bug-gnubg
[Top][All Lists]

## Re: [Bug-gnubg] RE: Bug-gnubg Digest, Vol 57, Issue 6

 From: Robert-Jan Veldhuizen Subject: Re: [Bug-gnubg] RE: Bug-gnubg Digest, Vol 57, Issue 6 Date: Thu, 23 Aug 2007 17:21:21 +0200

On 8/22/07, bob koca <address@hidden> wrote:

>
> In money games, one can expect from doubling theory that with perfect play,
> 2/3 of initial doubles are takes (for redoubles, it is 1/2). Simulations
> with GNUBG 2-ply playing against itself indeed get very close to this
> number.
>

Do your 2/3 and 1/2 figures come from early-late ratios?

One could see this as the result from the application of these early/late ratios I suppose.

I don't see the jump from the ratios
to the statement that those fractions of doubles are takes. Could you explain please?

You're probably looking for a mathematical proof or something, which I haven't seen yet. I can only say that for me it seems intuitively sort of clear that these early/late ratios will lead to these take/pass ratios (at least approximately), under the assumption that equity and equity change distributions in backgammon are not too skewed (and maybe other assumptions as well). GNUBG 2-ply simulations support the idea that 2/3 and 1/2 are (approximately) the right numbers, which for me is quite convincing as far as real life backgammon is concerned. This subject was discussed on GOL once, and from what I remember there was consensus about this relation beween early/late and pass/take ratios.

If anyone knows of a more theoretical approach to this problem or comes up with different ratios, I'd be interested in it.

--
Robert-Jan Veldhuizen
(Zorba on FIBS)