it's been quite a while now since I last touched my engine. The actual course of action was that I worked on it for about three months during the spring of 2006, participated at the Chess960 tournament at Mainz 2006 and then hardly continued to work on it at all, up until now.
Now, the fascination of chess programming caught me again and I've decided to continue.
What had always got to me back then is that my engine didn't see to crash and didn't seem to have search errors, but it just didn't seem to be able to get deep into the search tree.
Now I've begun my work by playing a couple of games against it whilst collecting some debug data and I'm not really able to make sense of them...
E.g. I suspect that I'll suffer from some sort of tree explosion, but then, the debug info says I'm only spending around 30-40% percent in quiescence search. Below I have to debug outputs, each in a 5 min blitz game, one after 1.d4 and one after 1.e4
My engine gets one ply deeper after 1.d4, and since 1.d4 positions are more quiet, I began to suspect the quiescence search...
Code: Select all
2.961.291<--1:info score cp -13 depth 9 nodes 653109 time 8845 pv d7d5 c1f4 g8f6 b1c3 c8f5 c3b5 b8a6 g1f3 c7c5
...
2.961.896<--1:info string Transposition Table: [Hits 21409 (3%) | Moves 76201 (11%) | IID calls: 54 (0%)]
2.961.943<--1:info string Evaluation: [Calls 640061 | Cache hits 210738 (32%)]
2.961.977<--1:info string Pawn evaluation: [Calls 429323 | Cache hits 394450 (91%)]
2.962.032<--1:info string Null move: [Tries 5582 | Cutoffs 2968 (53%) | Mate 0 (0%)]
2.962.094<--1:info string Reductions: [LMR 6069 (0%) | Futility 4067 (0%) | Razoring 109 (0%)]
2.962.162<--1:info string Quiescence Nodes: 78382 (11%)
2.962.235<--1:info string Cutoff rate: [1.22]
3.035.625<--1:info score cp -16 depth 7 nodes 309961 time 5039 pv d7d5 e4d5 d8d5 b1c3 d5e5 f1e2 c8f5
...
3.036.413<--1:info string Transposition Table: [Hits 12450 (2%) | Moves 61628 (11%) | IID calls: 90 (0%)]
3.036.453<--1:info string Evaluation: [Calls 513649 | Cache hits 215360 (41%)]
3.036.505<--1:info string Pawn evaluation: [Calls 298289 | Cache hits 267693 (89%)]
3.036.590<--1:info string Null move: [Tries 3772 | Cutoffs 1783 (47%) | Mate 0 (0%)]
3.036.664<--1:info string Reductions: [LMR 4013 (0%) | Futility 4067 (0%) | Razoring 116 (0%)]
3.036.734<--1:info string Quiescence Nodes: 63502 (12%)
3.036.805<--1:info string Cutoff rate: [1.17]
I'd appreciate some opinions. I recall that this board also had some keen testers. If someone'd like to check out on their own, I've put the source on the web, as well as a Windows executable.
http://www.koenig-nied.de/tmp/Kurt_271009.zip
Cheers to all of you.
Feels good to be back and learn (-:
