Cube 105 - Worth the risk?

[sixty_something]

If the board image does not appear in the question above, you may view it with the attached file below.

one vote per user .. results of the poll are shown after you have voted

then, open the text file below for a copy of the GNU analysis or you may enter the ID's below into GNU or Snowie to evaluate your answer:

Position ID: 2jaDAQYTt2MGAA   Match ID: MBmgACAAEAAA


Having once decided to achieve a certain task, achieve it at all costs of tedium and distaste. The gain in self-confidence of having accomplished a tiresome labor is immense.
-- Arnold Bennett

[socksey]

What are the pip counts? 

socksey



"Just living is not enough," said the butterfly, "One must have sunshine, freedom and a little flower." - Hans Christian Anderson

[Tomawaky]

What are the pip counts? 
socksey

hehe socksey  this is an online player question 

I voted for "Doubtful cube, awful pass"

hoping we will see answer for all of your submit one day ?

[sixty_something]

pip counts? according to dorbel in a shout when Zorba asked the same question "real gammoners don't need pipcounts" .. a little later in a Tell he said he got it wrong, messed up on the pip count .. hehehe .. sorry dorbel just too good to not share

When something that honest is said it usually needs a few minutes of silence to dissipate.
-- Pamela Ribon , Why Girls Are Weird, 2003 (i wonder if Pamela needs a pip count?)

[sixty_something]

Tomawaky read the instructions  the answers are in the attached text file - well, the GNUbg analysis is there .. it's up to you to figure out the answer from that or form an answer committee


Committee --a group of men who individually can do nothing but as a group decide that nothing can be done.
-- Fred Allen US radio comedian (1894 - 1956)

[Tomawaky]

Tomawaky read the instructions  the answers are in the attached text file - well, the GNUbg analysis is there .. it's up to you to figure out the answer from that or form an answer committee

Sorry I missed it 

[blitzxz]

Best cube is when you get your opponent to make terrible mistake. So great cube here.

[Tomawaky]

One point 

[Zorba]

This position qualifies as a standard doubling decision. If you want to get better at cube play, this is the kind of position you should really try to understand well. You'll often get similar cube decisions and using a position like this as a reference is a great help.

Position type

So, what's going on here? Red escaped both his backcheckers to the midpoint (and stripped it); White made the advanced anchor on the 21(4) point. This is called a holding game: White holds the anchor in Red's board and can hope to hit something for quite a few moves to come, while White mostly just wants to get home safely and avoid getting hit.

Race

In a situation like that, you really need the pipcounts (please post them, I don't see any advantage to leaving them out). Here, as you might expect, Red leads the race with 109 pips to go AND on roll, versus White's 122 pips.

The race is so important here, because if neither player hits something, Red is far from gin. White has plenty of rolls to pop out a 5-5 or 6-6, or perhaps even just a 6-5 and be a contender in the race, or even clearly ahead.

Based on the pipcounts, Red is 11.9% ahead here. If it was just a race, and a money game (ignoring the matchscore for now), that would indicate a big double, and just a small take. Red's winning chances would probably be around 78%, IF it were just a pips race.

Adjust for wastage (advanced)

Raw pipcounts are not all though, since there's also wastage involved when you get to the bear-off. I won't explain all that here, but the idea is that during the bear-off, you cannot always play all the pips the dice give you, because you don't have checkers on those points anymore. Like rolling a 5, when you only have a checker on your 6pt: you have to move 6/1 which is very inefficient.

In this position, Red's wastage (inefficiency) is worse than White's, mostly because Red already put two checkers on his acepoint this early, and because Red cannot put checkers on the useful 4pt, as long as White's anchored there. White on the other hand, already has useful checkers on his three highest points (6, 5 and 4) and figures to move his others checkers quite freely to the desired spots on these high points.

Using GNUBG's "One-sided rollout" (under race theory), the wastage factor here means that White's pip advantage is actually 3 pips less than the pipcount indicates. This adjustment means Red is about 75.1% to win based on the race alone, which is a clear double and clear take.


Hitting chances and effectiveness

Now comes the tricky bit: it's not just a race! There is still quite a lot of contact left, so either player could hit the other. Now which player is most favoured by the remaining contact? I guess it's clear here that White will get most of the benefit of the remaining contact: he can just sit on the anchor with hardly any risk, and might just get a shot when Red tries to clear the midpoint, 11 point and even later on, perhaps some residual chances to hit when White brings in the checkers from the 8 and 7pts.

The chance that Red can hit White in a favourable way is much, much smaller. One scenario is when White is pretty much forced to break his anchor later, then Red attacks the blot(s).

So, the remaining hitting chances are clearly in White's favour. The next question to ask yourself is (players often forget this!), what happens AFTER such a hit? How deadly will such a hit be? White already has a strong (home)board here, a perfect start: the four highest points and the next one slotted, with two checkers in direct range to cover. Getting hit means serious trouble for Red already, and might be almost lethal in just a few rolls. Red's own board building on the other hand leaves a lot to be desired: the fivepoint is still open and only the acepoint has been made yet.

Attempt to quantify the effect of hits

So, both the quantity AND quality of hits favours White enormously here. It's hard to adjust win percentages directly based on this analysis: you'll need experience, references and analyses to get a "feel" for it. You could try a very rough estimate just to see how it might work out: Let's say Red during his bear-in has to leave a blot in direct shot range 30% of the time and that the direct shot will get hit 30% of the time. That means White will hit a blot 9% of the time overall. Suppose then that White will win ~80% of games after a hit. Then we get ~7% additional wins for White, because of his hitting chances.

Decision

That would give 68% winning chances to Red here (32% for White). Obviously, it is a HUGE take now and the double is in question. Since White doesn't figure to have many big marketlosers, it might be better not to double here; better doubling opportunities might present themselves later, or, if things go wrong, Red will be happy not to have doubled.

Matchscore effects (advanced)

The matchscore here changes things a little compared to a money game. Basically you double a bit earlier and drop a bit quicker at 3-away both. If your money takepoint is 21.5%, then it's more like 23-24% here. Obviously still a huge take for White if the 32% is somewhat correct, but with White tending to drop cubes earlier at this matchscore, there is more incentive for Red to double now (probably more and bigger marketlosers around).

Rollout

And now... for GNUBG's 2-ply rollout verdict:


Cube analysis
Rollout cubeless equity  +0,383 (Money:  +0,383)

Cubeful equities:
1. No double            +0,735
2. Double, pass         +1,000  ( +0,265)
3. Double, take         +0,724  ( -0,011)
Proper cube action: No double, take (3,9%)
Rollout details:
Centered 1-cube:
   69,2%   2,2%   0,1% -  30,8%   2,2%   0,1% CL  +0,383 CF  +0,735
[  0,1%   0,1%   0,0% -   0,1%   0,1%   0,0% CL   0,001 CF   0,004]
Player You owns 2-cube:
   69,4%   1,9%   0,1% -  30,6%   2,3%   0,1% CL  +0,983 CF  +0,724
[  0,1%   0,1%   0,0% -   0,1%   0,1%   0,0% CL   0,004 CF   0,005]

Full cubeful rollout with var.redn.
2592 games, Mersenne Twister dice gen. with seed 1052117952 and quasi-random dice
Play:  2-ply cubeful prune [world class]
keep the first 0 0-ply moves and up to 6 more moves within equity 0,09
Skip pruning for 1-ply moves.
Cube: 1-ply cubeful prune


Conclusion

So, we see a little over 69% winning chances for Red here. Gammons are rare from here (for both sides) and not really a factor. With White's 30.6% winning chances, it is a HUGE take even at the matchscore (although not nearly as huge as GNUBG's 2-ply evaluation says: it's a good idea to always roll out interesting positions, especially cubes). It's probably not a double but it is VERY close to one. In practice, it looks like a good double since you lose only very little after a take, while you gain enormously if your opponent erroneously passes! GNUBG's figure in brackets after "No Double, Take" points this out: If there's at least a 3.9% chance that your opponent passes, then it's actually a good double!

[sixty_something]

  an excellent analysis, Zorba .. thanks for taking the time to put that together .. since starting this little problem series, i've been thinking it would be cool to have others elaborate on problems we all struggled with that might come up here accidentally in posting curious positions and situations .. i think you've just completed Chapter 1 .. with very little work, inserting the board image is about all, it would be a great article for adding to the fibsboard collection

again, great job and most appreciated .. also, i've locked the voting on this one .. Zorba's post is too good and too comprehensive to hide behind a spoiler button



Do not pursue what is illusory - property and position: all that is gained at the expense of your nerves decade after decade and can be confiscated in one fell night. Live with a steady superiority over life - don't be afraid of misfortune, and do not yearn after happiness; it is after all, all the same: the bitter doesn't last forever, and the sweet never fills the cup to overflowing.
-- Alexander Solzhenitsyn Russian author & dissident in US (1918 - )

[dorbel]

Well done Zorba, comprehensive, readable, informative, a model analysis.

[playBunny]

"In practice, it looks like a good double since you lose only very little after a take, while you gain enormously if your opponent erroneously passes! GNUBG's 3.9% figure in brackets after "No Double, Take" points this out: If there's at least a 3.9% chance that your opponent passes, then it's actually a good double!"

Now that's interesting. I've never known what to do with that number!