[computer-go] batch size and impact on strength

[Jason House]

On May 23, 2008, at 4:22 PM, Magnus Persson <magnus.persson at phmp.se>  
wrote:

> Quoting Carter Cheng <carter_cheng at yahoo.com>:
>
>> I remember seeing this in the archive before but I forget the  
>> actual  results of the experiment. Does processing simulations in  
>> groups of  say 5-10 have any impact on the strength of the program?
>>
>> Thanks in advance,
>>
>> Carter.
>
> Yes, at some point at least batch processing of simulations will be  
> inefficient.
> Every time you walk through the tree, the algorithm prefers to go  
> the best path possible to gain more information. With each extra  
> simulation in a batch there will be less and less useful  
> information. Maybe there is a tradeoff and you can get away with  
> small batches and gain something else. You have to test and see what  
> happens.

I expect batching to help when multithreading