[Computer-go] Fwd: News on Tromp-Cook ?

[valkyria at phmp.se]

Quoting Erik van der Werf <erikvanderwerf at gmail.com>:

> I think Yizao's policy is the primary reason why Mogo got strong so
> fast in 2006.

Yes, I agree on this. In my opinion UCT itself got too much credit for  
the strength of program for a long time. As many has empirically  
experience (well at least I did) is that the parameter for UCT goes to  
0 when the policy goes strong.

The policy of Valkyria was never influenced by Mogo. It was already  
much more complex and covered more patterns and tactics when it was  
published. But I on the other hand had to rewrite the search code  
several times and tune the parameters before it really started to work  
well.

Best
Magnus