Re: [Bug-gnubg] Problems using Stop rollout when one move appears to be best option

[Neil Robins]

I've just run it again with both boxes ticked and watched the rollout.
Absolutely identical result to what I had before. The move 13/9 6/5 is
outside the JSD limit after just over 1300 games and stays there but
continues to roll.
This is not a one off. All the three things I list that can happen are
commonplace in my experience. The feature worked differently in earlier
versions of GNU, but at least it once worked.
I would like to be able to be able to rollout my matches using this option
in conjunction with the CMark feature, but the unreliability makes that
impossible, from my point of view.

----- Original Message ----- 
From: "Christian Anthon" <address@hidden>
To: "Neil Robins" <address@hidden>
Cc: <address@hidden>
Sent: Saturday, August 22, 2009 9:14 PM
Subject: Re: [Bug-gnubg] Problems using Stop rollout when one move appears 
to be best option


> The code seems to work fine for me, but only if both kinds of
> stop-on-jsd is turned on. I'm checking in a simplification where there
> is only one kind.
>
> Please report back if the problems continue.
>
> Christian.
>
> On Sat, Aug 22, 2009 at 7:05 AM, Neil Robins<address@hidden> 
> wrote:
> > As I have previously said, there is often a problem with this when 
> > rolling
> > out several moves with this option. I am still encountering it using the
> > multithreaded code in the 20090817 Windows version.
> >
> > 1. A move that has come back within the set JSDs only sometimes not 
> > always
> > restarts rolling.
> > 2. A move that has gone way beyond the set JSDs sometimes continues to 
> > roll.
> > 3. The whole rollout can stop with some rolls neither rolled to maximum
> > number of trials nor outside the set JSDs.
> >
> > I think these problems happen generally when the plays don't fall out of
> > contention from the bottom of the list up. Stopping and restarting the
> > rollout with them reordered gets the right moves rolling again.
> >
> > The fourth move in the rollout below continued rolling another 1000 games 
> > or
> > more while above the 3.1 JSD limit set.
> >
> > GNU Backgammon Position ID: hucZAxDYzmDABg
> > Match ID : 8AmmAAAAIAAA
> > +13-14-15-16-17-18------19-20-21-22-23-24-+ O: gnubg
> > | X O O | | O X X O | 0 points
> > | X O O | | O X X O |
> > | O | | O |
> > | O | | O |
> > | | | |
> > v| |BAR| | 5 point match (Cube: 1)
> > | | | |
> > | | | |
> > | | | X |
> > | O X | | X X X | Rolled 41
> > | O X | | X X X O | 4 points
> > +12-11-10--9--8--7-------6--5--4--3--2--1-+ X: Neil
> >
> >
> > 1. Rollout 20/16 6/5 Eq.: +0.047
> > 0.532 0.192 0.027 - 0.468 0.153 0.005 CL +0.047 CF +0.047
> > [0.002 0.004 0.002 - 0.002 0.002 0.000 CL 0.003 CF 0.003]
> > Full cubeful rollout with var.redn.
> > 2592 games, Mersenne Twister dice gen. with seed 359697340 and
> > quasi-random dice
> > Stop when best play is enough JSDs ahead: limit 3.1 (min. 217 games)
> > Play: 0-ply cubeful [expert]
> > Cube: 0-ply cubeful [expert]
> > 2. Rollout 21/16 Eq.: +0.030 ( -0.017)
> > 0.523 0.177 0.021 - 0.477 0.153 0.004 CL +0.030 CF +0.030
> > [0.002 0.004 0.002 - 0.002 0.002 0.000 CL 0.003 CF 0.003]
> > Full cubeful rollout with var.redn.
> > 2331 games, Mersenne Twister dice gen. with seed 359697340 and
> > quasi-random dice
> > Stop when best play is enough JSDs ahead: limit 3.1 (min. 217 games)
> > Play: 0-ply cubeful [expert]
> > Cube: 0-ply cubeful [expert]
> > 3. Rollout 8/7 6/2* Eq.: +0.020 ( -0.027)
> > 0.518 0.179 0.017 - 0.482 0.154 0.003 CL +0.020 CF +0.020
> > [0.002 0.005 0.002 - 0.002 0.002 0.001 CL 0.005 CF 0.005]
> > Full cubeful rollout with var.redn.
> > 1486 games, Mersenne Twister dice gen. with seed 359697340 and
> > quasi-random dice
> > Stop when best play is enough JSDs ahead: limit 3.1 (min. 217 games)
> > Play: 0-ply cubeful [expert]
> > Cube: 0-ply cubeful [expert]
> > 4. Rollout 13/9 6/5 Eq.: +0.017 ( -0.030)
> > 0.515 0.172 0.016 - 0.485 0.132 0.003 CL +0.017 CF +0.017
> > [0.002 0.004 0.002 - 0.002 0.001 0.000 CL 0.003 CF 0.003]
> > Full cubeful rollout with var.redn.
> > 2592 games, Mersenne Twister dice gen. with seed 359697340 and
> > quasi-random dice
> > Stop when best play is enough JSDs ahead: limit 3.1 (min. 217 games)
> > Play: 0-ply cubeful [expert]
> > Cube: 0-ply cubeful [expert]
> > 5. Rollout 6/2*/1 Eq.: +0.010 ( -0.038)
> > 0.512 0.177 0.014 - 0.488 0.143 0.004 CL +0.010 CF +0.010
> > [0.004 0.010 0.004 - 0.004 0.004 0.001 CL 0.008 CF 0.008]
> > Full cubeful rollout with var.redn.
> > 426 games, Mersenne Twister dice gen. with seed 359697340 and
> > quasi-random dice
> > Stop when best play is enough JSDs ahead: limit 3.1 (min. 217 games)
> > Play: 0-ply cubeful [expert]
> > Cube: 0-ply cubeful [expert]
> > 6. Rollout 13/8 Eq.: +0.006 ( -0.041)
> > 0.509 0.180 0.012 - 0.491 0.118 0.003 CL +0.006 CF +0.006
> > [0.004 0.010 0.003 - 0.004 0.003 0.001 CL 0.008 CF 0.008]
> > Full cubeful rollout with var.redn.
> > 355 games, Mersenne Twister dice gen. with seed 359697340 and
> > quasi-random dice
> > Stop when best play is enough JSDs ahead: limit 3.1 (min. 217 games)
> > Play: 0-ply cubeful [expert]
> > Cube: 0-ply cubeful [expert]
> >
> > _______________________________________________
> > Bug-gnubg mailing list
> > address@hidden
> > http://lists.gnu.org/mailman/listinfo/bug-gnubg