GothicChessInventor wrote:Perhaps in the near future we can test Gothic Vortex 2.2.5 against it in a casual 2-game match. I'm still working on the winboard piece, and I'll have to change the entire time control code for this.
Vortex was not designed for blitz chess. I deliberately have the branching factor "bubble" below ply 10, and "thin out" on plies 10 and up. This allows the program to consider more tactical and very aggressive lines at the root of the tree, then "go deep" to make sure they are sound.
The tradeoff: Too sacrificial at shallower depths, and not reflective of its true strength at longer controls.
I'd hate to have to #define some completely new evaluation function if it is just going to play blitz winboard matches.
I know that in normal chess I could not find a single case with significant number of games when the difference in rating between long time control and blitz is more than 100 elo.
Even finding 2 programs when one is significantly better in 40/4 and one is significantly better in 40/40 is an hard task
The best example that I could find is colossus2007b and frenzee
and even in that case I guess that the programmer of colossus fixed the bug that cause it to perform worse at long time control.
Colossus 2007b 2674 +20 −20 47.5% +16.9 30.1% 890 40/4
Colossus 2007b 2588 +37 −38 46.5% +23.5 31.0% 242 40/40
Frenzee 3.0 2604 −61 +24 −24 48.0% +14.2 27.1% 616 40/4
Frenzee 3.0 2666 +27 −27 51.4% −16.8 33.1% 459 40/40
Uri