[IMG][/IMG]

This is the 2nd game of a match to 7 trailing 2-0.

My 1st instinct tells me to hold onto my anchor and build my board hoping to hit as GNU bears off possibly with 13-5(2) or 6-2, 13-5, 13-9

but GNU's top 5 moves all suggest giving up the anchor and leaving the 24 and 21 points blotted.

1. Cubeful 0-ply 21/17 16/12 13/9(2) Eq.: -1.169
0.158 0.002 0.000 - 0.842 0.521 0.025
0-ply cubeful prune [expert]
2. Cubeful 0-ply 21/17 13/9(2) 6/2 Eq.: -1.173 ( -0.004)
0.165 0.002 0.000 - 0.835 0.540 0.027
0-ply cubeful prune [expert]
3. Cubeful 0-ply 21/9 13/9 Eq.: -1.210 ( -0.041)
0.152 0.002 0.000 - 0.848 0.548 0.019
0-ply cubeful prune [expert]
4. Cubeful 0-ply 21/17 13/9(3) Eq.: -1.216 ( -0.047)
0.154 0.002 0.000 - 0.846 0.556 0.025
0-ply cubeful prune [expert]
5. Cubeful 0-ply 21/17 16/12 13/9 6/2 Eq.: -1.233 ( -0.064)
0.149 0.002 0.000 - 0.851 0.562 0.022
0-ply cubeful prune [expert]

I can't see how almost definately getting hit next turn can be to my advantage? If I leave the back checkers as is I at least have anchor to come out onto reducing the gammon possibilities while still keeping 2 points occupied for my OPP to deal with.

Can anyone explain why leaving 2 blots is a good move?

Hello PocketTrips,

that is a very interesting question, and I'm not sure if I would have been able to come up with GNU's suggestion over the board without seeing the rollout first. Having seen the rollout, it all makes sense to me.

At first you have to realize that you are about to lose a gammon (more than 50% of the time according to your rollout) no matter how you play the roll. If you lose a gammon you will be trailing 0-6. GNU therefore opts for active defense.

What GNU is suggesting is spreading your checkers across the outfield, which maximizes your chances to contain a checker once you hit it. This gives you the best chances to prime the hit checker and maybe turn the around but a least save the gammon.

Look at it like this. You need to hit very desperately. So let him hit you and then you will get a least a shot. He very likely won't leave a shot in the next two rolls anyway because he can play all his numbers safely. He probably doesn't want to hit open because his gammon chances are good enough even without hitting.

But if he does hit you picking and passing one of your checkers will be lurking on the bar, always ready to hit a blot he has been forced to leave.

Your original choice, 13/5(2), is correct, for all the obvious reasons.

GNU's plays are hopeless. I had heard GNU was weak in backgames, but this looks absurd even for GNU. I would check and make sure that the position you submitted for a rollout is the same as the one in your screen shot. I suspect that you altered the position in some way before starting the rollout.

Did you just click the Tutor button and copy that, because I did a rollout using the same match settings and got this:

1. Rollout 13/5(2) MWC: 18.03%
0.123 0.000 0.000 - 0.877 0.652 0.014 CL 17.11% CF 18.03%
[0.001 0.000 0.000 - 0.001 0.002 0.001 CL 0.06% CF 0.07%]
Truncated cubeful rollout (depth 11) with var.redn.
1296 games, Mersenne Twister dice gen. with seed 814630727 and quasi-random dice
Play: 0-ply cubeful prune [expert]
Cube: 0-ply cubeful prune [expert]
2. Rollout 21/17 13/9(2) 6/2 MWC: 17.06% ( -0.97%)
0.121 0.002 0.000 - 0.879 0.712 0.018 CL 16.13% CF 17.06%
[0.001 0.000 0.000 - 0.001 0.003 0.001 CL 0.07% CF 0.08%]
Truncated cubeful rollout (depth 11) with var.redn.
1296 games, Mersenne Twister dice gen. with seed 814630727 and quasi-random dice
Play: 0-ply cubeful prune [expert]
Cube: 0-ply cubeful prune [expert]
3. Rollout 21/17 13/9(3) MWC: 16.98% ( -1.05%)
0.116 0.001 0.000 - 0.884 0.703 0.020 CL 16.09% CF 16.98%
[0.001 0.000 0.000 - 0.001 0.002 0.001 CL 0.06% CF 0.07%]
Truncated cubeful rollout (depth 11) with var.redn.
1296 games, Mersenne Twister dice gen. with seed 814630727 and quasi-random dice
Play: 0-ply cubeful prune [expert]
Cube: 0-ply cubeful prune [expert]

1. Rollout 13/5(2) Eq.: -1.406
0.123 0.000 0.000 - 0.877 0.652 0.014 CL -1.475 CF -1.406
[0.001 0.000 0.000 - 0.001 0.002 0.001 CL 0.004 CF 0.005]
Truncated cubeful rollout (depth 11) with var.redn.
1296 games, Mersenne Twister dice gen. with seed 814630727 and quasi-random dice
Play: 0-ply cubeful prune [expert]
Cube: 0-ply cubeful prune [expert]
2. Rollout 21/17 13/9(2) 6/2 Eq.: -1.479 ( -0.073)
0.121 0.002 0.000 - 0.879 0.712 0.018 CL -1.549 CF -1.479
[0.001 0.000 0.000 - 0.001 0.003 0.001 CL 0.005 CF 0.006]
Truncated cubeful rollout (depth 11) with var.redn.
1296 games, Mersenne Twister dice gen. with seed 814630727 and quasi-random dice
Play: 0-ply cubeful prune [expert]
Cube: 0-ply cubeful prune [expert]
3. Rollout 21/17 13/9(3) Eq.: -1.485 ( -0.079)
0.116 0.001 0.000 - 0.884 0.703 0.020 CL -1.552 CF -1.485
[0.001 0.000 0.000 - 0.001 0.002 0.001 CL 0.004 CF 0.005]
Truncated cubeful rollout (depth 11) with var.redn.
1296 games, Mersenne Twister dice gen. with seed 814630727 and quasi-random dice
Play: 0-ply cubeful prune [expert]
Cube: 0-ply cubeful prune [expert]

This being a match I assume we want to use Match Winning Chance rather than equity, but 13/5(2) is best either way.

And I really believed there was something to GNU's suggestion. Are back games the only area where one shouldn't trust GNU?

I was playing with tutor mode and just hit the "hint" button and copied GNU's suggestions from there once all the plays that seemed reasonable to me came up as "doubtful." Glad to see I am not commpletely off-track with my thinking.

If it wasn't so easy for GNU to pick and pass on either point the next roll I could see your arguement. If He had say 5 or less checkers on his 5 and 6 points combined and my board was still as open as it is. I think this move would make sense since there is a better chance GNU will leave a shot while hitting. With the board as it is, 13-5(2) closes a 3rd point on my board in preparation for a hit and gives me 2 points to close the 2 point from on my next roll.

A three ply evaluation also says 13/5(2). I trusted the word "rollout" to much and tried to make sense of it...

Out of curiosity I ran this position in Snowie 4 and it doesn't even have splitting the anchor anywhere in its list of top 20 moves. 13/5(2) is clearly best followed by 13/9(3) 9/5, 16/12 13/5 6/2, 13/9(2) 9/5 6/2, and 16/12 13/9(2) 9/5. Snowie is also more optimistic about our overall match winning chances (over 19%, compared to 18.03% by GNU).