why didn't it double?

[badgolferman]

I was ahead 6 games to 1 and lost the Jacoby game to make it 6 games to
2. I started off with a 5-2 roll. GNUBG did not double but started
off with a double 5 to kill my two blots in it's home board. Why
didn't GNUBG double before it rolled? Did it know what was coming?




GNU Backgammon Position ID: M07wATDg8+ABYA
Match ID : cAngACAAMAAE
+13-14-15-16-17-18------19-20-21-22-23-24-+ O: gnubg
| X O | | O O O | 2 points
| X | | O O O |
| X | | O |
| X | | |
| | | |
v| |BAR| | 7 point match (Cube: 1)
| O | | X |
| O X | | X |
| O X | | X |
| O X | X | X O | On roll
| O X | X | X O | 6 points
+12-11-10--9--8--7-------6--5--4--3--2--1-+ X: Home

[badgolferman]

badgolferman wrote:

> lost the Jacoby game

I meant to say the Crawford game.

[Paul]

On Mar 31, 10:09 pm, "badgolferman" <REMOVETHISbadgolfer...@gmail.com>
wrote:

No, gnu did not know what was coming.

Gnubg lost nothing by delaying doubling. It's now a take whatever you
roll because of the match score.

See http://www.bkgm.com/faq/Matches.html#what_is_the_trick_ for more
explanation.

I think gnu doubled after your next roll. If you dropped then you
made a mistake.


Paul Epstein

[Walt]

No, gnubg didn't know what was coming - gnu doesn't know what rolls are
coming any more than you do.

After the 5-2 opening, it saw that you were ~51% to win and evaluated
the cube decision as "optional double". Doubling wasn't *wrong*, but it
didn't increase gnu's equity either, so it held.

It also understood that even if it rolled it's best number (and 55 is
probably as good as it gets) and you danced, you would still have a
take. Thus with no advantage in cubing, and no market losers it's
correct to hold. Yes, that's right, you have a TAKE here, even if you
dance.


The main issue is that your takepoint at odd-away post-Crawford is
really really small - you can take just about anything here. There's
something called 'the trick', which is waiting to double at odd-away
post-Crawford in hopes that the opponent will incorrectly drop. That's
what's going on here.

See http://www.bkgm.com/faq/Matches.html#what_is_the_trick_

Bottom line at post crawford

1) if your opponent is an even number away, drop anything less than 50%
2) if your opponent is an odd number away, take damn near anything
unless the gammon rate is large. (at 1-away 6 away, gammmons aren't
much of a factor)


Look at what happens if you dance:

GNU Backgammon Position ID: 4PPgAWAzTvABMA
Match ID : MAHgACAAMAAE
+12-11-10--9--8--7-------6--5--4--3--2--1-+ O: O

| X O | | O O O | 2 points

| X | | O O O | On roll
| X | | O |
| X | | |
| | | |
^| |BAR| | 7 point match (Cube: 1)

| O | | X |
| O X | | X |
| O X | | X |
| O X | X | X O |

| O X | X | X O | 6 points

+13-14-15-16-17-18------19-20-21-22-23-24-+ X: X




Cube analysis
2-ply cubeless equity +0.369 (Money: +0.651)
0.666 0.390 0.011 - 0.334 0.077 0.004
Cubeful equities:
1. Double, take +0.884
2. Double, pass +1.000 ( +0.116)
3. No double +0.830 ( -0.055)
Proper cube action: Double, take

In this position, gnu has rolled its best number and you rolled your
worst, but the cube action is still D/T. Since gnu has no market losers
in your original position, it has no reason to double.

Post-Crawford odd-away, it's going to wait until it has market losers to
double and since your takepoint is so low that may be awhile.


//Walt

[Paul]

On Apr 2, 2:58 am, Walt <walt_ask...@yahoo.com> wrote:
...

>
> Bottom line at post crawford
>
> 1) if your opponent is an even number away, drop anything less than 50%

...

I know you were only doing a beginner's summary. However 1) is
incorrect, and I think it can be quite seriously incorrect.
1) is accurate for 1 away 2 away. However, it's false for 1 away,
large even away.

If the underdog keeps winning then the favourite will have a choice as
to when the free drop is used. So the free drop shouldn't be used if
the game winning chances are > 49%. It might be worth waiting for a
better chance for a free drop as the match develops.

Paul Epstein