Is Glaurung 2.0.1 now the fastest mate solver?

[Tord Romstad]

I think 1.2.1 don't show correct mates always?

That depends on how you define "correct mates". Unless I have some unknown bug, Glaurung 1.2.1 will never display a mate score when there is no mate. However, I don't adjust mate scores from the hash table, so the number of moves to mate will not always be correct. I simply don't care about the number of moves to mate, and therefore don't bother to adjust hash table scores.

Question about Glaurung: what bitbases it is currently using

Only KP vs K. This bitbase is computed from scratch during program initialization, using a very inefficient algorithm. This is why Glaurung starts up somewhat slowly on old computers.

and can it use Scorpio 5 piece set may be!

It can't use the Scorpio bitbases, and it never will, because I didn't write the code. If I want to use bitbases or tablebases, I have to write the code for generating and probing them myself. It is possible that I will do this eventually, but it doesn't have very high priority at the moment.

Tord

[Tord Romstad]

Can you actually say what tricks make up a good mate-finder..

Are there any described algorithms?

Yes, there are, but I have never studied them. I simply don't care about mates, and optimizing my program to find them quickly has never been a priority.

I mean Null-move or similar isn't the key, so which part of the ordinary alpha beta search can be improved? Are there any good move ordering guidlines?

There is a very simple move ordering trick which most programs (including mine) uses: When a mate in 1 is found, this move is tried first (or directly after the move from the hash table, if such a move exists) at all sibling nodes. This sometimes speeds up the search considerably in positions where there is a forced mate, and usually has no measurable effect whatsoever in positions without a forced mate.

Tord