Open Side Menu Go to the Top
Register
obvious is not always right obvious is not always right

01-23-2015 , 07:31 AM
Code:
    1. 13/11 6/5                    Eq.:  -0,172 
       0,456 0,096 0,004 - 0,544 0,185 0,004 CL  -0,177 CF  -0,172 
    
    2. 7/5 6/5                      Eq.:  -0,321 ( -0,149) 
       0,414 0,097 0,004 - 0,586 0,149 0,004 CL  -0,223 CF  -0,321 

    3. 6/5 6/4                      Eq.:  -0,326 (-0,154)
       0,437 0,091 0,004 - 0,563 0,287 0,006 CL  -0,324 CF  -0,326
      
        Truncated cubeful rollout (depth 10) with var.redn.
        147 games, Mersenne Twister dice gen. with seed 769866802 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]
This problem looks very much like problem 23 . The difference here is that white has a good board. Less extreme looking as in problem 19 this is a high volatility play. Probably because of the reason that the merits and demerits of safe vs risky play have to be weighed in a straight way in these situations, which is very difficult to quantify for us humans, the best we can do is to listen to the bot, and use it as a reference position. Bot experimentation reveals that this is also the right move in case white's position is optimal:


Code:
    1. 13/11 6/5                    Eq.:  -0,363
       0,414 0,064 0,002 - 0,586 0,237 0,003 CL  -0,346 CF  -0,363
     
    2. 13/11 7/6                    Eq.:  -0,466 ( -0,103)
       0,366 0,058 0,002 - 0,634 0,130 0,002 CL  -0,340 CF  -0,466
obvious is not always right Quote
01-23-2015 , 03:23 PM
Hi yogiman -- I appreciate you posting positions and the rollouts, but I'd prefer it if you would roll out more plays. Like in the position in post #124, a lot of people would play 6/3* and 13/11 7/6. 6/3* was actually the first play that occurred to me. It would be nice to see how wrong those are, even if they are 4th and 5th best or something.

You can avoid rolling out plays that are dominated by other plays -- if there are two similar plays and one is clearly better than the other, rolling out the worse one is a waste of time, but you should try to rollout at least one play for each reasonable, but wrong concept. Like hitting and playing safe in post #124.

Thanks.
obvious is not always right Quote
01-23-2015 , 03:47 PM
Because of your mentioning I did a separate rollout for 6/5 6/4, which I added to my original rollout. You gave the right argument for not playing 6/3*, and didn't see a reason to pay further attention to it. But I will do my best in the next analysis.
obvious is not always right Quote
01-23-2015 , 04:07 PM
Position ID: ZLONKQC4DQDjTA Match ID: UgkOAAAACAAE



Black to Play 4-3
obvious is not always right Quote
01-23-2015 , 04:24 PM
B/22* 18/14.
Hitting, provocation, recirculation.
Let's go to war, we have the better inner board!
obvious is not always right Quote
01-23-2015 , 07:32 PM
After Black comes out to the 18pt, 6 is blocked, so he won't have to play those unless he can advance from the 18pt first. 3, 4, and 5 all play okay. 2 and 1 might be cracking numbers, although 21 itself plays inside.

His racing deficit is 68 pips, acceptable, but perhaps a tad low for a 4-2 backgame.

Should he hit? If he does hit, should it be: 1. as part of a plan to go forward, 2. a two-way plan, or 3. a plan to force White to hit him again?
  1. Going forward does not seem likely to succeed. Black will have trouble containing a single hit checker, so he would need to be successful in attacking White's other blots for the plan to work. Then he would need to roll the right numbers to extricate his rear checkers. The most efficient—but not the best—way to attack would be for Black to play bar/22* 18/14, moving within direct range of White's blots. If White then enters poorly or dances, Black might end up hitting a second checker. The downside is that he risks losing the 18pt for good. If that happens, all his chances, both forward and backward will be greatly diminished. So that means that bar/22*/18 is the best play for going forward.

  2. Two-way plans usually involve attacking plays that hit loose. If the opponent misses the return shot, and then you cover, you can switch to a forward game. Otherwise, getting hit won't hurt so much. It just commits you to the backgame. In this position, the attack is missing. Hitting on the 22pt, and then not being hit in return, is not going to switch Black's posture to a forward one. That said, bar/22*/18 is the best two-way play.

  3. It seems likely that Black will have to break his board if he wants to force White to hit him again. What about bar/22* 6/2? That would be fine if White rolls 52 or 42, but if he rolls a 6, Black may regret losing his 6pt and placing a semi-dead checker on the 2pt. Once again, it looks like bar/22*/18 is the best play if your plan is to hit in order to force an exchange of hits. But why bother? It looks like bar/18 (no hit) is almost just as good for forcing an exchange of hits downstream. Hitting would give White an additional 22 pips of timing, and even more when he dances.
All in all, it seems like Black loses more than he gains by hitting. I would play bar/18 (no hit), and await future developments. It may still be possible to force White to hit, but if not, Black may already have enough timing to win his backgame.

Mike

Last edited by Taper_Mike; 01-23-2015 at 07:41 PM.
obvious is not always right Quote
01-23-2015 , 09:42 PM
Bar/21/18 looks completely clear. White has 5 points out of 6 and Black needs to see that position break before creating contact.
obvious is not always right Quote
01-24-2015 , 06:21 AM
It is imaginable that the expert players see that not-hitting is the obvious move, and maybe a few will get a little doubt as sort of reversed psychology whether hitting could be right, and for the intermediate players (for whom it was meant) vice versa. I think the zenmaster would be pleased with me.

I have not included bar/22* 6/2 in the rollout, because we don't expect a reliable outcome. I have tried it, and it can work successfully. However, the thing is not so much that the checker disharmonizes the board buildup, as black has enough time to set things right and can possibly recirculate. However, many times black will look up against a prime and will have to break up his board and possible 3-point board because of bad doubles. But it is a thing to consider when the opponent is 1-away.

Code:
    1. bar/18                       Eq.:  -0,387 
       0,354 0,053 0,002 - 0,646 0,258 0,013 CL  -0,507 CF  -0,387 
     
    2. bar/22*/18                   Eq.:  -0,476 ( -0,089) 
       0,307 0,059 0,002 - 0,693 0,247 0,010 CL  -0,581 CF  -0,476 
       
        Truncated cubeful rollout (depth 10) with var.redn. 
        147 games, Mersenne Twister dice gen. with seed 769866802 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]
obvious is not always right Quote
01-24-2015 , 06:33 AM
Position ID: w+9NAABtIwFgGw Match ID: UgkHAAAACAAE




Black to Play 6-1

PS Could Taper_Mike or someone else include the code for XG on top of his reply?

Last edited by yogiman; 01-24-2015 at 06:42 AM.
obvious is not always right Quote
01-24-2015 , 07:14 AM
Quote:
Originally Posted by yogiman
Could Taper_Mike or someone else include the code for XG on top of his reply?
XGID=-ABBB--A-A------a-bdfBBBb-:2:1:1:61:0:0:0:0:10


Ask and ye shall receive. Actually, since both XG and GnuBg accept IDs from either one, you don't really need both IDs.

Mike
obvious is not always right Quote
01-24-2015 , 09:06 AM
[Afraid to venture a guess...]
obvious is not always right Quote
01-24-2015 , 05:03 PM
9/8 7/1 leave the back checkers alone and build a board
obvious is not always right Quote
01-24-2015 , 09:40 PM
White has some embarrassing 6s here, specifically 61, 63, 64, 66, and, especially, 65. White won’t like 32, 42, and 55 very much either.

Blue should make a four-point board now, with a plan to hit the next shot that comes up.

Mike
obvious is not always right Quote
01-25-2015 , 07:21 AM
Quote:
Originally Posted by uberkuber
[Afraid to venture a guess...]
LOL!

What I always tell myself is that it is important to have a reason for the plays I select. Even when wrong, I can learn something by positing an explanation. At the least, of course, I may learn that my explanation is wrong!

The other thing I constantly remind myself is that the hard plays are the reason I am attracted to backgammon. The game would not hold my interest for very long if all I had to do was play the 31s and 42s I love so much on the opening roll.

So, please, do keep posting, and do let us know your reasoning. The worst thing to do is to push checkers around on autopilot.

Mike
obvious is not always right Quote
01-25-2015 , 01:06 PM
As Taper_Mike already suggested, white is likely to leave a blot next turn, and even when he has got all his checkers in his board, blue probably is not out of hitting chances. Is it worth to give up the 3-point backgame with a board that looks deplorable? Covering the checker on the 1-point spoils the opportunity to get another checker, so he would like to recycle it. However, with a missing 5 and 6 point blue has no strong case, and on top of that you give white an opportunity to make his 3-point. Creating a 4-point board is much more important.
Code:
    1. 9/8 7/1                      Eq.:  -0,357 
       0,396 0,032 0,001 - 0,604 0,305 0,016 CL  -0,496 CF  -0,357 
    
    2. 21/15 7/6                    Eq.:  -0,395 ( -0,038) 
       0,412 0,029 0,001 - 0,588 0,367 0,027 CL  -0,540 CF  -0,395 
    
    3. 22/16*/15                    Eq.:  -0,520 ( -0,162) 
       0,363 0,044 0,001 - 0,637 0,385 0,034 CL  -0,647 CF  -0,520

Also if he would not be able to cover, he should not hit:


Code:
    1. 21/14                        Eq.:  -0,423
       0,407 0,026 0,001 - 0,593 0,379 0,027 CL  -0,565 CF  -0,423
      
    2. 22/16* 10/9                  Eq.:  -0,456 ( -0,033)
       0,374 0,050 0,001 - 0,626 0,359 0,023 CL  -0,584 CF  -0,456
Truncated cubeful rollout (depth 10) with var.redn.
145 games, Mersenne Twister dice gen. with seed 769929098 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]

In the thread when to hit in a 3-point backgame I will show in which cases hitting is right.

Last edited by yogiman; 01-25-2015 at 01:15 PM.
obvious is not always right Quote
01-25-2015 , 10:48 PM
Black to Play 6-5
XGID=--BBBBCA-A----b-ccbd-Ba---:0:0:1:65:0:0:0:0:10
White - Pips 114

Black - Pips 104

The usual: what would you play, and why?

Mike
obvious is not always right Quote
01-26-2015 , 01:14 AM
White won't leave a shot next roll and we might crush our board if we make the ace point now. If White hits loose, he'll face the risk of being on the bar against a 5-pt board.

We'll be ahead by 21 pips after the roll, so I would extract a back man now.

Of course, White can hit and cover with 13 rolls (55,44,33,22,11,51,41,31,21), but the cube still isn't turned, so we're not in gammon territory yet. (White might opt to complete the 6-prime with 11 instead of hitting.)

21/10
obvious is not always right Quote
01-26-2015 , 05:03 AM
I agree with uber, which unfortunately doesn't bode well for his choice. We have to run while the opportunity presents itself or risk collapsing our board.
obvious is not always right Quote
01-26-2015 , 10:18 AM
Black will be ahead 21 pips after the roll. Right now he has the best possible 5-point board, while White has a 1-point board with a blot. Can Black really think he'll have a better opportunity to run in the future? He has to run now.

Having said that, note that he has a choice. I'd play the simple 21/10, but 21/15 9/4 has a point to it, stopping White from making his 5-point with 61, 63, and 64. Might be worth it, but over the board I'd run all the way.
obvious is not always right Quote
01-26-2015 , 04:47 PM
id run as well
obvious is not always right Quote
01-27-2015 , 10:05 AM
So you are not that fat.
obvious is not always right Quote
01-27-2015 , 07:45 PM
Black to Play 6-5
XGID=--BBBBCA-A----b-ccbd-Ba---:0:0:1:65:0:0:0:0:10
White - Pips 114

Black - Pips 104

To a beginner, it may be obvious to make the ace point. To a more experienced player, it may be clear that Black must race, and take this opportunity to run a checker to the safety of his outer board. But neither of these is right. The top three plays all have Black breaking anchor with the 6, but then stopping on the 15pt, where White will have a direct shot at him! How can this be?

Two factors are at work. The first is White’s blot on his 3pt. With a nasty five-point board looming on the other side, White will not be in a mood to hit unless he can cover up at the same time. “So what?” you might say. “White will still hit when it is advantageous to him. Why give him the free shots?”

The answer may found by looking at the second factor, which is White’s rolls of 61, 63, and 64. When the 15pt is empty, White will use those to make his 5pt. When, however, a Black checker is stationed on the 15pt, White will be forced to play safe, moving one or two of his valuable spares behind Black’s straggler. Among non-doublets, White has only 6 other rolls that make his 5pt. Killing White’s 6s cuts the number of non-doublets that make the 5pt in half.

Like the posters who replied to this problem, my OTB play was 21/10. At first, I was mystified by 21/15 9/4, but then I noticed the 6s. Running all the way is only a small error, sacrificing about 22 millipoints of equity.

5k XG2 Rollout – Unlimited Game , No Jacoby, No Beavers
Code:
XGID=--BBBBCA-A----b-ccbd-Ba---:0:0:1:65:0:0:0:0:10

X:Player 1   O:Player 2
Score is X:0 O:0. Unlimited Game
Pip count  X: 104  O: 114 X-O: 0-0
Cube: 1
X to play 65

    1. Rollout¹    21/15 9/4             eq:-0.158
      Player:   45.49% (G:10.40% B:0.12%)
      Opponent: 54.51% (G:4.49% B:0.13%)
      Confidence: ±0.0069 (-0.1651..-0.1514) - [96.6%]

    2. Rollout¹    21/15 6/1             eq:-0.167 (-0.009)
      Player:   45.25% (G:10.92% B:0.14%)
      Opponent: 54.75% (G:6.12% B:0.32%)
      Confidence: ±0.0062 (-0.1732..-0.1608) - [3.2%]

    3. Rollout¹    21/15 7/2             eq:-0.172 (-0.013)
      Player:   45.04% (G:9.82% B:0.11%)
      Opponent: 54.96% (G:4.65% B:0.14%)
      Confidence: ±0.0068 (-0.1784..-0.1649) - [0.2%]

    4. Rollout¹    21/10                 eq:-0.181 (-0.022)
      Player:   44.63% (G:8.59% B:0.11%)
      Opponent: 55.37% (G:2.79% B:0.08%)
      Confidence: ±0.0067 (-0.1872..-0.1739) - [0.0%]

¹  2592 Games rolled with Variance Reduction.
   Dice Seed: 32175630
   Moves: 3-ply, cube decisions: XG Roller

eXtreme Gammon Version: 2.10
Mike

Last edited by Taper_Mike; 01-27-2015 at 08:02 PM.
obvious is not always right Quote
01-28-2015 , 06:36 AM
Position ID: /LwTACBsWxAANw Match ID: UokWAAAACAAE

White - Pips 88

Black - Pips 172
Black to Play 5-5
Created with www.BGdiagram.com
obvious is not always right Quote
01-28-2015 , 10:45 AM
I'll guess that the obvious play is supposed to be 12/7 23/8, making a 5-prime and slotting the back of the prime. The problem here is that White will safety his blot on his 8-point, and then Black will have to release checkers quickly, giving White a chance to perhaps make his 2-point on Black's head, possibly gaining time to release his back checker.

I like 12/7 23/18(3). Now Black has plenty of checkers to move in peace. His goal is to flood the outfield and either slot and cover the 8-point or get a double or triple shot if White gets out.
obvious is not always right Quote
01-28-2015 , 01:43 PM
Why didn't Black double before the roll? Is this even a take? All numbers that hit or make a 5-prime are extremely strong.

12/7 23/18(3) looks good to me. No chance of cracking or being attacked.

A Gnu backgammon rollout may not get the right answer here as it screws up the cube action in containment positions, thinking the trailer is stronger than he is, especially when multiple checkers can be hit.
obvious is not always right Quote

      
m