Stockfish Testing Queue

Pending - 0 tests 0.0 hrs

None

Active - 0 tests

Finished - 1350 tests

20-10-17 sg separate_captures2 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 21930 W: 3989 L: 4020 D: 13921
sprt @ 10+0.1 th 1 Don't separate captures only at 1/128 of the nodes.
20-10-17 sg separate_captures2 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 11911 W: 2066 L: 2141 D: 7704
sprt @ 10+0.1 th 1 Nearly the same as my first test but for some randomness use the node count instead position key (So the same position can be handled differently). I want compare which of both approaches seems better. Later i will try a lower fraction to effect less nodes.
20-10-17 sg separate_captures diff
LLR: -2.94 (-2.94,2.94) [0.00,5.00]
Total: 8187 W: 1403 L: 1493 D: 5291
sprt @ 10+0.1 th 1 Do on average at each 32th position separate not between good and bad captures in move sorting (so good and bad captures are tried together). Using position key for this decision to guarantee that the same position is always handled in the same way. I hope that so we have a better chance to find a bad capture which is in reality is good without hamper the search to much
19-10-17 sg pawn_material_move_tabl diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 9766 W: 1682 L: 1766 D: 6318
sprt @ 10+0.1 th 1 Use combined pawn and material key as index in the new move table
19-10-17 sg material_move_table diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 11425 W: 2021 L: 2098 D: 7306
sprt @ 10+0.1 th 1 Use now material instead of pawn key
19-10-17 sg pawn_structure_move_tab diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 5743 W: 958 L: 1059 D: 3726
sprt @ 10+0.1 th 1 Also update table if there not move is found for the position
18-10-17 sg pawn_king_structure_mov diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 20836 W: 3680 L: 3717 D: 13439
sprt @ 10+0.1 th 1 Use pawn structure and king positions as table key (bench fixed)
18-10-17 sg pawn_structure_move_tab diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 20961 W: 3674 L: 3711 D: 13576
sprt @ 10+0.1 th 1 Update table only if last move was a pawn move (Suggested by VoyagerOne)
18-10-17 sg pawn_king_structure_mov diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 13186 W: 2274 L: 2344 D: 8568
sprt @ 10+0.1 th 1 Update table only if last move was a pawn or king move
18-10-17 sg pawn_structure_move_tab diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 7698 W: 1330 L: 1423 D: 4945
sprt @ 10+0.1 th 1 Now use only own pawn structure as table key
18-10-17 sg pawn_structure_move_tab diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 8167 W: 1391 L: 1482 D: 5294
sprt @ 10+0.1 th 1 Introduce new pawn structure move hash table which saves best moves dependent on pawn key and use them for quiet move order. (Bug fixed)
18-10-17 sg weakQueen2 diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 14072 W: 2483 L: 2549 D: 9040
sprt @ 10+0.1 th 1 weakQueen penalty depended on if pinned piece is double defended or not
17-10-17 sg KBPPKB diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 43291 W: 7677 L: 7621 D: 27993
sprt @ 10+0.1 th 1 In the recent TCEC game between Laser and Booot a opposite bishop KBPPKB endgame arises which was draw. Laser could detect it but not Booot. Stockfish also can't detect the draw (score > 1). So add to the KBPPKB endgame function following case. If one pawn is on a or h file and promotion is covered by the king and the other pawn is covered by the bishop the position is draw. (Bench differs on my big bench)
17-10-17 sg KBPPKBPs diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 20335 W: 3576 L: 3616 D: 13143
sprt @ 10+0.1 th 1 Based on my KBPPKB patch extend this to handle any number of pawns for the weak side.
17-10-17 sg weakQueen diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 23988 W: 4235 L: 4259 D: 15494
sprt @ 10+0.1 th 1 Give no weakQueen penalty if bishop is pinned by bishop or rook is pinned by rook because in this cases we the pinned pieces can capture the pinner.
16-10-17 sg pawn_break diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 42225 W: 7678 L: 7623 D: 26924
sprt @ 10+0.1 th 1 Only midgame bonus
16-10-17 sg pawn_break diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 23489 W: 4136 L: 4162 D: 15191
sprt @ 10+0.1 th 1 Fixed version because i forget to check if the pawn can push. Only midgame bonus. Stopped other tests.
16-10-17 sg pawn_break2 diff
LLR: -1.18 (-2.94,2.94) [0.00,5.00]
Total: 19118 W: 3448 L: 3417 D: 12253
sprt @ 10+0.1 th 1 Half midgame bonus
16-10-17 sg pawn_break2 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 23124 W: 4206 L: 4232 D: 14686
sprt @ 10+0.1 th 1 Double midgame bonus
16-10-17 sg pawn_break diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 17986 W: 3215 L: 3264 D: 11507
sprt @ 10+0.1 th 1 Bonus for possible pawn break
13-10-17 sg LongRangedBishop diff
LLR: -2.94 (-2.94,2.94) [0.00,5.00]
Total: 21333 W: 3747 L: 3782 D: 13804
sprt @ 10+0.1 th 1 bonus +50%
13-10-17 sg LongRangedBishop3 diff
LLR: -3.14 (-2.94,2.94) [0.00,5.00]
Total: 9401 W: 1651 L: 1744 D: 6006
sprt @ 10+0.1 th 1 Use the full bonus
13-10-17 sg LongRangedBishop diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 21361 W: 3770 L: 3805 D: 13786
sprt @ 10+0.1 th 1 Half the bonus
13-10-17 sg LongRangedBishop diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 38854 W: 6861 L: 6823 D: 25170
sprt @ 10+0.1 th 1 Also smaller bonus if only one center square is seen which has an opponent pawn not defended by a pawn
13-10-17 sg LongRangedBishop2 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 13469 W: 2347 L: 2415 D: 8707
sprt @ 10+0.1 th 1 Also smaller bonus if only one center square is seen which has an own pawn which is not blocked
11-10-17 sg draw2 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 14635 W: 2585 L: 2648 D: 9402
sprt @ 10+0.1 th 1 Less reduction only at PV nodes
11-10-17 sg draw diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 14390 W: 2483 L: 2548 D: 9359
sprt @ 10+0.1 th 1 Less reduction only if best value >= alpha
11-10-17 sg draw diff
LLR: -3.09 (-2.94,2.94) [0.00,5.00]
Total: 8688 W: 1504 L: 1598 D: 5586
sprt @ 10+0.1 th 1 Less reduction if current best value is a draw. The most draw score are by repetition so the idea is to look closer at the other moves if detected draw.
09-10-17 sg tempo diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 29298 W: 5169 L: 5227 D: 18902
sprt @ 10+0.1 th 1 Refactor to use score instead of flat value for tempo. set temp to (40, 0). This doubles tempo for midgame and gives no tempo bonus in endgame. (bench fixed)
09-10-17 sg tempo2 diff
LLR: -3.37 (-2.94,2.94) [0.00,4.00]
Total: 11088 W: 1937 L: 2078 D: 7073
sprt @ 10+0.1 th 1 Try now for midgame same tempo value 20 as master but us zero for endgame
02-10-17 sg master diff
ELO: 32.61 +-1.6 (95%) LOS: 100.0%
Total: 40000 W: 6431 L: 2688 D: 30881
40000 @ 60+0.6 th 1 Regression test until "Good bishops on the main diagonals "
04-10-17 sg pvExactOrLower diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 85876 W: 15299 L: 15064 D: 55513
sprt @ 10+0.1 th 1 Additionally less reduction if lower bound with ttValue < beta
05-10-17 sg pvExact diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 19520 W: 2352 L: 2409 D: 14759
sprt @ 60+0.6 th 1 LTC: The version with ttValue in the window failed yellow. Currently Elberto test ttValue > alpha which currently struggles. So also try the opposite case of ttValue < beta
05-10-17 sg pvExact diff
LLR: 2.94 (-2.94,2.94) [0.00,5.00]
Total: 11628 W: 2112 L: 1940 D: 7576
sprt @ 10+0.1 th 1 The version with ttValue in the window failed yellow. Currently Elberto test ttValue > alpha which currently struggles. So also try the opposite case of ttValue < beta
04-10-17 sg bad_outposts diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 35156 W: 6310 L: 6286 D: 22560
sprt @ 10+0.1 th 1 No outpost bonus for bishop at a6,h6 which is defended by own pawn because no retreat is possible.
04-10-17 sg pvExactOrUpper diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 24702 W: 4299 L: 4321 D: 16082
sprt @ 10+0.1 th 1 Additionally less reduction if upper bound with ttValue > alpha
03-10-17 sg pvExact diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 61412 W: 10851 L: 10720 D: 39841
sprt @ 10+0.1 th 1 Apply pvExact only if ttValue lies in the search window
03-10-17 sg lmr_statScore diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 16924 W: 3010 L: 3109 D: 10805
sprt @ 10+0.1 th 1 Use ONE_PLY=2. Modify statScore reduction to allow half plies (with compensation) (Fixed bench)
01-10-17 sg cutNode2 diff
LLR: -2.94 (-2.94,2.94) [0.00,4.00]
Total: 44223 W: 7924 L: 7930 D: 28369
sprt @ 10+0.1 th 1 Use ONE_PLY=2. Two and a half more reductions at cut nodes. (Park with prio -1)
02-10-17 sg escape2 diff
LLR: -2.98 (-2.94,2.94) [0.00,4.00]
Total: 52455 W: 9318 L: 9299 D: 33838
sprt @ 10+0.1 th 1 Use ONE_PLY=2. Two and a half plies less reduction for escapes (Park with prio -1)
02-10-17 sg escape diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 35572 W: 6301 L: 6337 D: 22934
sprt @ 10+0.1 th 1 Use ONE_PLY=2. Only one and a half plies less reduction for escapes (Park with prio -1)
01-10-17 sg cutNode diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 32610 W: 5674 L: 5721 D: 21215
sprt @ 10+0.1 th 1 Use ONE_PLY=2. Only one and a half more reductions at cut nodes. (Park with prio -1)
01-10-17 sg mch diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 23775 W: 4166 L: 4242 D: 15367
sprt @ 10+0.1 th 1 Use ONE_PLY=2. Only half ply less reduction if last move was late (Park with prio -1)
01-10-17 sg ttCapture2 diff
LLR: -2.94 (-2.94,2.94) [0.00,4.00]
Total: 23035 W: 4057 L: 4135 D: 14843
sprt @ 10+0.1 th 1 Use ONE_PLY=2. One and a half plies reduction for ttCapture.
01-10-17 sg mch2 diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 9622 W: 1639 L: 1763 D: 6220
sprt @ 10+0.1 th 1 Use ONE_PLY=2. One and a half plies less reduction if last move was late (Park with prio -1)
01-10-17 sg ttCapture diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 10361 W: 1766 L: 1887 D: 6708
sprt @ 10+0.1 th 1 Use ONE_PLY=2. Only half ply reduction for ttCapture
01-10-17 sg unsupportable_pawns diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 25527 W: 4490 L: 4508 D: 16529
sprt @ 10+0.1 th 1 Doubling failed fast so try half bonus
01-10-17 sg unsupportable_pawns diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 6610 W: 1165 L: 1262 D: 4183
sprt @ 10+0.1 th 1 Double up supportable bonus
30-09-17 sg unsupportable_pawns diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 32767 W: 5811 L: 5798 D: 21158
sprt @ 10+0.1 th 1 Give now a bonus for supportable pawns instead penalty for unsupportable (suggested by Rocky)
30-09-17 sg pawns_rank_3 diff
LLR: -2.94 (-2.94,2.94) [0.00,4.00]
Total: 27687 W: 4935 L: 4997 D: 17755
sprt @ 10+0.1 th 1 The last failed strong yellow. So the eval has a slighlty problem with pawns at rank 3. Try now a decrease (-5,-5) of psqt for pawns at rank 3.