GnuBG recommends running with both back checkers, but that makes no sense to me. Black is going to get hit 75% of the time, greatly increasing the chances of a gammon. Even when black isn't hit, there's little chance of safetying both blots on the next roll, so white likely has multiple shots from this move. The one advantage I can see is that it gets both checkers past the 4-point prime, but that advantage seems minimal when you're likely to get at least one checker put back on the Bar on the next play.

I chose to play 13/2, covering the blot in my inner board and preserving the defensive anchor. My blot on the 13-pt is a favorite not to get hit, and I still have chances of picking up a blot as white tries to get home or rolling a big double to extricate myself.

Am I missing a key advantage to the recommended running play, or is this a case of the search including bad plays in its evaluation (slotting the 3- or 4-pts in White's home board at the same time Black is hit, giving Black return shots) that artificially inflate the worth of the move?

Look a couple rolls into the future. You're going to have to abandon the anchor very soon, and if you abandon the midpoint white can distribute his checkers to maximize the chances of hitting you as you run. Furthermore, that 4-prime in front of your anchor could cause your board to crack, which will allow white to attack more aggressively. By running now, if you happen to be hit, you preserve your chances of maximizing the benefit of any return shot you might have.

By running both checkers instead of just one, you do not allow white to point on your head, and thereby accomplish two very good things at once. He now needs to hit in the outfield, which means you're going to have two entering numbers instead of one.

I think this is a very interesting position. I'm sure, I would make your play OTB as well.

Looking at the numbers for games won and gammons lost I think the score is a big factor. The cube hasn't been turned yet, but white is probably closing in on a double here. When he doubles getting gammoned isn't gonna hurt that much (because you would be down 6-2 if you lose anyway).

I think 13/2 would be the clear play for money, because of the gammon threat.

btw, I still think it's better to post positions without "results", and then post evaluations/rollouts later.

I don't really think this is true. If white misses he has to lift his blot ofcourse. After that there are plenty of rolls, that safeties at least one of the blots, and only leaves an indirect shot.

I can see what you're saying, Aaron, but I'm still skeptical. White is pretty flat, and only has three checkers to distribute. Given my play, if he hits my blot, I could get stuck behind the 4-ptrime and crash my board, but I'm a favorite to get out. If he can't hit my blot (more likely), then he's got to move his checkers, and flat as he is he'll have to break a point or move the one free checker he has to shoot at me in the outfield without leaving a blot of his own.

If I run, as GnuBG suggests, then 75% of the time I get hit. In that case, I'm stuck on the bar over 44% of the time, and White shoots at my second blot. Even when I get in I'm still facing the same 4-prime and he gets chances to point on me. Even the 25% when White misses, I still have to get past his anchor on my 7-pt while he probably strengthens his inner board.

I'm going to try a roll out -- I'll report back the results.

-Phil in KC

I make GNU's play. How else can you win?

There's a theme of connectedness here. There might even be some examples in Modern backgammon that look like this.

If white misses, he's going to spend the next roll cleaning up that blot. And then what happens for you? You've got to run YOUR blot to safety (if you can). Then his next roll will be to break his back anchor to maximize his shots if you roll any 6 (or maybe even a 5). The only way you get out of it is to roll 55/66. Otherwise, you're going to have an awful time getting your back checkers home. In other words, you're moving towards giving up the game and simply playing not to get gammoned.

Also, the match score matters. I should have said this at the beginning, but because I have never played a cash game, it's basically intuitive for me to look at the score and incorporate that into my thinking. White doesn't want to cube because he's 3-away and you're 'big'-away and there's contact (Stick's law?). He has little to gain by doubling but stands to lose a lot by a potential recube, so that cube is probably just going to sit on the bar. For money (as mute points out), this could well be the right play BECAUSE you might face a cube which leads to a costly gammon. It would be interesting to look at this position at DMP (1-away/1-away). Since losing a gammon in that situation is no worse than losing a regular game, it might once again swing back to GNU's running play. After all, you'll be ahead in the race, so you should be looking to race.

Also, white's play once you jump his prime is to stay back since you've got a decent racing lead. So he's not going to be running in the outfield. Instead, he's going to take the next roll or two to try to fill in his board.

It took a while for me to really appreciate the differences in play at different scores (especially in short games). I'm still learning about this sort of thing.

Just what Stick's fragile ego needs, a law named after him.

My rollout for the two plays, using World Class checker play, gives 21/16, 21/15 the edge by .004, which is within the margin of error. The rollout was for money, so it might have been different for the actual match score. Interestingly, 21/16, 21/15 wins 5.6% more games, but loses gammons 25% of the time, while 13/2 loses gammons only .2% of the time, thus making up for the fewer victories.

Considering that black gets hit 75% of the time with 21/16, 21/15, I'm amazed at the number of wins for black, though I'm not surprised at how often black is gammoned.

I just sat down to calculate the point differential between the two moves in a money game, based on the following rollout results:

Roll W Gam BGam L Gam BGam
21/16, 21/15 .355 .082 .004 .645 .253 .005
13/2 .299 .029 .000 .701 .002 .004

By my calculations, here are the points won and loss in 100 games:

21/16, 21/15 won 44.5 points, and lost 90.8 points, for a net of -46.3 points
13/2 won 32.8 points, and lost 71.1 points, for a net of -38.3 points

This doesn't match the equity results, which were almost equal, so my first thought was that the difference must be based on the cube -- with 13/2, Black accepts the cube more often, making some of the losses more expensive. However, when I checked the cube analysis (4-ply) for the resulting positions, white should double and black should take for either position, so the cube shouldn't matter. After 21/16, 21/15, in 8.6 games black wins a gammon or backgammon, gaining 9 points if he doubles back at the right time, without ever loosing his market. After 13/2, black would gain 2.9 points using the same assumptions. The adjusted points are -37.3 for 21/16, 21/15 and -35.4 for 13/2. So, with perfect cube play, 13/2 still has a smaller negative equity. I'm confused about how GnuBG comes up with its equity calculations.

Incidentally, over the board I won this game in a race. White ended up breaking his 4-prime, giving up his 7-pt, and I ended up leaving one direct shot and a couple of indirect shots that were missed as I maneuvered past his checkers.

Aaron, with over 12,000 posts, I guess you'd know. Thanks for the tip!

-Phil in KC