Position # 54

[PersianLord]

Position Information:

Pip-counts: Red 103-37 White
Score:       Red     1-2  White
Match:                3-pointer

Red to play 3-3

[PersianLord]

Spoiler

Position ID: 3QAAILZtgyACAA , Match ID: QYltACAACAAA

This position actully occured in my match against NIHI. I was playing Red and trying to pick up NIHI's second man at his ace point, I hit loose with 17/11-7/1*. Being extremely unlucky, I lost the game and match and wondered whether my breaking the full-prime was actually right or not. (We had a brief discussion on the right move) To my surprise my move is right according to GNUBG:

   1. Cubeful 2-ply    17/11 7/1*                   Eq.:  +0.254
        62.7%   0.0%   0.0% -  37.3%   6.8%   0.9%       
    2. Cubeful 2-ply    17/14 12/9 7/1*              Eq.:  +0.239 ( -0.016)
        61.9%   0.0%   0.0% -  38.1%   7.3%   0.8%
     
    3. Cubeful 2-ply    20/14 7/1*                   Eq.:  +0.235 ( -0.020)
        61.7%   0.0%   0.0% -  38.3%   7.3%   0.8%
       
    4. Cubeful 2-ply    12/6 7/1*                    Eq.:  +0.231 ( -0.024)
        61.5%   0.0%   0.0% -  38.5%   8.5%   0.8%
       
    5. Cubeful 2-ply    17/14 7/1* 7/4               Eq.:  +0.206 ( -0.049)
        60.3%   0.0%   0.0% -  39.7%   8.4%   0.5%
       
Well, even though GNUBG agrees to going for hunting the 2nd blot, I think playing safe with 7/1*(2) is not as bad as GNUBG's analysis indicates:

    7. Cubeful 2-ply    7/1*(2)                      Eq.:  +0.173 ( -0.082)
        58.6%   0.0%   0.0% -  41.4%   1.4%   0.0% 
   
NIHI and I agreed that with his 8 men off, my chances to win the game with 1 of his men closed out was about 50%. But according to the bot, I was a ~59% favorite to win. Had I known this, I would have played safe. BTW, I wonder how does GNUBG think that there is a 1.4% chance for Red to getting gammoned after closing out right now?!!

[close]

[playBunny]

Spoiler

BTW, I wonder how does GNUBG think that there is a 1.4% chance for Red to getting gammoned after closing out right now?!!

Try rolling 6-6 twice and you'll discover a blot on your 6-point.

[close]

[blitzxz]

Yabba dabba, I guessed this one right but I can't understand these positions. Sometimes it's right sometimes it's not. It depends from the opponent's structure. Is there a blot, how many checker out, is it crushed and how badly and so on. The position of the cube can also make big difference. But I can't make good general rules for these positions.

And BTW after close out, gammon chances are, of course, zero. It's not possible to leave a blot without bearing out any checkers. Gnu can look ahead only 2-ply and can't see the inevitable result. Gnu tends to make these kind of errors in closed out/late-shot positions which can lead to bad cube handling.

[Tomawaky]

Spoiler

I played 17/11-12/6 and leave no joker (6-1) to my opponent. I think we have time to hit with more control.

[close]

[stog]

but perhaps by then white can cover their 1......i think red needs 2 back to maximise winning potential ..

good problems thx all

[lewscannon]

but perhaps by then white can cover their 1......i think red needs 2 back to maximise winning potential ..

good problems thx all

White can maybe cover the 1, but just as easily leave the 3 or eventually 2 open. I'd bring my back men forward and definitely not leave the 6-1 heartbreak roll to chance

[socksey]

White can maybe cover the 1, but just as easily leave the 3 or eventually 2 open. I'd bring my back men forward and definitely not leave the 6-1 heartbreak roll to chance

Me too!     But, what do we know?!   

socksey



"Everyone who changes is often a traitor in the eyes of those who can never change." - Amoz Oz

[dorbel]

Haven't we all come back to the chouette to find that the captain broke the prime and is being slated because it has all gone horribly wrong? However, it is often correct and it is clearly right here, even though Snowie 3-ply plays 7/1*(2).  Bots tend to over rate the winning chances of the side with the prime in these "after the hit" positions and both Gnu and Snowie do that here. Gnu's 62% is far too high and the Snowie rollout suggests that Red wins around 57 % if he tries to pick up the second checker, but just under 54% if he closes the board now. Not hitting at all is clearly inferior to either of these choices, as you can't just leave White alone to tidy up her blot inboard. Besides, if you don't hit now, in some variations you don't close the board at all!
It doesn't seem to matter much which hit-loose play you choose, but the rollout marginally selects 20/17, 12/9, 7/1*. I can't say why that should be better than anything else and making the 17pt doesn't seem to have much merit, apart from blocking 5-5 after a White 6-1, bar/24*/18*.
If the bot creators could add a propbot to fibs, what a wonderful tool that would be for settling arguments! inim?

[NIHILIST]

This is a good instructional position for several reasons. One not discussed is to what extent the NUMBER OF CHECKERS OFF  white has. It's good discipline to know the winning probabilities based on how many opponent has off after a blot is hit.

Here, I'll just discuss general rules of thumb based on crashed positions ( remaining checkers on the 1-3 points ) that will guide you to making the correct play. Thus, the most significant aspect here is 8 checkers off for white. 8 off in a crushed position is a near 50-50 proposition; if you played perfectly, closed out 1 blot, it's a near coin flip. Anyone who's played longer than a few weeks knows his winning chances improve if he can pick up the second blot, so in this case the hit is best...........you hope opponent rolls a 1, hitting you, and you go back and hit the second blot. An interesting side question would be WHAT'S YOUR PLAY IF THERE WERE NO BLOT IN WHITE'S BOARD ? Do you close the board or work the position a few rolls hoping opponent exposes a second blot for you to go after ? That's a question for another day.

Now let's assume opponent has only 7 checkers off with the other checker on the 2 point and the blot still exposed. While the position wouldn't seem to have changed much, it has; 7 off reduces his winning chances to considerably less than 50% and makes closing the board clearly best.

So, now that you know this, you know the correct play depending on number of checkers off.............chase the second blot with 8 or more off, close the board with 7 or less off.


Bob

[PersianLord]

Well, NIHI provided his 'rule of thumb', now it's my turn. PL's rule of thumb: It's always good to know NIHI's verdict. The '8-men off rule' is just one of the many important nuances I have learned from this veteran BG player and leviathan of FIBS.

But, another question regarding this especial position is this: 'at what expense Red should go for the other blot, knowing his chances is about 50% with one man trapped?'. In other words, was Red/PL right in hitting loose White/NIHI's blot WHICH LED TO THE BREAKING OF HIS FULL-PRIME? This is the crucial question.

Regards

[PersianLord]

Haven't we all come back to the chouette to find that the captain broke the prime and is being slated because it has all gone horribly wrong? However, it is often correct and it is clearly right here, even though Snowie 3-ply plays 7/1*(2).  Bots tend to over rate the winning chances of the side with the prime in these "after the hit" positions and both Gnu and Snowie do that here. Gnu's 62% is far too high and the Snowie rollout suggests that Red wins around 57 % if he tries to pick up the second checker, but just under 54% if he closes the board now. Not hitting at all is clearly inferior to either of these choices, as you can't just leave White alone to tidy up her blot inboard. Besides, if you don't hit now, in some variations you don't close the board at all!
It doesn't seem to matter much which hit-loose play you choose, but the rollout marginally selects 20/17, 12/9, 7/1*. I can't say why that should be better than anything else and making the 17pt doesn't seem to have much merit, apart from blocking 5-5 after a White 6-1, bar/24*/18*.
If the bot creators could add a propbot to fibs, what a wonderful tool that would be for settling arguments! inim?

I think 20/17-12/9-7/1* is clearly inferior to 17/11-7/1*. With 17/11-7/1*, Red will have 2 builders in his 11 and 12 pts to re-build his full-prime and restart the hunting spree with safety and joy, but with 20/17-12/9-7/1* he will have just one builder in his 9 pt, which is cleary less constructive to repair the prime. The reason behind 20/17-12/9-7/1*, i.e. blocking 5-5 after a White 6-1, is of least importance IMO. Just think about it and calculate how likely is this weird scenario to take place: White should roll a 6-1 roll first (2/36), then Red will have to miss hitting the blot in White's ace point (25/36) and then White should roll a 55 (1/36). Personally I won't exchange maximizing my chances to rebuild my broken prime for preventing such a miraculous story to happen. I think this is yet another position in which Snowie's rollout result is even more inferior than it's actual first hand analysis.

Regards

[PersianLord]

Well, I just learned how to set up positions. I just wanted to know what's the probabilities in positions wherein  my opponent has borne off 7,8 and 9 men and I have managed to close out one man and have NO chances of picking up more blots. In all of these case it has been assumed that my opponent's remaining men are all stacked in his ace point and that I have reached an ideal position for bear-off (3 men on 6,5 and 4 pts and 2 men on the low pts)

1- With 7 checkers off: (1st pic)

        Eq.:  +0.267
        59.5%   0.0%   0.0% -  40.5%   0.0%   0.0%
       
2- With 8 checkers off: (2nd pic)

        Eq.:  +0.015
        50.5%   0.0%   0.0% -  49.5%   0.0%   0.0%

3- With 9 checkers off: (3rd pic)

        Eq.:  -0.235
        41.6%   0.0%   0.0% -  58.4%   0.0%   0.0%
       

[dorbel]

Yes it is clear that 7/1* is essential and the other two threes should be played somewhere in the outfield. As I said too, as a sideline, it doesn't matter much where you play the other two threes. I only highlighted Snowie's "best in rollout" as a curiosity. 20/14 and 17/11 are very close, well within a margin of error, but it is unlikely that either will be clearly superior with a longer rollout. Of course the blocking of 5-5 next turn after 20/17, 12/9 isn't very important, but the fact that White won't have two more blots to shoot at after her joker 6-1 may be.

[blitzxz]

WHAT'S YOUR PLAY IF THERE WERE NO BLOT IN WHITE'S BOARD ?

Yes, this is the interesting question and I would like to see the roll-outs for it. My guess is that you try work (keep the prime and no hit) the other checker if there is possibility of leaving it with only one roll and own checkers far enough and yet again enough checkers bear-off. But what is exactly enough? And the other big question is what about if white has no checkers bear-off at all. In some cases it's correct to try to work the other checker because gammon chances but this depends of the structure and I would like to know good rules of thumb to that too. Thanks for 8-off rule.

[dorbel]

If you close White's board by moving the spare to the ace point, then hitting at all is wrong. You fiddle around for a turn or two hoping that White will roll a 1 or 2 and expose another blot. If that happens, then you hit loose on the ace to try to recirculate and pick up a second checker. It doesn't work that often, but it's worth a try as it is relatively risk free, even if you have to break the prime to hit as we have seen here.

If White has no checkers off then the immediate closeout is the goal. At some scores it might be worthwhile to try and pick up another checker for gammons, but not here of course.

[blitzxz]

If White has no checkers off then the immediate closeout is the goal. At some scores it might be worthwhile to try and pick up another checker for gammons, but not here of course.

Yes, of course, in this score the gammons value nothing... it's double match point. But in every score were gammons value it might be an option. But it's not only score related. For example, if white has checkers outside home field it's usually correct to close out immediatly and not give time to move them to closer. But in a different set up it might be worth to try to get second checker. Quite often I find myself in these position with no idea what is correct and end up closing the board immediatly because it's just simpler.

Another close out problem is that when white has crashed board and one point anchor. In these you want to give white a chance to break anchor but what gnu gives usually the best is leaving blot or even two blots(!) (breaking two points in the prime) as best option. These fancy plays are not my choices even if know that gnu thinks them best. But now I'm off topic and will give example position in different topic later.

[Zorba]

GNUBG 2-ply rollout results:

Spoiler

    1. Rollout          17/11 7/1*                   Eq.:  +0,2044
        60,22   0,00   0,00 -  39,78   8,34   2,30 CL  +0,2044
      [  0,44   0,00   0,00 -   0,44   0,67   0,59 CL   0,0088]
        Full cubeless rollout (trunc. at one-sided bearoff) with var.redn.
        216 games, Mersenne Twister dice gen. with seed 815107159 and quasi-random dice
        Play:  2-ply cubeless prune
        keep the first 0 0-ply moves and up to 10 more moves within equity 0,06
        Skip pruning for 1-ply moves.
        Cube: 2-ply cubeful prune
    2. Rollout          20/17 12/9 7/1*              Eq.:  +0,1913 ( -0,0131)
        59,56   0,00   0,00 -  40,44   7,41   1,64 CL  +0,1913
      [  0,42   0,00   0,00 -   0,42   0,32   0,39 CL   0,0084]
        Full cubeless rollout (trunc. at one-sided bearoff) with var.redn.
        216 games, Mersenne Twister dice gen. with seed 815107159 and quasi-random dice
        Play:  2-ply cubeless prune
        keep the first 0 0-ply moves and up to 10 more moves within equity 0,06
        Skip pruning for 1-ply moves.
        Cube: 2-ply cubeful prune
    3. Rollout          20/14 7/1*                   Eq.:  +0,1876 ( -0,0168)
        59,38   0,00   0,00 -  40,62   7,87   2,30 CL  +0,1876
      [  0,43   0,00   0,00 -   0,43   0,40   0,63 CL   0,0086]
        Full cubeless rollout (trunc. at one-sided bearoff) with var.redn.
        216 games, Mersenne Twister dice gen. with seed 815107159 and quasi-random dice
        Play:  2-ply cubeless prune
        keep the first 0 0-ply moves and up to 10 more moves within equity 0,06
        Skip pruning for 1-ply moves.
        Cube: 2-ply cubeful prune
    4. Rollout          12/6 7/1*                    Eq.:  +0,1862 ( -0,0182)
        59,31   0,00   0,00 -  40,69   9,71   2,63 CL  +0,1862
      [  0,46   0,00   0,00 -   0,46   0,34   0,65 CL   0,0092]
        Full cubeless rollout (trunc. at one-sided bearoff) with var.redn.
        216 games, Mersenne Twister dice gen. with seed 815107159 and quasi-random dice
        Play:  2-ply cubeless prune
        keep the first 0 0-ply moves and up to 10 more moves within equity 0,06
        Skip pruning for 1-ply moves.
        Cube: 2-ply cubeful prune
    5. Rollout          17/14 12/9 7/1*              Eq.:  +0,1850 ( -0,0194)
        59,25   0,00   0,00 -  40,75   8,10   2,88 CL  +0,1850
      [  0,42   0,00   0,00 -   0,42   0,54   0,77 CL   0,0085]
        Full cubeless rollout (trunc. at one-sided bearoff) with var.redn.
        216 games, Mersenne Twister dice gen. with seed 815107159 and quasi-random dice
        Play:  2-ply cubeless prune
        keep the first 0 0-ply moves and up to 10 more moves within equity 0,06
        Skip pruning for 1-ply moves.
        Cube: 2-ply cubeful prune
    6. Rollout          20/17 7/1* 7/4               Eq.:  +0,1711 ( -0,0333)
        58,56   0,00   0,00 -  41,44  11,30   1,19 CL  +0,1711
      [  0,47   0,00   0,00 -   0,47   0,70   0,38 CL   0,0093]
        Full cubeless rollout (trunc. at one-sided bearoff) with var.redn.
        216 games, Mersenne Twister dice gen. with seed 815107159 and quasi-random dice
        Play:  2-ply cubeless prune
        keep the first 0 0-ply moves and up to 10 more moves within equity 0,06
        Skip pruning for 1-ply moves.
        Cube: 2-ply cubeful prune
    7. Rollout          17/14 7/1* 7/4               Eq.:  +0,1672 ( -0,0372)
        58,36   0,00   0,00 -  41,64  11,44   2,20 CL  +0,1672
      [  0,47   0,00   0,00 -   0,47   0,74   0,66 CL   0,0093]
        Full cubeless rollout (trunc. at one-sided bearoff) with var.redn.
        216 games, Mersenne Twister dice gen. with seed 815107159 and quasi-random dice
        Play:  2-ply cubeless prune
        keep the first 0 0-ply moves and up to 10 more moves within equity 0,06
        Skip pruning for 1-ply moves.
        Cube: 2-ply cubeful prune
    8. Rollout          12/9 7/1* 7/4                Eq.:  +0,1618 ( -0,0426)
        58,09   0,00   0,00 -  41,91  11,47   2,72 CL  +0,1618
      [  0,47   0,00   0,00 -   0,47   0,81   0,64 CL   0,0093]
        Full cubeless rollout (trunc. at one-sided bearoff) with var.redn.
        216 games, Mersenne Twister dice gen. with seed 815107159 and quasi-random dice
        Play:  2-ply cubeless prune
        keep the first 0 0-ply moves and up to 10 more moves within equity 0,06
        Skip pruning for 1-ply moves.
        Cube: 2-ply cubeful prune
    9. Rollout          17/11 12/6                   Eq.:  +0,1304 ( -0,0740)
        56,52   0,00   0,00 -  43,48   5,10   0,48 CL  +0,1304
      [  0,38   0,00   0,00 -   0,38   0,55   0,18 CL   0,0076]
        Full cubeless rollout (trunc. at one-sided bearoff) with var.redn.
        216 games, Mersenne Twister dice gen. with seed 815107159 and quasi-random dice
        Play:  2-ply cubeless prune
        keep the first 0 0-ply moves and up to 10 more moves within equity 0,06
        Skip pruning for 1-ply moves.
        Cube: 2-ply cubeful prune
   10. Rollout          7/1*(2)                      Eq.:  +0,0808 ( -0,1236)
        54,04   0,00   0,00 -  45,96   0,00   0,00 CL  +0,0808
      [  0,24   0,00   0,00 -   0,24   0,04   0,00 CL   0,0047]
        Full cubeless rollout (trunc. at one-sided bearoff) with var.redn.
        216 games, Mersenne Twister dice gen. with seed 815107159 and quasi-random dice
        Play:  2-ply cubeless prune
        keep the first 0 0-ply moves and up to 10 more moves within equity 0,06
        Skip pruning for 1-ply moves.
        Cube: 2-ply cubeful prune

[close]