Game 2, move 3: Forum 5-5

[Zorba]

sHPGAQPE58IBJA:cIm2ABAAAAAA

Looks like I found the "FIBS dice" generator in GnuBG! 

[diane]

Spoiler

ugh, I made my choice more based on what will move, rather than anything else.  I don't want to move anything off the 13, because it isn't adding anything to our position - which pretty much just leaves the 6-1[2], which I hate, but it is a point on the board! 8-3, 9-4, 6-1[2]

[close]

[ah_clem]

Spoiler

I voted for 13/8 9/4 8/3(2) , which is the same thing as 13/3 8/3 9/4 (which I think is the usual notation.)

Running the backmen with a high doubles is best, but alas they can't move.

The next best choice is to safety the blot on the 9. and make the three point.  So far, so good.  What to do with the other two 5s?  The choice is between stripping the midpoint with 13/3 or  burying a pair with 6/1.  Neither is particularly inviting, and I'm not sure which is better.  I chose 13/3, but I'm perfectly ok with 6/1(2).

[close]

[socksey]

Ditto, ah_clem, on this one (at least, so far). 

socksey



"Right is right, even if everyone is against it; and wrong is wrong, even if everyone is for it." - William Penn

[roadkillbooks]

Good move.  Thats what I would have picked.  Why not make the one point?  Its double fives early in the game.  Slotting it when you can make it cant be right.  Pilling extra pieces on the 3 point can only be useful if you are going to make the one or two point.  This position looks solid and relatively balanced.

I say NO DOUBLE and let the dice roll on my move.

[blitzxz]

Spoiler

I would have doubled already. We're about to crash complitely and doubled gammon would go for victory. There's a big chance that markets are lost with the next change of rolls. I'm not even sure was it a take. Propably but I wouldn't liked it.

[close]

[roadkillbooks]

Spoiler

I would have doubled already. We're about to crash complitely and doubled gammon would go for victory. There's a big chance that markets are lost with the next change of rolls. I'm not even sure was it a take. Propably but I wouldn't liked it.

[close]

With almost any of the other moves I think Id double, not with this one.  When you say you would have doubled already do you mean now or before?  Now forum has a 3 point board i have a 2 point board.  Albeit I have both 5 points but im also done in the race and of course I have better position.  I think if forum didnt have the 3 point board Id double.  Cant really say more about my reasons without giving away too much.  I can of course use the doubled gammon but so can the forum and they'd have a sweet recube of a premature double.  Plus I know that the forum would take here..just by following the arguements from my last double...so i might as well wait a move or two.  IF it becomes too good then Ill play for gammon...if i blow it and the forum turns it around..I can then drop and be tied 1-1.  Anyway...I cant wait for my dice...if I dont see them soon itll have to wait until 3am my time.

[blitzxz]

With almost any of the other moves I think Id double, not with this one.  When you say you would have doubled already do you mean now or before?  Now forum has a 3 point board i have a 2 point board.  Albeit I have both 5 points but im also done in the race and of course I have better position.  I think if forum didnt have the 3 point board Id double.  Cant really say more about my reasons without giving away too much.  I can of course use the doubled gammon but so can the forum and they'd have a sweet recube of a premature double.  Plus I know that the forum would take here..just by following the arguements from my last double...so i might as well wait a move or two.  IF it becomes too good then Ill play for gammon...if i blow it and the forum turns it around..I can then drop and be tied 1-1.  Anyway...I cant wait for my dice...if I dont see them soon itll have to wait until 3am my time.

5-5 was one of the worst rolls in my mind.   I meant I would have doubled now. If you know opponent will take that's no way to tell you don't want to double. It may sound that I'm judging your choice but I'm just telling how I would have done.  

[blitzxz]

Quick 0-ply rollout.And I seem to be wrong as usual. 

Rollout details:
Centered 1-cube:
  0,633 0,189 0,011 - 0,367 0,094 0,003 CL  +0,421 CF  +0,687
[0,002 0,002 0,001 - 0,002 0,001 0,001 CL   0,005 CF   0,015]
Player Forum owns 2-cube:
  0,637 0,192 0,013 - 0,363 0,095 0,003 CL  +1,067 CF  +0,644
[0,002 0,003 0,002 - 0,002 0,001 0,000 CL   0,013 CF   0,019]
Full cubeful rollout with var.redn.
1296 games, Mersenne Twister dice gen. with seed 3290832632 and quasi-random dice
Play: 0-ply cubeful [expert]
Cube: 0-ply cubeful [expert]

[socksey]

*Sigh*  " 'splain Lucy"!

socksey



"The way to get things done is not to mind who gets the credit for doing them." - Benjamin Jowett (1817-1893), British theologian and classicist

[Zorba]

move ROLLOUT:

Spoiler

It's best by quite a large margin to build the board. As rkb wrote, if you slot the onepoint, might as well make it. The preserved spare on the midpoint could come in handy later on. Blots in the homeboard behind your opponent's backcheckers may look harmless, but once the shot-hitting starts, they can be very dangerous.


    1. Rollout          9/4 8/3 6/1(2)               Eq.:  -0,6831
        36,56   9,51   0,36 -  63,44  18,76   1,35 CL  -0,4186 CF  -0,6831
      [  0,09   0,08   0,05 -   0,09   0,13   0,10 CL   0,0029 CF   0,0076]
    2. Rollout          13/3 6/1(2)                  Eq.:  -0,7572 ( -0,0741)
        34,00   9,29   0,44 -  66,00  18,11   2,52 CL  -0,4744 CF  -0,7572
      [  0,11   0,08   0,05 -   0,11   0,16   0,17 CL   0,0037 CF   0,0104]
    3. Rollout          13/3 9/4 6/1                 Eq.:  -0,8202 ( -0,1372)
        34,19   8,83   0,54 -  65,81  20,13   3,07 CL  -0,5048 CF  -0,8202
      [  0,12   0,08   0,06 -   0,12   0,17   0,18 CL   0,0040 CF   0,0109]

       Full cubeful rollout with var.redn.
        3888 games, Mersenne Twister dice gen. with seed 862504334 and quasi-random dice
        Play: 0-ply cubeful prune [expert]
        Cube: 0-ply cubeful prune [expert]

[close]

For Lucy  : Eq. is equity, the expected value of the game. Winning a single game is equity +1, losing it is -1, at the start of the game your equity is 0.

Here, -0.68 means you'll lose 68 points from here if you play it out 100 times, on average.

The second row numbers are percentages: games won total, (back)gammons won, backgammons won - games lost total, (back)gammons lost, backgammons lost.

The third row numbers in square brackets [ ] represent standard errors for all the values above it, similar to a standard deviation: a number representing the statistical inaccuracy range of the bot's rollout numbers. A quick way to interpret them is to multiply them by two (1.96 to be more exact) and add a +/- sign in front of it; you then have a so-called 95% confidence interval (CI).

Example:

The first move has equity -0.6831, the SE below it says 0.0076. The 95% CI is then -0.6831 +/- 0.0152 which is [-0.6983 to -0.6679]. It is 95% likely that the 'true value' of the rollout would be within this range. The more games you let GnuBG roll out, the smaller this error/range will become. It's similar to the margin of error you see in presidential polls, etc.

The rollout settings at the bottom are GnuBG stuff, I refer you to a good tutorial or manual for those. The simple answer is, don't worry about Mersenne Twister and variance reduction; the more games played, the better (you get smaller standard errors); 0-ply gives decent results in not too much time but 2-ply plays significantly better (alas more than ten times slower).

HTH 

[stog]

nice explanation Z thx
the blackberry and elderberry wine i was given this evening means that i shall have to revisit this post again however
thx (will forward to my bg chum (who gave me this vino gallactico and beat me 6 0 ce soir)

[Zorba]

roadkillbooks CUBE decision after our 5-5 (same rollout as blitzxz but higher settings):

Spoiler

It's close, but probably just a bit better to hold that cube for now.

Rollout details:
Centered 1-cube (NO DOUBLE):
   63,89  18,68   1,41 -  36,11   9,79   0,28 CL  +0,4290 CF  +0,7153
[  0,08   0,13   0,14 -   0,08   0,09   0,05 CL   0,0026 CF   0,0064]
Player FIBSBoard forum owns 2-cube (DOUBLE/TAKE):
   64,56  19,05   2,10 -  35,44   9,35   0,48 CL  +1,1233 CF  +0,6954 (-0,0199)
[  0,09   0,19   0,19 -   0,09   0,09   0,06 CL   0,0064 CF   0,0083]
Full cubeful rollout with var.redn.
3888 games, Mersenne Twister dice gen. with seed 862504334 and quasi-random dice
Play:  2-ply cubeful prune [world class]
keep the first 0 0-ply moves and up to 8 more moves within equity 0,12
Skip pruning for 1-ply moves.
Cube: 2-ply cubeful prune [world class]

[close]

[blitzxz]

Very close indeed. I think it's a lot about matter of taste and playing style. I like simple and aggressive style which means I double early when there are lots of gammons. I don't want to lose my markets in my favorite type of game. And I'm not very scared about the recube because often it happens that the opponent will redouble way too late in this score and lose much of his/her potential.