Edit menu:

Edit > Copy ID to clipboard > GnuBgID
Edit > Paste ID

In Windows, you can use the usual shortcut keys, Ctrl+C and Ctrl+V. I am not sure whether those work under Linix, but you should find the shortcut keys listed on the Edit menu.

360-trial GnuBg Truncated Rollout
No Jacoby, No Beavers. This trunc uses the same settings as an XGR++ eval.

Code:

    1. Rollout          13/9 10/9                    Eq.:  -0.018424
       0.494188 0.159354 0.002314 - 0.505812 0.145013 0.005859 CL  -0.000828 CF  -0.018424
      [0.000871 0.002320 0.000124 - 0.000871 0.003052 0.000304 CL   0.002990 CF   0.004021]

    2. Rollout          13/9 11/10                   Eq.:  -0.103539 ( -0.085116)
       0.469528 0.161869 0.002057 - 0.530472 0.173520 0.006223 CL  -0.076760 CF  -0.103539
      [0.000806 0.001532 0.000089 - 0.000806 0.001570 0.000222 CL   0.002499 CF   0.004292]

    3. Rollout          11/10 6/2                    Eq.:  -0.128662 ( -0.110239)
       0.470393 0.094227 0.000814 - 0.529607 0.160537 0.004627 CL  -0.129338 CF  -0.128662
      [0.000941 0.001332 0.000049 - 0.000941 0.001615 0.000189 CL   0.002688 CF   0.004319]

    4. Rollout          11/6                         Eq.:  -0.141015 ( -0.122591)
       0.456597 0.127929 0.001185 - 0.543403 0.152073 0.004595 CL  -0.114359 CF  -0.141015
      [0.000838 0.001643 0.000068 - 0.000838 0.001869 0.000165 CL   0.002493 CF   0.003598]

    5. Rollout          10/5                         Eq.:  -0.145776 ( -0.127353)
       0.460878 0.137467 0.001481 - 0.539122 0.173414 0.006325 CL  -0.119034 CF  -0.145776
      [0.000877 0.001508 0.000055 - 0.000877 0.002119 0.000309 CL   0.002632 CF   0.004073]

        Truncated cubeful rollout (depth 5) with var.redn.
        360 games, Mersenne Twister dice gen. with seed 764333941 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]
        Different evaluations after 2 plies:
        Play:  1-ply cubeful prune
        keep the first 0 0-ply moves and up to 8 more moves within equity 0.16
        Cube: 2-ply cubeful prune [world class]

1k 3-ply XG Rollout
No Jacoby, No Beavers

Code:

XGID=-BBBa-C---AAcB---cBcabb---:0:0:1:41:0:0:0:0:10

X:Player 1   O:Player 2
Score is X:0 O:0. Unlimited Game
Pip count  X: 113  O: 121 X-O: 0-0
Cube: 1
X to play 41

    1. Rollout¹    13/9 10/9                    eq:-0.0113
      Player:   49.36% (G:16.13% B:0.19%)
      Opponent: 50.64% (G:14.55% B:0.62%)
      Confidence: ±0.0110 (-0.0223..-0.0003) - [100.0%]
      Duration: 3 minutes 19 seconds

    2. Rollout¹    13/9 11/10                   eq:-0.1055 (-0.0942)
      Player:   46.92% (G:16.00% B:0.18%)
      Opponent: 53.08% (G:17.42% B:0.70%)
      Confidence: ±0.0099 (-0.1154..-0.0956) - [0.0%]
      Duration: 3 minutes 30 seconds

    3. Rollout¹    11/10 6/2                    eq:-0.1131 (-0.1018)
      Player:   47.17% (G:9.73% B:0.08%)
      Opponent: 52.83% (G:16.61% B:0.56%)
      Confidence: ±0.0101 (-0.1232..-0.1030) - [0.0%]
      Duration: 3 minutes 31 seconds

    4. Rollout¹    10/5                         eq:-0.1333 (-0.1220)
      Player:   46.12% (G:13.88% B:0.13%)
      Opponent: 53.88% (G:17.20% B:0.63%)
      Confidence: ±0.0099 (-0.1433..-0.1234) - [0.0%]
      Duration: 3 minutes 09 seconds

    5. Rollout¹    11/6                         eq:-0.1435 (-0.1322)
      Player:   45.64% (G:12.82% B:0.12%)
      Opponent: 54.36% (G:15.09% B:0.58%)
      Confidence: ±0.0093 (-0.1528..-0.1342) - [0.0%]
      Duration: 3 minutes 13 seconds

¹  1296 Games rolled with Variance Reduction.
   Dice Seed: 4594854
   Moves: 3-ply, cube decisions: XG Roller

eXtreme Gammon Version: 2.10

Mike

Last edited by Taper_Mike; 01-02-2015 at 06:33 AM.

Quote

Page 2 of 13

First

1 2 3 4 5 6 7 12

Last

Post Reply Subscribe

...

Page 2 of 13

First

1 2 3 4 5 6 7 12

Last