[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Superko detection (was Re: [computer-go] M.Boon-Library)
GHI seems to be much more serious in GO than in chess. I am wondering
how it's normally handled in a GO program?
I have thought of a couple of "solutions" which I don't believe are
good enough:
1. Never use values in the hash table that do not match the depth
of the position you are attempting to score.
2. Do not store (or retrieve) positions where a ko is possible.
I don't know how much either of these help, but I'm sure they don't
"solve" the problem and I doubt there is a complete solutions that
doesn't cripple the benefits of the hash tables.
- Don
chrilly wrote:
>>Does MadLab's search use transposition tables? If so, how does it deal
>>with Graph-History Interaction?
>>Erik
> I have never really understand the point behind GHI. One can reach the same
> position over 2 different variations/graphs. But what bad playing behaviour
> is caused by GHI?
> Stupid side-Question: What is SuperKo? Is it general repetition of position
> and not just recapturing back one stone?
> If it is really general repetition and if GHI has some effects on it, than
> GHI would be also harmfull in chess. GHI seems to be a favorite topic of the
> Maastricht research-group. Maybe Erik can explain it.
If you store evaluation results for a position, then the result doesn't depend
on just the position itself, but also on its history.
For instance, position A is good for me because in some variation I can move to
the winning position B. So I store this information in a transposition table.
However, if in another part of the search, I reach position A *after* having
already reached B, then I cannot use the transposition table entry for A,
since A is not so good for me now that I am not allowed (by superko) to move to
B again.
regards,
-John
_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/
_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/