Henk wrote:Skipper was very lucky against Embla for Embla caused threefold repetition in a more than totally won endgame (two queens and a knight more). So Skipper should have been last place.
Probably the test set that Skipper uses (WAC) is worthless.
What depth skipper reach with the starting position with only material in eval ? 14 ?15?
Not much ... Isa is not bitboard and reach depth 17 with the minimal eval
Your bitboard is slow no ?
Depth 24 in about 2 seconds.
ok for first move
then , the depth fall to 16 , 15
(see the first game against Isa , saturday)
Your bitboard implementation is SLOW
Perhaps. Depth reached depends on position. Also is not that important. Just like nodes per second is.
Skippers first problem is that it should not crash.
Henk wrote:Skipper was very lucky against Embla for Embla caused threefold repetition in a more than totally won endgame (two queens and a knight more). So Skipper should have been last place.
Probably the test set that Skipper uses (WAC) is worthless.
What depth skipper reach with the starting position with only material in eval ? 14 ?15?
Not much ... Isa is not bitboard and reach depth 17 with the minimal eval
Your bitboard is slow no ?
Depth 24 in about 2 seconds.
ok for first move
then , the depth fall to 16 , 15
(see the first game against Isa , saturday)
Your bitboard implementation is SLOW
Perhaps. Depth reached depends on position. Also is not that important. Just like nodes per second is.
Skippers first problem is that it should not crash.
Here, I agree with Henk.
FIRST make it right. Then make it faster.
To know where the speed bottleneck is, the way to find out is to profile.
Of course, we can learn a lot about the move generator with perft.
Even more so by profiling perft.
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.
Henk wrote:Problem with WAC test is that counting material only gives good results on that test. So you think you are making progress. But positional play is important too.
WAC is called a tactical test suite for a reason, it is all about tactics. Which is certainly a critical part of a chess engine. If you overlook tactics, the best positional evaluation on the planet won't save you...
For positional adjustments, just use one of the many known positional test suites like the STS stuff.
Henk wrote:Skipper was very lucky against Embla for Embla caused threefold repetition in a more than totally won endgame (two queens and a knight more). So Skipper should have been last place.
Probably the test set that Skipper uses (WAC) is worthless.
What depth skipper reach with the starting position with only material in eval ? 14 ?15?
Not much ... Isa is not bitboard and reach depth 17 with the minimal eval
Your bitboard is slow no ?
Depth 24 in about 2 seconds.
ok for first move
then , the depth fall to 16 , 15
(see the first game against Isa , saturday)
Your bitboard implementation is SLOW
Not reaching a sufficient search depth is usually caused by a high effective branching factor, not by some slow implementation parts. You will never get 6 or 8 plies deeper by speeding up bitboard code etc.
Henk wrote:Skipper was very lucky against Embla for Embla caused threefold repetition in a more than totally won endgame (two queens and a knight more). So Skipper should have been last place.
Probably the test set that Skipper uses (WAC) is worthless.
What depth skipper reach with the starting position with only material in eval ? 14 ?15?
Not much ... Isa is not bitboard and reach depth 17 with the minimal eval
Your bitboard is slow no ?
Don't bother giving Henk any advice. He routinely ignores it or "misunderstands" it. Anybody who says that WAC is "useless" clearly isn't serious. That's my recommendation, anyway....
Thanks John but I'have noticed that too
bests
dany
