Search found 150 matches
- Thu Aug 03, 2017 8:52 pm
- Forum: Computer Chess Club: Programming and Technical Discussions
- Topic: Black magic bitboards
- Replies: 21
- Views: 8954
Re: Black magic bitboards
Also I don't understand why you have more magic factors to chose from!? I was wrong with this. I verified that the hit rate for testing random numbers for beeing black or white magics is about the same. Nevertheless black magics give more compact lookup tables than white ones. I'll give more detail...
- Thu Aug 03, 2017 3:24 pm
- Forum: Computer Chess Club: Programming and Technical Discussions
- Topic: Black magic bitboards
- Replies: 21
- Views: 8954
Black magic bitboards
In my engines I use magic bitboards with a fixed shift of 12 bits for rooks and 9 bits for bishops. The lookup tables for each square have many unused entries. To save space, they are all combined overlapping in one big table. So the lookup tables for one square can use the unused entries of the loo...
- Tue Jul 11, 2017 7:55 pm
- Forum: Computer Chess Club: Programming and Technical Discussions
- Topic: Looking for dense magics
- Replies: 3
- Views: 2813
Re: Looking for dense magics
Basically it is just more patience. Some tricks may help: - Look for magics with few bits set, with most bits set or with regions where only few bits are set and other regions with most bits set. - When you have found two similar magics, also try all magics between them, for example with 11000010 an...
- Tue Feb 28, 2017 8:08 pm
- Forum: Computer Chess Club: Programming and Technical Discussions
- Topic: Move ordering statistics
- Replies: 8
- Views: 2593
Re: Move ordering statistics
Move ordering in an ALL-node is irrelevant. You will always search all of them anyway. I read that many times here and I always thought it is true. But now I think it is not. It is true by definition. An ALL node is a node where you search all moves because none of them cause a cut-off. It is true ...
- Tue Feb 28, 2017 6:31 pm
- Forum: Computer Chess Club: Programming and Technical Discussions
- Topic: Move ordering statistics
- Replies: 8
- Views: 2593
Re: Move ordering statistics
I read that many times here and I always thought it is true. But now I think it is not. It should be advantageous to keep moves together, that are refuted by the same killer move.Evert wrote:Move ordering in an ALL-node is irrelevant. You will always search all of them anyway.
- Tue Feb 28, 2017 6:24 pm
- Forum: Computer Chess Club: Programming and Technical Discussions
- Topic: Sorting losing captures ?
- Replies: 7
- Views: 1712
Re: Sorting losing captures ?
Long ago I tried the following. I am not claiming it is the best :). Deuterium Version 09.01.26.492 (21Dec09) [...] 8. Added capture killer moves, these are lossing capture moves that produces a cutoff. [...] Making captures killers works for me. The only captures that are excluded from kil...
- Mon Jan 16, 2017 9:06 pm
- Forum: Computer Chess Club: Programming and Technical Discussions
- Topic: Move ordering ?
- Replies: 6
- Views: 2683
Re: Move ordering ?
When you don't have a TT-move you can sort the move that captures the last moved piece with the least valuable attacker before the normal MVV-LVA.
- Sat Jan 14, 2017 6:33 am
- Forum: Computer Chess Club: Programming and Technical Discussions
- Topic: On-line engine blitz tourney January
- Replies: 41
- Views: 8153
Re: On-line engine blitz tourney January
Are you sure you did not accidentally login both under the name rpiArminius? If a registered player logs in while already logged in, the existing login is disconnected. Yes, I am sure. I don't remember the password for rpiArminius and that password is not in the scripts on the non rpi Arminius side...
- Thu Jan 12, 2017 7:40 pm
- Forum: Computer Chess Club: Programming and Technical Discussions
- Topic: On-line engine blitz tourney January
- Replies: 41
- Views: 8153
Re: On-line engine blitz tourney January
After having played some games Arminius vs. various engines with xboard 4.8.0 I also tried playing Arminius vs. rpiArminius. This worked for some moves but then the game was adjourned. Now I cannot resume the match. **** rpiArminius has arrived - you can't both be logged in. **** Disconnected from i...
- Thu Jan 12, 2017 6:28 pm
- Forum: Computer Chess Club: Programming and Technical Discussions
- Topic: On-line engine blitz tourney January
- Replies: 41
- Views: 8153
Re: On-line engine blitz tourney January
I use icsdroneng. I think I need to use -feedbackCommand kibitz for that. The default of that switch is whisper. I use icsdroneng with -feedbackCommand kibitz too with rpiArminius but the only kibitz from Arminius I see in the logfile is "This is Arminius 2017-01-01". Is it necessary to implement k...