Yes a had a little more of a think about how much precision I would
actually need, and you're right. I will give this parallelisation a
go. Thanks for suggesting a totally different way to go about the
problem. In any case it was good to learn how trees can be implemented
in J.

On 3 March 2011 16:06, Devon McCormick <[email protected]> wrote:
> Note that for (!13), exact probabilities would give you better than nine
> digits of precision.  This would be helpful if you are differentiating
> between strategies that differ by only one part in a billion, but I've found
> that three or four digits are all you need for most practical distinctions.
>
> On Thu, Mar 3, 2011 at 10:48 AM, Justin Paston-Cooper <
> [email protected]> wrote:
>
>> I don't know whether this would be sufficient. It would be nice to
>> have exact probabilities, but I haven't really reasoned about the
>> space required. 13! is already 6227020800, and there's going to be a
>> table for each stage in each game. Ultimately I would like to use the
>> probabilities to play this game on-line, finding bets which go against
>> the calculated probabilities. I'd start worrying about the randomness
>> of the games.
>> ...
>>
> --
> Devon McCormick, CFA
> ^me^ at acm.
> org is my
> preferred e-mail
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm
>
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to