Search found 405 matches

by Chan Rasjid
Thu Jun 20, 2013 8:09 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: First post (and FailHigh question!)
Replies: 89
Views: 17723

Re: First post (and FailHigh question!)

Hello Sven, You get a hash hit when you are searching THAT position. And you get and backup the score of infi- 5. That's all. But in the parnet node, you will adjust the score to infi-5-12, infi-17, and that score is still above beta, so another cut-off occurs, and if a cut-off occurs you must store...
by Chan Rasjid
Tue Jun 18, 2013 11:45 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: First post (and FailHigh question!)
Replies: 89
Views: 17723

Re: First post (and FailHigh question!)

Hello Natale, Hi Chan, There is something which I still cannot understand with Natale's output. From what I now understand, at depth 20, the +mate score that search may have is limited to infi - 19 - 19, or infi - 38; ie MATE19. So I don't understand why the output has scores like MATE76. Chan, you'...
by Chan Rasjid
Tue Jun 18, 2013 4:12 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: First post (and FailHigh question!)
Replies: 89
Views: 17723

Re: First post (and FailHigh question!)

Hello Natale, I think I now understand why search (without QS) may see +mate score of mate-in-25(50 plies from the root) when it is still in the root iteration for a low depth of, say, 15. It is the working of chess position transposition in combination with TT hashing. If we have not encountered su...
by Chan Rasjid
Tue Jun 18, 2013 1:37 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: First post (and FailHigh question!)
Replies: 89
Views: 17723

Re: First post (and FailHigh question!)

Hello Natale, In this case we are at ply=26 and can get a hash hit (draft condition is OK and hash signature is OK) returning a score back that is 31979. At this point, we will adjust the score by subtracting 26 plies from 31979 ---> 31953 ---> which is a mate in 47 plies even if horizon is at ply 2...
by Chan Rasjid
Mon Jun 17, 2013 6:13 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: First post (and FailHigh question!)
Replies: 89
Views: 17723

Re: First post (and FailHigh question!)

Hello Natale, I see Sven too had just replied, but I only briefly browse through his post without raking my brain too much. Let me explain it more precisely... We already know that original position is mate in 11, and there's no doubt about it. So suppose a ply 22 search returned already that mate s...
by Chan Rasjid
Mon Jun 17, 2013 10:54 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: First post (and FailHigh question!)
Replies: 89
Views: 17723

Re: First post (and FailHigh question!)

Hello Natale, I first though it was impossible to have mate scores < MATE_IN_11 (less than 32000-21), but tonight (dusing sleep) I thought about it and I'm convinced it's possible. The "problem" is the hashtable. Low mate scores get overwritten over and over with scores less than MATE_IN_11, that's ...
by Chan Rasjid
Mon Jun 17, 2013 7:56 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: First post (and FailHigh question!)
Replies: 89
Views: 17723

Re: First post (and FailHigh question!)

Hi Natale, according to your report I am fully convinced that you have a very common bug which you need to fix as follows: Whenever you store a mate score "mated in M plies from root" or "mate in M plies from root" at ply N you need to store a value that specifies "mated (mate)" in M-N plies relati...
by Chan Rasjid
Sun Jun 16, 2013 9:29 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: First post (and FailHigh question!)
Replies: 89
Views: 17723

Re: First post (and FailHigh question!)

I agree, in case it is really free and really helps. From the Wiki Fail-Soft has the reputation for searching less nodes than Fail-Hard, but might also require some care regarding to search instability issues in conjunction with transposition tables and various pruning-, reduction- and extension te...
by Chan Rasjid
Fri Jun 14, 2013 9:42 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: First post (and FailHigh question!)
Replies: 89
Views: 17723

Re: First post (and FailHigh question!)

One more thing. The rule of thumb in debugging is "homing in" exactly what we are certain of. In your case, scores >= 650.00 can only be +mate scores. But in a research, it failed to find the mate, but found only a score of 600.00. So the bug is likely related to TT; error with storing hash scores o...
by Chan Rasjid
Fri Jun 14, 2013 9:29 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: First post (and FailHigh question!)
Replies: 89
Views: 17723

Re: First post (and FailHigh question!)

... I need to clarify certain things. I assume you use an aspiration window of 50cp as you mention so you do root search with [alpha, beta] where beta-alpha = 50 cp. Fail high with score 6.00 means your root search returned 6.00 >= beta and you did a research; but the score returned is again 6.0 as...