In reviewing GnuBG's analysis of a match, I came across this curious position:

12 11 10 9 8 7 6 5 4 3 2 1
o o o o o o
o o o
o o o
o


x
x x x x x x o
x x x x x x x x o
12 11 10 9 8 7 6 5 4 3 2 1

X was behind 2-4 in a match to 7, and O owned the cube at 2. X rolled a 2-1. In the match, X played 9/7, 8/7 to complete the prime. In GnuBG's analysis, using a 2-ply search, completing the prime was the 7th best move, loosing -.78% equity. The best move was 8/5, and some of the better moves included slotting the 7-point. The extra gammons and backgammons don't amount to much -- .165 to .151 and .006 to .004 respectively. Is there something I should learn from this position, or is this an anomaly of GnuBG's evaluation?

-Phil in KC

p.s. I noticed that someone had posted an image of a position -- any pointers on how to do this so I can make any future inquires clearer with a screen shot?

If you post the alphanumeric GNUBg ID: string we can see it for ourselves.

Edit: Screenshots

Hit PrintScreen, Edit/Paste into MSPaint, and rehost at imageshack/tinypic/photobucket/etc.

Here's the image:



Thanks for the pointer on posting it!

-Phil in KC

You really don't want the bar point. It's just another point you're going to have to clear in a couple of rolls. Instead, you should be looking to give yourself the best chance of bearing off safely by getting your high points with spares on them. Since the ace point is blocked, 5s are going to be your unhappy numbers, and therefore, spares on the 5 are the most important.

Aaron, thanks for your response. Having to break the 7-point is an issue, but so is getting hit! With three blots, I'm not likely to cover them all in one roll, either, so there are multiple turns where I leave a target. Granted, I get to shoot at his blots if I get hit, but I'll miss his blots almost one third of the time (11/36ths).

I don't know how to find out the percentages for leaving a shot with a 7-2 prime against a two checker anchor on the 1-pt. There sure are a lot of variables in BG!

-Phil in KC

It looks like if you complete the prime villain still has plenty of time to build a prime of his own while you bear in and bear off. Be advised that I suck at backgammon. I was surprised to see this position is a 50/50 coinflip.

If he hits in return, he's going to have to use both checkers to hit, which means that you're going to have a double-direct shot (plus 61) to return. Not only that, but then he'll have a missing anchor and you'll have a chance to close him out which will drive up your gammon chances in a huge way.

I don't know of any good way of computing those chances. There are loads of combinations to consider. I only know (sort of) how to minimize the times that it happens. I think I learned this particular bit from one of Kit Woolsey's books.

Falcon, this isn't a 50/50 split. Black wins over 85% of the games from this position. The 49% represents the Match Winning Chance (MWC), and the differences in that column indicate how a poor move degrades your chances. I'm still new at using GnuBG, so I'm not sure how far to trust its recommendations, especially ones that are counterintuitive like this one.

-Phil in KC

I suck more than I thought.

This position is an aberration.

In most similar positions, Black will want to play 9/7 8/7, both to protect any outfield blots from an indirect shot, and to make sure that White can't roll 6-6 and catch up in the race. What makes this position a weird one is that White is both 42 pips behind in the race (so blocking 6-6 isn't necessary) and has three blots (so safeguarding any Black outfield blots isn't necessary, since White is better off not hitting with rolls like 6-1 or 6-2.)

If we move a White checker from the 3-point back to the 4-point, and from the 6-point to the 5-point, producing a strong board with the same pip count, then 9/7 8/7 is the better play. If we instead move some White checkers into the inner board so the pip count differential is only 20 pips or so, then 9/7 8/7 is again the right play.

Difficulty clearing the bar-point later isn't the issue. In general, the full prime is a big asset when bearing home. It's the other factors that make this position exceptional.

Mr. R, thanks for your view on this position. If I understand this correctly, GnuBG gives a counterintuitive move the higher score because it includes clear mistakes in its search -- white won't hit with a 61 or 62, but the position sure looks good for black when he does, and those moves are included in the evaluation.

I'm new to working with backgammon programs and their analyses, so this is a revealing situation for me. I begin to see why rollouts are needed in some cases to clarify these types of anomalies.

-Phil in KC

Thanks for posting--very interesting position!

Here's my guess:

If you just make the bar point your opponent has the potential to develop a well-timed ace point game if he doesn't roll awkward numbers; but letting him hit you with 61, 62, or 63 gives you a chance to close him out, neutralizing his ace point game. You would also be guaranteed a double-direct shot from the bar after he hits you, so you're sending a 3'rd man back in the majority of the sequences where he rolls 61, 62, or 63--significantly bolstering your gammoning chances whether you close him out or not.

His board is very vulnerable and your board is very strong-- and as Bill was saying, even 66 isn't really a joker when you're this far ahead in the race-- so there are only a few very unlikely losing sequences even after he hits you (i.e. you fanning on re-entry, followed by him hitting you again and covering, followed by you missing the remaining home board blot, followed by him hopping your prime, etc etc).

Since it's more likely that his well-timed ace point game would succeed than it is that one of these sorts of long-shot losing sequences would occur, you're more likely to win by trying to force him to abandon his ace-point game-- it's the winningest play whether you count the extra gammons or not!

so let's say black plays like GNU wants and white rolls 1-6 . should white hit? or what ?

Definitely not at this match score. White would much rather lose a normal game for 2 points and still have an equity edge on the match instead of losing a gammon and have to win two games in a row.

I think for money, I would also sit tight and build the home board, but I think it would be a lot closer.

you' re right. I usually don't play matches , so i didn't considered the score. I've had a similar situation some time ago in a $game , and went to hit and get gammoned. Will try to recover the position to have some thoughts about.

I don't think white should hit for money either or at pretty much any score (and I don't think it's close at all), fwiw. It's just plain suicide gammonwise, and I'm not even sure if it wins more games.