low hanging fruit

Discussion of chess software programming and technical issues.

Moderators: hgm, Dann Corbit, Harvey Williamson

flok

Re: low hanging fruit

Post by flok »

matthewlai wrote:
flok wrote:
matthewlai wrote:Killer moves will give you significant speedup while still guaranteeing the same result (no tradeoff).
Do you mean https://chessprogramming.wikispaces.com ... +Heuristic ?
[/url]
Yeap!
Ok thanks!
What is written there looks so obvious I wonder why I did not think of this myself :-)
JVMerlino
Posts: 1352
Joined: Wed Mar 08, 2006 10:15 pm
Location: San Francisco, California

Re: low hanging fruit

Post by JVMerlino »

flok wrote:
JVMerlino wrote:Such a simple question, and yet it can be interpreted multiple ways. Do you mean nodes per second or time to depth? Do you JUST want this speedup or, more likely, would you like a strength increase as well? :)
It currently does 500-750k nodes/s. Not bad I think for a non-bitboard engine. But it takes days to reach eg depth 20. Depth 10 takes 5,2 seconds. So that probably can be improved.
Probably the easiest way to increase time-to-depth and HOPEFULLY get a strength increase as well is to implement reductions and pruning. But be forewarned, because you'll be tuning them forever.
What kind of red/pru are you thinking of?
These can get you started:
http://chessprogramming.wikispaces.com/ ... Reductions
http://chessprogramming.wikispaces.com/Futility+Pruning

Implementations vary greatly among engines based on how aggressively or conservatively they reduce/prune, so YMMV.

jm
flok

Re: low hanging fruit

Post by flok »

Thanks!