bug-gnubg
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Réf. : RE: [Bug-gnubg] Analysis bug ?


From: Massimiliano . Maini
Subject: Réf. : RE: [Bug-gnubg] Analysis bug ?
Date: Fri, 10 Mar 2006 16:06:51 +0100

>On Fri, Mar 10, 2006 at 10:01:23AM +0000, Jonathan Kinsey wrote:
>> I've had a quick look and the code that decides if it's above TG or
below
>> DP uses the CashPoint function, this reuturns .632 as the cash point,
it's
>> about .657 to play on.
>
>If winning chances are above the cashpoint it must be a pass. Something
>must be wrong with the numbers here, though 0.632 is in the fully live
>to dead cube window. The logic seems to be that if a double is wrong,
>but cashable  it must be too good. Which should be okay, except for the
>fact that gnubg doesn't think it is cashable atall.
>
>Anyway a consistent view could be obtained by checking if the optimal
>cube action has a value below or above 1.0.

I agree. It is too good only if the equity is > 1.0

What surprises me is that the analysis/hint code gets it right (eq < 1.0
--> it's not too good, in the example, it wasn't even a double), while
the code that summarizes the statistics for the match/game gets it wrong,
as if it was doing his own computations instead of reusing the ones
already done (and stored in the analysis) ... or at least, the statistics
code is doing something different wrt the analysis/hint code ...

MaX.







reply via email to

[Prev in Thread] Current Thread [Next in Thread]