Here's the position:



I chose one play, and GnuBG prefers another. I'll post my choice and GnuBG's recommendation tomorrow.

I want to double White on the next roll, so let's play close to vest here. I like the single hit, bar/20*/15(2).

After the hit, I'll be up by 23 pips, so I don't need to take too many more chances if I can escape both rear checkers. If I were hit two White checkers instead, playing bar/20*, bar/10*, White would have a few doubles that could force me to hold back the cube.

Yeah I agree with this. I just go for the double hit.

double hit seems mandatory. white have other two blots to hit in next roll .

Yep, looks like a double hit for me too: B/20* B/10*

After that, you follow up with scooping more outside blots or pointing on White entering checkers if you can.

I can see the logic behind B/20* (2) 8/3 (2), hoping that White doesn't enter both checkers, for the sake of building our board, but I don't think it's powerful enough. With a big joker like 5-5, you want to swing things around and be really aggressive. The double hit would allow to either prime or attack based on how we roll.

The double hit was my choice, with plans to turn the cube. GnuBG prefers Bar/20(2)*, 20/15(2), as shown below.

As it happens, my opponent was only able to enter one checker on the next roll, but took the cube anyway (big blunder). I wasn't able to close out my board, but had three white checkers trapped on the 1-pt, and missed the gammon by one pip.

I really don't understand GnuBG's evaluation function. It estimates that the double hit will win five less games in a hundred, but will win 12 more gammons. That's a worthwhile exchange, in my book, and seems like it should have a higher equity, but it doesn't. If anyone can explain why GnuBG's equity evaluation makes sense, I'd appreciate it.



Edit: I just did the math, and factoring in the extra lost gammons, by GnuBG's estimations, the single hit garners 51 points in 100 games, while the double hit garners 49 points, despite winning almost twice the number of gammons. I still don't get GnuBG's evaluation function, though, because this looks to me like an equity difference of 2 points in 100 games, or .02, not the .09 shown in the table.

Yeah, I was playing around with this last night. I would have played B/20 B/10 as well. What's interesting here is that the cubeless equity is pretty close, but the cubeful equity diverges way in favor of B/15*2. I analyzed it as a jacoby-free cash game just in case there was some weirdness with the match score.

B/15*2 Cubeless .505, Cubeful .820
B/20 B/10 Cubeless .462, Cubeful .650

So the equity difference is about 4 times as big cubeful as cubeless, even though the cube is centered. What's going on here?

After B/15*(2) there is little variance. You do very well on any response but 5-5 and even that is not too bad.

After the double hit, a doublet by opponent is a disaster and on the other hand his worst numbers are capped at -1.00 for a double/pass.

This is what Taper_Mike explained. When you are near a strong double, you tend to make small plays. If the average equity is 0.8 in both cases for instance, you prefer the 36 outcomes of the response to fall between, say 0.6 and 1.0 rather than 0.0 and an hypothetical 1.6 that is in fact capped at 1.0.

A nice tool to look at this is the "Temp.Map" button of the Hint panel. You select a move and this button shows you the 36 responses.

Thanks, plm. This says it better than I did.

Interestingly, my first take was to go with the aggressive play. Wamy Einehouse quoted me before I edited my initial response. If this had not been posed as a problem, I probably would not have reconsidered. But then I focused on the cube...

Is this another instance where the Jacoby Rule "interferes" with natural play? Or is the conservative play correct in all cases?

It's not a jacoby issue. It's actually a match spot (5-away/7-away), but the non-jacoby cash analysis and match analysis are basically identical. If equity goes up to 1.6, it isn't capped. It might be a similar issue though. You have a chance to go from no double to too good (and stay there), and not be able to use the cube, while White will actually get to double and only rarely jump from no double to too good.

Grunch. 25/20*(2), 8-3 (2).