[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: [Bug-gnubg] 0-ply (expert) Cube Decision Rollouts
From: |
Eberlein, Robert (REBERLEI) |
Subject: |
RE: [Bug-gnubg] 0-ply (expert) Cube Decision Rollouts |
Date: |
Mon, 5 May 2003 09:41:33 -0400 |
I can think of two ways to fix the problem:
1) Ideal way - Figure out a way to calculate
on-roll-but-I've-decided-not-to-double 0-ply equities. These would have a
smaller equity for "no double" than "double, take", whenever gnubg 0-ply
thinks doubling is correct.
2) Quick way - Make sure that the 0-ply rollouts turn the cube in these
cases. Either subtract 0.001 (minimum value that always shows a difference)
or an even smaller number, or just change the decision logic for the case
where the two equities match. That would restore the value of a 0-ply
cubeful rollout. The displayed value for a 0-ply "no double" cubeful
evaluation would still be wrong, but that is a very small problem,
comparatively.
I would favor option 2. That would make it easy to gather 0-ply rollout
data to implement option 1)
-----Original Message-----
From: Joern Thyssen [mailto:address@hidden
Sent: Sunday, May 04, 2003 3:43 PM
To: Eberlein, Robert (REBERLEI)
Cc: 'address@hidden'
Subject: Re: [Bug-gnubg] 0-ply (expert) Cube Decision Rollouts
On Thu, May 01, 2003 at 12:58:19PM -0400, Eberlein, Robert (REBERLEI) wrote
> I have found a bug in the 0-ply (expert) cube decision evaluation.
Whenever
> the 0-ply evaluation for an initial cube should be Double-Take, the result
> is coming back either No Double - Take, or Optional Double - Take, with
the
> equity for No Double always exactly matching (to the reported number of
> decimals) the equity for Double - Take. As a result, when a 0-ply cube
> decision rollout is performed, no double from a centered cube is ever
taken,
> and the rollout statistics show only drops in the centered cube case, with
> "cube efficiencies" (actually take rates) always 0. Because of this
massive
> market loss, the equity for the No Double case that the 0-ply rollout
> reports is very much too low.
The error lies in the way we use Janowski's formulae for cubeful
equities. Janowski's formulae produce on-roll equities, whereas we need
on-roll-but-I've-decided-not-to-double equities for cube decisions.
In your examples the equity for "no double" is equal to "double, take"
because it's a double on 0-ply.
At this point I'm not sure how to fix it.
I can add that this bug has no consequence for higher-ply evaluations or
cubeful rollouts with higher-ply cube decisions.
Jørn