Match 1, Game 3, Move 5 Forum

[Zorba]

I won't change my vote in any case, I'd rather have my mistakes and blunders recorded for history!

[lewscannon]

Spoiler

I voted otherwise, but the arguments here have convinced me that the double hit is a good idea. If this goes wrong, however, I'll be blaming everyone else.

[close]

[stog]

yes i'm willing to change my vote too

Spoiler

4 the double hit!

[close]

if we are quick i wonder if we can change the outcome of the vote - do u need a moderator to change your vote lews? or others if u need us to change your vote - post your request quickly!

[playBunny]

Spoiler

I'd probably do the double hit if I was feeling like some reckless fun and it was a longer match and if my opponent wasn't formidable factotum. ;-)

There are things that I don't like about the double hit here. One is that one of the hits is deep and another is that we strip our midpoint in order to do so. Next roll we'd have only one builder to capitalise on the potential gain (although a 2 would give us 4/2 and a deep home point, whoopee) and might have to bust the 8-point. The 2-point hit is almost sacrificial but I don't think that it gains enough. Another point against hitting is that if we get double-tapped in return, we're looking to have our 11-point blot zapped while we scramble to our feet over in our opponent's home. I know I'm focusing only on the negatives here but they do seem fairly substantial to me.

[close]

[ah_clem]

do u need a moderator to change your vote lews? or others if u need us to change your vote - post your request quickly!

There's a "Remove Vote" link that works.  After you've removed your vote, you can vote again.  No need to bother the mods. Thanks to blitzxz for implementing that functionality.

[Zorba]

After playBunny's analysis, I suggest everybody changes their votes once more  

[ah_clem]

Don't tempt me.  I just might change my mind again!  (c:

Anyway, I see that the votes for safe plays are spread across four different moves while the bold voters are stacked up on one.  If anybody's going to change votes, I'd suggest that the "safe" voters consolidate their support behind one move.  Or go over to the dark side  (c:

[diane]

Spoiler

I'd probably do the double hit if I was feeling like some reckless fun and it was a longer match and if my opponent wasn't formidable factotum. ;-)

There are things that I don't like about the double hit here. One is that one of the hits is deep and another is that we strip our midpoint in order to do so. Next roll we'd have only one builder to capitalise on the potential gain (although a 2 would give us 4/2 and a deep home point, whoopee) and might have to bust the 8-point. The 2-point hit is almost sacrificial but I don't think that it gains enough. Another point against hitting is that if we get double-tapped in return, we're looking to have our 11-point blot zapped while we scramble to our feet over in our opponent's home. I know I'm focusing only on the negatives here but they do seem fairly substantial to me.

[close]

This is very much what I am feeling here - and couldn't see what dorbel meant by his comment that 'either of the two jokers were excellent rolls for white'...if he could get back in time to explain that to me, I might be more persuaded.

[dorbel]

If White is not on the bar, he has no bad rolls and 4-4 and 2-2, the rolls that you fear both play very well. He'll probably make the 9pt and 4pt with 4-4, while 2-2 anchors on the 21pt and makes the 4pt. It's an interesting paradox that the thing that you most fear is being on the bar, yet you are reluctant to put your opponent there!
I don't actually think that 21/15, 13/11 is a bad play by the way. For all I know it may be best! If that's what you like go for it!

[diane]

Spoiler

It isn't a fear of being put on the bar as such - for us to put him on the bar, we have to leave two blots - if he rolls either of those two jokers, he puts two of our men on the bar against a two point board and leaves us nothing - with the option on taking our third one, as we waste our roll coming off the bar [if we dont fan]. It is just how bad this potentially is, whilst he holds the cube, that turns me away from it.  I also really don't like giving him time to consolidate his position - but I also think we are ahead and safeish, and that we don't need the hits to take these two points.  I think I will stick with my current vote.

[close]

[lewscannon]

Spoiler

It isn't a fear of being put on the bar as such - for us to put him on the bar, we have to leave two blots - if he rolls either of those two jokers, he puts two of our men on the bar against a two point board and leaves us nothing - with the option on taking our third one, as we waste our roll coming off the bar [if we dont fan]. It is just how bad this potentially is, whilst he holds the cube, that turns me away from it.  I also really don't like giving him time to consolidate his position - but I also think we are ahead and safeish, and that we don't need the hits to take these two points.  I think I will stick with my current vote.

[close]

It looks like Diane's league matches against me have psychologically scarred her.

[ah_clem]

Rollout:

Spoiler

The safe play is the best, but not by alot.  As I was watching the rollout progress, the double-hit went into the lead several times, even though the safe play won in the long run. And the double-hit was the winner in the evaluation.  No wonder so many of us had a hard time with this one.


    1. Rollout          21/15 13/11                  MWC:  52.05%
       0.595 0.177 0.009 - 0.405 0.099 0.003 CL  54.50% CF  52.05%
      [0.001 0.001 0.000 - 0.001 0.000 0.000 CL   0.04% CF   0.04%]
        Truncated cubeful rollout (depth 11) with var.redn.
        1296 games, Mersenne Twister dice gen. with seed 858129907 and quasi-random dice
        Play: 0-ply cubeful prune [expert]
        Cube: 0-ply cubeful prune [expert]
    2. Rollout          8/2* 6/4*                    MWC:  51.77% ( -0.28%)
       0.564 0.236 0.015 - 0.436 0.121 0.007 CL  54.52% CF  51.77%
      [0.001 0.001 0.000 - 0.001 0.001 0.000 CL   0.05% CF   0.04%]
        Truncated cubeful rollout (depth 11) with var.redn.
        1296 games, Mersenne Twister dice gen. with seed 858129907 and quasi-random dice
        Play: 0-ply cubeful prune [expert]
        Cube: 0-ply cubeful prune [expert]
    3. Rollout          21/13                        MWC:  51.47% ( -0.58%)
       0.587 0.161 0.007 - 0.413 0.095 0.003 CL  53.99% CF  51.47%
      [0.001 0.001 0.000 - 0.001 0.000 0.000 CL   0.03% CF   0.04%]
        Truncated cubeful rollout (depth 11) with var.redn.
        1296 games, Mersenne Twister dice gen. with seed 858129907 and quasi-random dice
        Play: 0-ply cubeful prune [expert]
        Cube: 0-ply cubeful prune [expert]
    4. Rollout          21/15 6/4*                   MWC:  51.01% ( -1.04%)
       0.565 0.192 0.011 - 0.435 0.115 0.005 CL  53.69% CF  51.01%
      [0.001 0.001 0.000 - 0.001 0.000 0.000 CL   0.04% CF   0.04%]
        Truncated cubeful rollout (depth 11) with var.redn.
        1296 games, Mersenne Twister dice gen. with seed 858129907 and quasi-random dice
        Play: 0-ply cubeful prune [expert]
        Cube: 0-ply cubeful prune [expert]
    5. Rollout          24/18 13/11                  MWC:  50.99% ( -1.06%)
       0.576 0.161 0.009 - 0.424 0.104 0.004 CL  53.58% CF  50.99%
      [0.001 0.001 0.000 - 0.001 0.000 0.000 CL   0.04% CF   0.04%]
        Truncated cubeful rollout (depth 11) with var.redn.
        1296 games, Mersenne Twister dice gen. with seed 858129907 and quasi-random dice
        Play: 0-ply cubeful prune [expert]
        Cube: 0-ply cubeful prune [expert]

[close]

[playBunny]

 
Rollout 1296 trials, 2-ply for 12 plies, 0-ply thereafter, 2-ply cube throughout

Spoiler

# .. Move .......... Equity . (Diff) ... Win .... WinG ... WinBg -- Lose ... LoseG . LoseBg
1 .. 21/15 13/11 ... -0.110 ............ 59.7% .. 25.4% ... 4.7% -- 40.3% ... 9.9% ... 1.0%
2 .. 8/2* 6/4* ..... -0.148 (-0.038) ... 58.0% .. 28.3% ... 6.9% -- 42.0% .. 11.1% ... 2.1%
3 .. 21/13 ......... -0.162 (-0.052) ... 58.2% .. 23.1% ... 2.7% -- 41.8% ... 9.5% ... 0.9%
4 .. 24/16 ......... -0.188 (-0.079) ... 57.6% .. 22.8% ... 3.9% -- 42.4% .. 10.2% ... 1.0%
5 .. 24/18 13/11 ... -0.214 (-0.104) ... 56.9% .. 24.6% ... 5.6% -- 43.1% .. 10.8% ... 1.1%
Std Err ............. 0.008 ............. 0.2% ... 0.5% ... 0.4% --- 0.2% ... 0.1% ... 0.2%

Full cubeful rollout with var.redn.
1296 games, Mersenne Twister dice gen. with seed 858261885 and quasi-random dice
Play: supremo 2-ply cubeful prune [world class]
keep the first 0 0-ply moves and up to 16 more moves within equity 0.32
Skip pruning for 1-ply moves.
Cube: 2-ply cubeful prune [world class]
Different evaluations after 12 plies:
Play: 0-ply cubeful prune [expert]
Cube: 2-ply cubeful prune [world class]

[close]

[dorbel]

Interestingly, a 2-ply cubeless Snowie rollout (equivalent to Gnu 1-ply), supported the double hit that Snowie 3-ply will make. The win rates were broadly the same but Snowie won a lot less gammons (than Gnu) with either play. However, for those who didn't want to play the double hit because they thought that it lost too many gammons, as you can see here, Snowie agrees that the gammon losses are more or less equal.
I will do another rollout using cubeful 3-ply, but it is worth noting that Gnu is usually reckoned to be better for this sort of position, because of its ability to rollout according to score.

[dorbel]

And rollout fans may be interested to know that a long 3-ply cubeful rollout gave the same result.
The double hit won 57.3 games with 23.4 gammons, 21/15, 13/11 won 59.5 games with 19.7 gammons. Both plays lost 10 and a bit gammons. This gives 8/2*, 6/4* a 0.033 advantage cubeless, but only 0.01 cubeful, both results lying well within the margin of error in any case.
It isn't clear to me why Gnu wins so many more gammons. I would be slightly more inclined to believe Gnu's results from experience, but I have no evidence for that. Perhaps the two plays are about equal.

[Zorba]

My 2-ply GnuBG rollout is still running, but lower settings rollouts and evaluations all pretty much agreed so far that the double hit is about a 0.06 error. I'll post it later.

BTW, ah_clem's rollout looks like it was done with the wrong score or cube level, our MWC should be much higher, also, on my GnuBG all evaluations prefer 21/15 13/11 by a good margin.

That Snowie cubeless has the plays close makes sense, I'm sure GnuBG would get a similar result. The difference is the matchscore, where our gammons don't have full value on the 2-cube, and zero value once the cube reaches 4 (with factotum's gammons always valuable). Since factotum should be quick on the redouble here (again because of the score), relatively many games will get played on a 4-cube.

One that cube reaches 4, factotum can afford to stay back with  his backcheckers forever. That's why we will win many more gammons when this is played out according to score, with a live cube, as GnuBG plays it. It's the kind of gammons that are not useful for us, so that explains the apparent paradox: Gnubg wins many more gammons after the double hit, but it doesn't show up in the cubeful equity, more like the opposite. It's because once gammons don't count, we will win many more of them.

The effect is even clearer when looking at backgammons, that already have lost any value for us on a 2-cube. We win much more of them, because of that.