checks in qsearch - final results

[bob]

results:

Code: Select all

Rank Name               Elo    +    - games score oppo. draws  ---------LOS----------
   1 Glaurung 2.1       181    5    5 23346   77%   -33   17%     100100100100100100100
   2 Fruit 2.1           69    5    4 23345   64%   -33   23%    0   100100100100100100
   3 opponent-21.7       27    4    4 23346   59%   -33   33%    0  0    99100100100100
   4 Glaurung 1.1 SMP    14    5    5 23346   56%   -33   21%    0  0  0   100100100100
   5 Crafty-22.2R3      -31    3    4 38910   43%    20   22%    0  0  0  0    55 99100
   6 Crafty-22.2R2      -31    4    4 38910   43%    20   22%    0  0  0  0 44    99100
   7 Crafty-22.2R1      -37    4    4 38909   42%    20   23%    0  0  0  0  0  0   100
   8 Arasan 10.0       -190    4    5 23346   29%   -33   19%    0  0  0  0  0  0  0   

R1 is normal crafty, normal (no check) q-search, null move R=2~3. R2 add qsearch checks, all else is identical. R3 changes null move R=3. So adaptive null move is now very slightly worse, and the difference is nearly unmeasurable. So, +7 elo (ignoring the error bar / uncertainty) which is surprising considering all the past comments I have seen. It is worth something it appears, but _not_ very much...

I may go back and resurrect some old stuff and test in this same way, to see what null-move threat extension does,, or the old "cheap singular extensions" and such, since I kept all of that old code.

Meanwhile we need to get back to dealing with the new eval and get everything converted to the "fruit interpolation" approach and re-enable all the code...