[computer-go] Remarks on the Indigo-Paper

[Chrilly]

Dear Don,

>
> Sometimes in alpha/beta searching you can compare search results.  This
> was a real help when I did computer chess.
Yes, but as you know from Cilkchess parallel code is not deterministic any 
more. At least the Hydra search is not. This is a real pain in the ass. The 
only thing I can do is to make at least the single-version relative stable. 
Althought even there I detect sometimes bugs where I think afterwards: Why 
has the programm ever played one reasonable game. But alpha-beta is a big 
filter and one needs a lot of bugs or bad luck till it really produces 
remarkable results. Cock de Gorter once joked about Diep: Its a great 
programm, but it has too many bugs per second.

As a curiosity I won at the WC in 1993 a fantastic endgame against Rebel due 
to a bug. Nimzo sacrificied the last piece - a knight - and won the 
resulting pawn-endgame. Ed Schroeder tried for months to find a way to 
implement this in Rebel. In fact it was the white/black bug. I had the rule: 
If I have an unstopable passed pawn and the opponent has no piece to stop 
it, give a big bonus.
Due to the bug Nimzo checked for his own pieces and sacrificed the knight to 
get this high bonus. When I told this Ed at the next AEGON he was really 
relieved that no magic was behind this sacrifice. But it was the most 
brilliant bug I know off.

Apples and Trees:
I have not forgotten to answer you, but I am busy with programming and then 
I am coming in a strange mood. Its certainly some sort of addiction. One 
consequence is, that I am getting very narrow minded and "forget" everything 
what is not related to programming. Please be still a little bit patient 
with me.

Best Regards
Chrilly