Game 6, move 1 : Forum 4-3

[diane]

Well , after rkb's save, we are into game 6, and the forum, blue,  has 4-3 to open with.

Again, there are more options, but hopefully all the ones we might want to think about are there.

4HPwATDgc/ABMA:8ImxACAAIAAA

[ah_clem]

Spoiler

 24/21 13/9 is the book move for double match point.

"Aha!" you may say.  "But it's not DMP."

True, but it's very like DMP. Gammon wins don't count for us, and gammon losses don't hurt us very much - the only difference between match score of 4-3 and 4-4 is that we get a free drop at 4-3.  That's not enough of a difference for me to change my play based on it.

So, I say go with the book move.  
 24/21 13/9

As this game continues, let's bear in mind that gammons are basically irrelevant for both sides.

[close]

[roadkillbooks]

Not that I care which move you make but what book are you talking about.  Most first move things I've seen say 24/20, 13/10.  Of course there are times when I move 24/21. 13/9 but it's mainly with people Im not sure or not if there smack me on their 4 point. 
roadkillbooks

[spielberg]

Spoiler

At DMP with this I'd drop two from the mid with 13 9 , 13 10. Here at Crawford, 3 away that's too aggressive so I'll split my back men and drop one from the mid point. I much prefer 24 20, 13 9 to 24 21, 13 10. We're better diversified with the move to his Golden point I'm sure.

[close]

[dorbel]

Spoiler

This has been extensively rolled out by several people, including me. The best play at dmp, which this effectively is, is 24/20, 24/21!

[close]

[ah_clem]

Not that I care which move you make but what book are you talking about. 

See  http://www.bkgm.com/openings.html#opening43

[diane]

See  http://www.bkgm.com/openings.html#opening43

Spoiler

OOh, I found it without the book, that was good reading though, and I do so like it when I am right  - thanks for the link clem 

[close]

[spielberg]

Spoiler

This has been extensively rolled out by several people, including me. The best play at dmp, which this effectively is, is 24/20, 24/21!

[close]

Dorbel, interesting thoough that is (and I can't wait to play it), this is way away from DMP. We lead 4-2 Crawford hence my Crawford, -3 summary of the match score.

See  http://www.bkgm.com/openings.html#opening43

Spoiler

The book's on generic openings though. At this score I'm sure it's 24 21 , 13 10 - it will be a small difference but I bet it's better than 24 20, 13 9 which will be second best.

[close]

[ah_clem]

Spoiler

Dorbel, interesting thoough that is (and I can't wait to play it), this is way away from DMP. We lead 4-2 Crawford hence my Crawford, -3 summary of the match score.

The key factor here is that gammons don't count.  The only thing that counts is whether we win the game or not.  If we win, the match is over.  If we lose, we play one more game for the match.  That makes it play exactly like DMP, and we should choose our checker play accordingly. 

Splitting both back men is good for gammon-save. 
Bringing down two builders is good at gammon-go or for money play.

At DMP, one of the hybrid moves is the way to go. 

The book's on generic openings though. At this score I'm sure it's 24 21 , 13 10 - it will be a small difference but I bet it's better than 24 20, 13 9 which will be second best.

We'll see.  I agree that they're close, but I'll stick with 24/20, 13/9 based on the Tom Keith site and my certainty that this plays like DMP.


BTW, Dorbel, are you sure you remember this correctly?  You might want to check your notes.

[close]

[diane]

Spoiler

The book's on generic openings though. At this score I'm sure it's 24 21 , 13 10 - it will be a small difference but I bet it's better than 24 20, 13 9 which will be second best.

[close]

Spoiler

far from it, the summary table at the bottom has the equity for each match status as well as move...the green tick for DMP is firmly in the 24/21, 13/9 box...

[close]

[dorbel]

Spoiler

Yes I am absolutely sure that I remember this correctly, having rolled it out some 10,000 times (Snowie 3-ply precise) during the project to supply data for openings for the Nack-Peever opus. It was clearly best at dmp. This was about five years ago, has later data supplanted it?
As far as I can recall, my opening chart having disappeared in my latest hard drive crash, this is also the least gammonish play, so would also be correct at Crawford, 2-away if I remember correctly.

[close]

[ah_clem]

Spoiler

Yes I am absolutely sure that I remember this correctly, having rolled it out some 10,000 times (Snowie 3-ply precise) during the project to supply data for openings for the Nack-Peever opus. It was clearly best at dmp. This was about five years ago, has later data supplanted it?
As far as I can recall, my opening chart having disappeared in my latest hard drive crash, this is also the least gammonish play, so would also be correct at Crawford, 2-away if I remember correctly.

I'm sorry to hear that you lost the data.  This is entirely from a self-interested perspective, not pathos; I would have liked to have seen it.  (c:

Basically, all I know about opening is on the Tom Keith page, and he has a little green checkmark in the column for DMP for 24/31 13/9 with the   24/21, 24/20 play in last place of the four considered - that's what caused me to question whether you remember your result correctly.  He also says that " the rollout is far from definitive", so take it for what it's worth.  The article is dated April 2006, so that would be newer than your 5 year old rollouts, but I wouldn't go so far as to say it "supplanted" your results.

The article doesn't say much about his methodology for doing the rollouts, so I'm a bit unsure about how they came about.  In particular, I'm not sure if he did a separate rollout for each match score (dmp, gammon-go, etc) or just did one rollout and compared win rates and gammon rates against a MET.  If it's the latter (and I suspect that it is) then the rollouts would not take into consideration differing checker play for different match scores and thus the results would be distorted.*

I'm doing my own rollout right now, but I'll withhold the results until the vote  closes.
I doubt it will be dispositive, neither should we take the Tom Keith page as the last word.

*UPDATE: it's the latter.  See http://www.bkgm.com/openings/rollouts.html

[close]

[roadkillbooks]

.0014 of a difference rolled out at 2ply is a statistical tie.  This is not a .01 error or a .05 blunder but a .0014 difference.  I like it though and do it all the time.  The advantage of dorvel´s play 24/20, 24/21 is that the relplies are often errors.  Knowing that it´s a good idea to go out and learn the correct replies to it so one isnt suckered.

roadkillbooks

[ah_clem]

gnubg rollout

Spoiler

Just what the world needs - another rollout of an opening position.  (c:

Dorbel's move comes out on top, but barely.  Forum choice is a very close second.

Bottom line here is that the differences are all within the margins of error; a more rigorous rollout might be more definitive, but all four reasonable plays are viable ways to play an opening of 4-3.   It's more a matter of taste as to which you choose.



    1. Rollout          24/21 24/20                  Eq.:  -0.014
       0.502 0.197 0.040 - 0.498 0.125 0.005 CL  -0.014 CF  -0.014
      [0.003 0.006 0.004 - 0.003 0.002 0.001 CL   0.006 CF   0.006]

    2. Rollout          24/21 13/9                   Eq.:  -0.020 ( -0.006)
       0.500 0.208 0.040 - 0.500 0.138 0.006 CL  -0.020 CF  -0.020
      [0.003 0.006 0.004 - 0.003 0.002 0.001 CL   0.006 CF   0.006]

    3. Rollout          13/10 13/9                   Eq.:  -0.026 ( -0.012)
       0.498 0.216 0.037 - 0.502 0.142 0.006 CL  -0.026 CF  -0.026
      [0.003 0.006 0.004 - 0.003 0.002 0.001 CL   0.006 CF   0.006]

    4. Rollout          24/20 13/10                  Eq.:  -0.027 ( -0.013)
       0.495 0.193 0.029 - 0.505 0.130 0.004 CL  -0.027 CF  -0.027
      [0.003 0.006 0.003 - 0.003 0.003 0.001 CL   0.005 CF   0.005]
        Full cubeful rollout with var.redn.
        1296 games, Mersenne Twister dice gen. with seed 872235963 and quasi-random dice
        Play: 0-ply cubeful prune [expert]
        Cube: 0-ply cubeful prune [expert]

[close]

[dorbel]

As you say, the problem with Tom Keith's tables is that the results are extrapolated from a rollout where the checker play is not made according to score. As the requirements of the checker plays are so very different in each column and as there is a whole game in which the inevitable errors will be magnified, I don't believe that this is a good method.
Ah Clem's considerably shorter rollout is at 0-ply. This is not really good enough IMO as at this level Gnu makes some pretty poor plays. I am replicating TK's 2-ply rollout, but rolling out according to score. Don't hold your breath while waiting for the answer, but with 922 of the 46,656(!) trials complete, 24/20, 24/21 is leading. I do agree that either of the split plays is very close, but I wouldn't play 13/10, 13/9 at this score. You do need to stay pure cubeless and it is so much easier to do that with a high anchor, which the split plays cater for of course.