game 4 - move 1 -forum to play 21

[stog]

game 4 - move 1 -forum to play 21
let me know of any illegal options or if u need another option adding

[vegasvic]

Spoiler

early in the match i like action so the play would be 13/11 6/5 for moi. as we know 24/23 13/11 is also ok.

[close]

[stog]

4211vote so will leave for a few hours more= see if any of us shift our votes?

[stog]

blitzxz to play 63

[blitzxz]

Hmm. I don't remember what's the correct opening move here but I guess I just play the normal looking 24/18 13/10.

[blitzxz]

The correct reply to 1-2 slot with 6-3 is weird looking 24/21 13/7. I usually slot 1-2 opening and I haven't seen anyone ever to move that move. I can't understand how this move could be the best here, when it's clearly inferior as opening move. Differences are very small though. I will continue full 2/2 gnubg rollout to make cubeful equities statictically significant. Cubeless equities are already.

1. Rollout          24/21 13/7                   Eq.:  -0,484
      0,438 0,132 0,010 - 0,562 0,182 0,014 CL  -0,248 CF  -0,484
     [0,002 0,002 0,001 - 0,002 0,002 0,001 CL   0,007 CF   0,015]
       Full cubeful rollout with var.redn.
       1296 games, Mersenne Twister dice gen. with seed 1734030735 and quasi-random dice
       Play: world class 2-ply cubeful prune [world class]
       keep the first 0 0-ply moves and up to 8 more moves within equity 0,16
       Skip pruning for 1-ply moves.
       Cube: 2-ply cubeful prune [world class]
   2. Rollout          24/15                        Eq.:  -0,498 ( -0,013)
      0,446 0,125 0,010 - 0,554 0,190 0,009 CL  -0,276 CF  -0,498
     [0,002 0,003 0,002 - 0,002 0,002 0,001 CL   0,006 CF   0,014]
       Full cubeful rollout with var.redn.
       1296 games, Mersenne Twister dice gen. with seed 1734030735 and quasi-random dice
       Play: world class 2-ply cubeful prune [world class]
       keep the first 0 0-ply moves and up to 8 more moves within equity 0,16
       Skip pruning for 1-ply moves.
       Cube: 2-ply cubeful prune [world class]
   3. Rollout          24/18 13/10                  Eq.:  -0,511 ( -0,026)
      0,439 0,131 0,014 - 0,561 0,192 0,012 CL  -0,297 CF  -0,511
     [0,002 0,003 0,003 - 0,002 0,002 0,002 CL   0,007 CF   0,014]
       Full cubeful rollout with var.redn.
       1296 games, Mersenne Twister dice gen. with seed 1734030735 and quasi-random dice
       Play: world class 2-ply cubeful prune [world class]
       keep the first 0 0-ply moves and up to 8 more moves within equity 0,16
       Skip pruning for 1-ply moves.
       Cube: 2-ply cubeful prune [world class]
   4. Rollout          24/21 24/18                  Eq.:  -0,522 ( -0,037)
      0,438 0,116 0,006 - 0,562 0,189 0,008 CL  -0,287 CF  -0,522
     [0,002 0,003 0,002 - 0,002 0,002 0,001 CL   0,008 CF   0,017]
       Full cubeful rollout with var.redn.
       1296 games, Mersenne Twister dice gen. with seed 1734030735 and quasi-random dice
       Play: world class 2-ply cubeful prune [world class]
       keep the first 0 0-ply moves and up to 8 more moves within equity 0,16
       Skip pruning for 1-ply moves.
       Cube: 2-ply cubeful prune [world class]

[dorbel]

Splitting to the bar in response to a slot is generally wrong, as it gives the starter extra hit and cover numbers.
Slotting a point of your own on the other hand often works well as here, but it is rare to see. I have been running 24/15 with this, thanks for the tip. best to keep this sort of thing to yourself!

[roadkillbooks]

This is a very complcated position indeed.

i rolled it out in 1296 games.. i now lost those rollouts because i extended them to 2592..


the results switched..

the top 4 at 1296 were

24/21, 24/18
24/21, 13/7  -.002
24/15          -.005
24/18, 13/7

they were all close but the 4th was -.016 less
now with 2592  i have:

24/18, 13/10
24/15    -.007
24/21, 24/18  -.017
24/21, 13/7 -.019

now illl extend it again and let you know.. im gonna post this though for fear of losing it

[roadkillbooks]

now with 5184 games rolled out it's

24/15
24/18, 13/10 -.001
24/21. 24/18 -.012
24/21, 13/7   -.020

so essentially it switched again but do to the tendency i like both 24/15 and  24/18, 13/10.. the advantage of the other play however is that people wont be used to it and may inturn blow it so i may just play one of the other 2

[blitzxz]

What are your rollout settings? And what is the bot you're using? What are the standard errors?   I want to be sure this is rolled out with very high settings to be sure of the results. You must have superfast computer or you're using some kind of lower settings then I was.

[blitzxz]

I continued the rollout to 2500 games and there was still no difference at all between the top two plays, 24/15 and 24/21 13/7. Even the fourth best play wasn't inside 95 % confidence. Then I accidentaly closed my computer and lost the results.  So basically I didn't get any results what's the best play. The top moves are just so close to each other. But to me this is interesting result in itself.

[dorbel]

I rolled this out 46656 times using Gnu 0-ply. 24/15 came out best, but 24/21, 13/7 was only 0.005 behind. These plays are essentially equal and superior by a small but clear amount to any play that involves splitting to the opponent's bar. 24/15 wins  44.9% with 10.1% gammons, while 24/21, 13/7 wins 44.1% with 11.4% gammons. Gammon losses are almost the same, 17.1% for 24/15, 17.2% for 24/21, 13/7. On this information 24/15 is clearly best at dmp and gammonsave, 24/21, 13/7 is clearly best at gammongo and either play will do for money or level match.