Stockfish Testing Queue

Pending - 4 tests 15.6 hrs

24-07-16 aj tune_king_safety diff
3298/30000 iterations
6867/60000 games played
60000 @ 60+0.6 th 1 Tune king safety at LTC. Run at priority -2(for an empty framework). Use much lower ck values for parameters around the lingering.
24-07-16 SC double_reductions diff
LLR: 1.81 (-2.94,2.94) [0.00,5.00]
Total: 60751 W: 11891 L: 11541 D: 37319
sprt @ 10+0.1 th 1 And take 3: adjust history contribution to have a reasonable bench. I'll wait the results of these three before deciding whether to stop or go on.
24-07-16 lb tune diff
20459/40000 iterations
41332/80000 games played
80000 @ 20+0.2 th 1 tune relative weights of cm, fm, fm2
15-07-16 aj smp_thread_skip diff
LLR: -0.05 (-2.94,2.94) [0.00,5.00]
Total: 13651 W: 2070 L: 2022 D: 9559
sprt @ 5+0.05 th 23 Don't let threads deviate more than 6 PLY from main thread: STC (Profiling shows that this happens only at very high thread counts. At low threads counts this hardly makes a difference)

Active - 60 machines 265 cores 1.59M nps (422.08M total nps) 274 games/minute

Machine Cores MNps System Version Running on Last updated
fastgm
11
1.48 Windows 10 59 iid_low_depth 19 minutes ago
hyperbolic.tom
3
1.36 Linux 3.13.0-86-generic 59 iid_low_depth seconds ago
sunu
2
1.55 Linux 4.6.0-4.slh.3-aptosid-amd64 59 iid_low_depth seconds ago
anst
4
1.91 Windows 8 59 iid_low_depth seconds ago
lem
3
1.23 Linux 4.5.4-1-ARCH 59 iid_low_depth seconds ago
fastgm
4
1.43 Windows 8 59 iid_low_depth seconds ago
marrco
7
1.71 Windows 10 59 iid_low_depth seconds ago
mibere
3
1.16 Linux 3.16.0-4-amd64-4 59 iid_low_depth seconds ago
chris
3
2.19 Windows 10 59 iid_low_depth seconds ago
biffhero
1
1.28 Linux 3.16.0-4-amd64 59 iid_low_depth seconds ago
cw
5
0.72 Windows 7 59 iid_low_depth seconds ago
homyur
2
2.43 Windows 7 59 ft seconds ago
ako027ako
3
1.87 Windows 10 59 ft 5 minutes ago
mgrabiak
3
1.53 Windows 8 59 ft seconds ago
cw
3
1.17 Windows 7 59 ft seconds ago
fastgm
11
1.61 Windows 8 59 ft 19 minutes ago
mibere
14
1.55 Linux 3.16.0-4-amd64-2 59 ft seconds ago
slakovv
12
1.60 Linux 4.4.0-28-generic 59 ft seconds ago
JojoM
3
1.97 Linux 3.19.0-65-generic 59 ft seconds ago
Freja
3
1.74 Darwin 15.6.0 59 ft seconds ago
CSU_Dynasty
11
1.36 Windows 7 59 ft seconds ago
spams
3
2.43 Windows 7 59 ft 5 minutes ago
ttruscott
3
1.62 Windows 7 59 ft seconds ago
davar
3
1.86 Windows 8 59 ft seconds ago
stocky
3
2.27 Linux 3.13.0-87-generic 59 ft seconds ago
JBoyd333
3
1.67 Windows 8 59 ft seconds ago
Thanar
3
1.70 Windows 7 59 ft seconds ago
mhoram
2
0.76 Linux 3.16.0-4-amd64 59 ft seconds ago
ctoks
3
2.40 Windows 8.1 59 ft seconds ago
snicolet
3
1.66 Darwin 13.4.0 59 ft seconds ago
velislav
2
1.66 Linux 3.10.0-123.8.1.el7.x86_64 59 ft seconds ago
JojoM
1
1.97 Windows 10 59 ft seconds ago
mibere
7
1.39 Linux 3.16.0-4-amd64-3 59 ft seconds ago
CSU_Dynasty
7
0.87 Windows 7 59 ft seconds ago
cw
3
1.50 Windows 7 59 ft seconds ago
mibere
7
1.09 Linux 3.16.0-4-amd64-6 59 ft seconds ago
Bobo1239
1
0.97 Linux 4.6.4-1-ARCH 59 ft seconds ago
SC
3
2.19 Linux 4.2.0-41-generic 59 ft seconds ago
mibere
7
1.51 Linux 3.16.0-4-amd64-7 59 ft seconds ago
drabel
3
2.32 Windows post2008Server 59 ft seconds ago
robnjr
5
2.44 Windows 7 59 ft seconds ago
vdbergh
5
1.53 Linux 2.6.32-504.23.4.el6.x86_64 59 ft seconds ago
amicic
3
2.05 Windows 7 59 ft seconds ago
cw
3
1.38 Windows 7 59 ft seconds ago
sqrt2
3
1.93 Linux 3.16.0-4-amd64 59 ft seconds ago
mibere
5
1.65 Linux 3.16.0-4-amd64-1 59 ft seconds ago
psk
3
2.55 Linux 3.13.0-88-generic 59 ft seconds ago
VoyagerOne
5
2.28 Windows 10 59 ft seconds ago
marrco
7
1.50 Windows 2012Server 59 ft seconds ago
crunchy
7
1.39 Windows 7 59 ft seconds ago
mibere
7
0.95 Linux 3.16.0-4-amd64-5 59 ft seconds ago
Thanar
3
1.57 Windows 7 59 ft seconds ago
mibere
7
1.63 Linux 3.16.0-4-amd64-8 59 ft seconds ago
fastgm
15
1.76 Windows 10 59 lazy_high_density seconds ago
TheDarkHorde
2
1.65 Linux 4.6.4-1-ARCH 59 double_reductions seconds ago
ctoks
2
1.08 Windows 8.1 59 master seconds ago
modolief
1
1.71 Windows 8 59 master 2 minutes ago
cw
1
1.82 Windows 7 59 master 2 minutes ago
fp53fish
1
1.07 Linux 3.19.0-32-generic 59 master 3 minutes ago
IgorLeMasson
2
1.04 Windows 7 59 master seconds ago
25-07-16 Vo ft diff
LLR: -0.89 (-2.94,2.94) [0.00,5.00]
Total: 21681 W: 3048 L: 3009 D: 15624
sprt @ 60+0.6 th 1 Ltc futility margin tweak
25-07-16 ci iid_low_depth diff
LLR: -1.83 (-2.94,2.94) [0.00,5.00]
Total: 44932 W: 8636 L: 8515 D: 27781
sprt @ 10+0.1 th 1 Research in IID for ttMove at low depth.
24-07-16 mc master diff
ELO: 48.30 +-1.8 (95%) LOS: 100.0%
Total: 36003 W: 7203 L: 2230 D: 26570
40000 @ 60+0.6 th 1 Regression test
24-07-16 SC double_reductions diff
LLR: -1.48 (-2.94,2.94) [0.00,5.00]
Total: 59565 W: 11334 L: 11135 D: 37096
sprt @ 10+0.1 th 1 Take 2: rounding instead of converting.
21-07-16 pb lazy_high_density diff
ELO: 0.67 +-4.6 (95%) LOS: 61.2%
Total: 5162 W: 619 L: 609 D: 3934
10000 @ 20+0.2 th 15 As queue is almost empty, trying high_density at LTC with 15 threads to see if we have the same effect as we had on 7 threads. (Last try if the outcome is negative)

Finished - 12116 tests

24-07-16 Vo ft diff
LLR: 2.96 (-2.94,2.94) [0.00,5.00]
Total: 41192 W: 8053 L: 7742 D: 25397
sprt @ 10+0.1 th 1 futility margin tweak
24-07-16 tv mcpruning diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 11418 W: 2165 L: 2241 D: 7012
sprt @ 10+0.1 th 1 Take 2
24-07-16 SC double_reductions diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 8109 W: 1530 L: 1621 D: 4958
sprt @ 10+0.1 th 1 I have verified that using double-valued reductions array does not seem to slow down. So we could avoid rounding the Reductions array, such that we can sum with a double valued history contribution. Take 1.
24-07-16 tv mcpruning diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 13344 W: 2507 L: 2575 D: 8262
sprt @ 10+0.1 th 1 Less movecount pruning when static eval near eval
24-07-16 SC see_pruning_red diff
LLR: -2.94 (-2.94,2.94) [0.00,5.00]
Total: 3156 W: 535 L: 647 D: 1974
sprt @ 10+0.1 th 1 Increase reductions if SEE results is negative, but not enough to prune the move.
24-07-16 My bb3.1 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 7818 W: 1123 L: 1215 D: 5480
sprt @ 10+0.1 th 1 last was close so another try
24-07-16 lb test diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 82169 W: 15743 L: 15609 D: 50817
sprt @ 10+0.1 th 1 cutNodes: reduce only 1 extra ply
24-07-16 lb test diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 15787 W: 2975 L: 3077 D: 9735
sprt @ 10+0.1 th 1 capture escape: decrease reduction 1 ply only
23-07-16 My bb3 diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 47274 W: 4740 L: 4732 D: 37802
sprt @ 60+0.6 th 1 Really curious to see how this (and scaling stuff generally) goes at LTC low throughput.
23-07-16 lu mg_sf_openfiles diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 5857 W: 1096 L: 1197 D: 3564
sprt @ 10+0.1 th 1 Rough take 2 as queue is running out of tests...
23-07-16 Vo rmcr diff
LLR: -0.92 (-2.94,2.94) [0.00,5.00]
Total: 463 W: 75 L: 113 D: 275
sprt @ 10+0.1 th 1 Root Move Count Reduction idea...
23-07-16 SC fp_shallow diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 18954 W: 3603 L: 3646 D: 11705
sprt @ 10+0.1 th 1 Some variations: even less depth, even more pruning. Take 4 and call it a day.
23-07-16 Vo dangerSqs diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 22373 W: 4274 L: 4301 D: 13798
sprt @ 10+0.1 th 1 stc
23-07-16 lu mg_sf_openfiles diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 15799 W: 3033 L: 3089 D: 9677
sprt @ 10+0.1 th 1 Does scaling midgame scores make sense? Let's try with central open files... Low throughput.
23-07-16 SC fp_shallow diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 7479 W: 1379 L: 1473 D: 4627
sprt @ 10+0.1 th 1 Some variations: a quadratic formula with limit at predicted depth 10. Take 3.
23-07-16 SC fp_shallow diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 19975 W: 3855 L: 3893 D: 12227
sprt @ 10+0.1 th 1 Some variations: less depth, more pruning. Take 1.
23-07-16 SC fp_shallow diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 4256 W: 780 L: 888 D: 2588
sprt @ 10+0.1 th 1 Some variations: all depths, exponential margin. Take 2.
23-07-16 aj king_safety_tuned2 diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 16838 W: 2350 L: 2452 D: 12036
sprt @ 60+0.6 th 1 Pick only some stable values from previous tuning session. Especially those that involve attacks around the kingring: LTC
23-07-16 SC fp_shallow diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 9437 W: 1760 L: 1845 D: 5832
sprt @ 10+0.1 th 1 It has worked for SEE pruning. Perhaps it works also for futility.
23-07-16 My bb3 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 109431 W: 16663 L: 16380 D: 76388
sprt @ 10+0.1 th 1 last one
23-07-16 aj king_safety_tuned2 diff
LLR: 2.96 (-2.94,2.94) [0.00,4.00]
Total: 29503 W: 5964 L: 5696 D: 17843
sprt @ 10+0.1 th 1 Pick only some stable values from previous tuning session. Especially those that involve attacks around the kingring: STC
22-07-16 Vo yc diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 23212 W: 3164 L: 3204 D: 16844
sprt @ 60+0.6 th 1 LTC: Yellow Combo http://tests.stockfishchess.org/tests/view/574642a50ebc59301a3550e2 http://tests.stockfishchess.org/tests/view/573f52d00ebc59301a354f80
22-07-16 pb limit_skipsize diff
ELO: -1.11 +-4.0 (95%) LOS: 29.2%
Total: 10000 W: 1695 L: 1727 D: 6578
10000 @ 5+0.05 th 7 Limit ply skipping on high depths.
21-07-16 aj tune_king_safety diff
21264/30000 iterations
43260/60000 games played
60000 @ 60+0.6 th 1 Tune king safety at LTC. Run at priority -1(for an empty framework)
22-07-16 lu thread_skip_log diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 10233 W: 1670 L: 1752 D: 6811
sprt @ 10+0.1 th 3 May limiting thread depth deviation be useful with a low thread count? Final take 2 (priority: -1)
22-07-16 Vo yc diff
LLR: 2.95 (-2.94,2.94) [0.00,5.00]
Total: 19899 W: 3938 L: 3722 D: 12239
sprt @ 10+0.1 th 1 Yellow Combo http://tests.stockfishchess.org/tests/view/574642a50ebc59301a3550e2 http://tests.stockfishchess.org/tests/view/573f52d00ebc59301a354f80
22-07-16 aj king_safety_tuned diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 9875 W: 1346 L: 1431 D: 7098
sprt @ 60+0.6 th 1 Tuned values King safety parameters after 20K iterations: LTC
22-07-16 SC see_pruning diff
LLR: 2.97 (-2.94,2.94) [0.00,5.00]
Total: 15578 W: 2243 L: 2070 D: 11265
sprt @ 60+0.6 th 1 Bugfix (wrong sign). LTC.
22-07-16 II SMP diff
LLR: -0.15 (-2.94,2.94) [0.00,5.00]
Total: 609 W: 96 L: 100 D: 413
sprt @ 5+0.05 th 7 My 2 cents for better SMP.
22-07-16 aj king_safety_tuned diff
LLR: 2.96 (-2.94,2.94) [0.00,4.00]
Total: 7275 W: 1493 L: 1308 D: 4474
sprt @ 10+0.1 th 1 Tuned values King safety parameters after 20K iterations: STC
22-07-16 My bb2 diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 31347 W: 4851 L: 4855 D: 21641
sprt @ 10+0.1 th 1 Second go at scaling out opp bishop & pawns w 8moves bk
22-07-16 SC see_pruning diff
LLR: 2.96 (-2.94,2.94) [0.00,5.00]
Total: 35366 W: 7011 L: 6724 D: 21631
sprt @ 10+0.1 th 1 Bugfix (wrong sign).
22-07-16 SC see_pruning diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 15616 W: 2938 L: 3041 D: 9637
sprt @ 10+0.1 th 1 In step 13 we prune all bishop for knight exchanges because of see pruning. Try to avoid this.
22-07-16 SC see_pruning diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 513 W: 54 L: 180 D: 279
sprt @ 10+0.1 th 1 Another variation: see pruning at all depths but with an increasing threshold.
21-07-16 II rf_new diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 48135 W: 9238 L: 9151 D: 29746
sprt @ 10+0.1 th 1 A local test was +3.4 Elo with 5000@5+0.05.
21-07-16 lu thread_skip_log diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 12673 W: 2103 L: 2175 D: 8395
sprt @ 10+0.1 th 3 May limiting thread depth deviation be useful with a low thread count? (fixed bench)
21-07-16 Ro Outpost6 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 8258 W: 1523 L: 1613 D: 5122
sprt @ 10+0.1 th 1 Another outpost experiment
21-07-16 Vo cft-hp diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 15259 W: 2866 L: 2925 D: 9468
sprt @ 10+0.1 th 1 Use cft to assist in history pruning. Based off cft.
21-07-16 Vo ttStats diff
LLR: -2.95 (-2.94,2.94) [-3.00,1.00]
Total: 12427 W: 2298 L: 2479 D: 7650
sprt @ 10+0.1 th 1 Don't update stats for ttMoves...
21-07-16 aj tropsim_param_tweak diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 10147 W: 1894 L: 2017 D: 6236
sprt @ 10+0.1 th 1 Running a param tweak for Snicolet. : STC
21-07-16 aj tropism2 diff
LLR: -2.94 (-2.94,2.94) [0.00,5.00]
Total: 2558 W: 495 L: 613 D: 1450
sprt @ 10+0.1 th 1 Compute tropism only when ei.KingAttackersCount[Them] = 0 : STC
21-07-16 pb clearance_history diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 9833 W: 1838 L: 1921 D: 6074
sprt @ 10+0.1 th 1 As queue is almost empty, trying a clearance stats on top of passed cft
21-07-16 aj tropism diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 1442 W: 241 L: 364 D: 837
sprt @ 10+0.1 th 1 Move tropism code inside evaluate_king: STC See commit notes for some implications
20-07-16 pe tm diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 43587 W: 6061 L: 6090 D: 31436
sprt @ 60+0.6 th 1 Direct LTC. increse time per move by 5%. Low throughput
20-07-16 aj thread_history diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 14819 W: 2447 L: 2511 D: 9861
sprt @ 5+0.05 th 7 Use history and countermoves from main thread. But they don't share it: STC (fixed hash setting)
21-07-16 Vo cftStats diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 18675 W: 3458 L: 3550 D: 11667
sprt @ 10+0.1 th 1 Adjust some parameters for the new cft table...test against cft.
12-07-16 pb lazy_big_map diff
ELO: -1.28 +-4.5 (95%) LOS: 28.8%
Total: 5140 W: 569 L: 588 D: 3983
10000 @ 20+0.1 th 31 Re-spin at 20+0.1, see pull #727
20-07-16 II delta_new diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 9058 W: 1689 L: 1776 D: 5593
sprt @ 10+0.1 th 1 The second last try with this idea.
12-07-16 Ma timesimplification diff
LLR: -0.62 (-2.94,2.94) [-3.00,1.00]
Total: 128000 W: 26838 L: 27117 D: 74045
sprt @ 15 th 1 Test time management at sudden death time controls.
20-07-16 aj thread_history diff
LLR: 0.68 (-2.94,2.94) [0.00,5.00]
Total: 1505 W: 263 L: 229 D: 1013
sprt @ 5+0.05 th 7 Use history and countermoves from main thread. But they don't share it: STC