TalkChess.com

Hi Michael,

Michael Sherwin wrote: ↑

Fri Apr 19, 2019 3:25 pm

(...) I do expect it to shine in the normal ab search though! We will see.

For it to shine in alpha / beta it will need to efficient at validating the hash move as legal and generating captures. Do you think this will be the case?

Cheers,

Steve

Hey Michael,

This looks interesting. How does it perform in perf tests?

BTW: if you'd like to write it up I'd be happy to publish it as a blog article on chessprogramming.net

Steve

It's interesting that Komodo seems to excel in this position. Clearly Mark and Larry have devised some clever logic. The position reminds me of the game Jonny vs. Komodo from the WCCC in 2015; a blocked game where Komodo was eventually out maneuvered. I wonder if Mark and Larry developed the heurist...

Yes - default settings with 8 Gb of hash and six threads.

If anyone is interested, here's the mate in 82 that Komodo 12.3 found: [FEN "8/1p1p2p1/5p1p/5P1P/1p1p1P2/pP1P1P2/P1k1B3/4K3 w - -"] 1. Bf1 Kc1 2. Bh3 Kc2 3. Ke2 Kc1 4. Bg2 Kc2 5. Bf1 Kc3 6. Ke1 Kc2 7. Be2 Kb1 8. Kd2 Kb2 9. Bd1 Kb1 10. Bc2+ Kb2 11. Kd1 Ka1 12. Kc1 Kxa2 13. Bd1 Ka1 14. Be2 Ka2 15. Kc2...

Amazingly Komodo 12.3 announces mate in 82 after 44 seconds on my relatively fast laptop (i9)! Note there wasn't any forward and backward analysis. I did a clean restart of the Shredder GUI. Here's the output: 8/1p1p2p1/5p1p/5P1P/1p1p1P2/pP1P1P2/P1k1B3/4K3 w - - Engine: Komodo 12.3 64-bit [6-Threads...

I am not sure I understand this. What information is in the lookup tables that would not be provided by the bitboard generation already? Is it that you store a lot of info on every move, that it pays to have an intermediate level of indirection through the pointers (so that later you would only hav...

I think that move generators of this type will always be slow. Because the overwhelming majority of nodes in a search tree will be QS nodes, where you are only need captures. And the lookup tables do not provide a way to directly skip to the end of a slider ray; you would still need to test the boa...

I've updated the download link. You can download from this page:

http://www.chessprogramming.net/the-baron/

Steve

A development version of SF (Jan 6th) running on my fast laptop finds Qd8 with a reasonable score. r5k1/1ppq1bp1/2n4r/1BBn1p2/p1NPp2p/P3P1P1/1PQ2P1P/2R3RK b - - Engine: Stockfish 060119 64 BMI2 [6 Thread] (2048 MB) by T. Romstad, M. Costalba, J. Kiiski, G. 38/73 1:15 0.00 1...Qe6 2.Bxc6 bxc6 3.Ne5 B...