Tests

Active 6 Workers / 0 Training / 77 Test threads / 95.08 MNPS

Author	Engine	Identifier	Code	Time control	Results
Priority 0
Ellie	WakWak	main #8852 Active Error Duck960	diff ↗	N=5000	Generated 1632294/10000000 Games Elo: -0.37 +- 0.46 (95%) [N=10000000] Games: 1632294 W: 606845 L: 608592 D: 416857
nocturn9x	Heimdall	master #8814 Active Error	diff ↗	N=12000	Generated 2346518/5000000 Games Elo: -0.13 +- 0.36 (95%) [N=5000000] Games: 2346518 W: 780014 L: 780887 D: 785617
Tecci	WakWak	frac-tunables #8883 Active Duck960	diff ↗	40.0+0.40	LLR: -0.06 (-2.25, 2.89) [0.00, 4.20] Games: 494 W: 161 L: 162 D: 171 Ptnml(0-2): 12, 47, 130, 46, 12
remove the tuning stuff
Carbec	Zangdar	aspiration-fail-low #8881 Active	diff ↗	8.0+0.08	LLR: -0.42 (-2.25, 2.89) [0.00, 5.00] Games: 9270 W: 2359 L: 2340 D: 4571 Ptnml(0-2): 20, 1106, 2368, 1117, 24
7.46a + aspiration, fail low : beta calculé avec l'ancien alpha, avant de baisser alpha total nodes = 28827506
Priority -1
dan	Hobbes	main #7901 Active Error	diff ↗	N=12000	Generated 3435856/5000000 Games Elo: -0.06 +- 0.29 (95%) [N=5000000] Games: 3435856 W: 1103207 L: 1103829 D: 1228820
dan	WakWak	feature/cutnode-lmr #8831 Active Error Duck960	diff ↗	8.0+0.08	LLR: -0.35 (-2.25, 2.89) [0.00, 4.20] Games: 10726 W: 3801 L: 3779 D: 3146 Ptnml(0-2): 287, 911, 2928, 967, 270
dan	WakWak	feature/tt-alpha-negext #8828 Active Error Duck960	diff ↗	8.0+0.08	LLR: -0.61 (-2.25, 2.89) [0.00, 4.20] Games: 13156 W: 4673 L: 4654 D: 3829 Ptnml(0-2): 410, 1178, 3375, 1213, 402

Finished 25 shown

Author	Engine	Identifier	Code	Time control	Results
Carbec	Zangdar	probcut-cutnode #8880 Passed Error	diff ↗	8.0+0.08	LLR: 2.92 (-2.25, 2.89) [0.00, 5.00] Games: 10416 W: 2678 L: 2536 D: 5202 Ptnml(0-2): 27, 1178, 2661, 1310, 32
7.45b + ProbCut : l'enfant est cherché en !cut_node, comme les autres enfants d'un nœud de coupure total nodes = 28206940
Tecci	WakWak	frac-tunables #8875 Failed Duck960	diff ↗	8.0+0.08	LLR: -2.26 (-2.25, 2.89) [0.00, 4.20] Games: 25842 W: 9061 L: 9081 D: 7700 Ptnml(0-2): 798, 2412, 6487, 2460, 764
remove the tuning stuff
MasterStruct	Eunoia	feat/iir #8879 Failed	diff ↗	8.0+0.08	LLR: -2.26 (-2.25, 2.89) [0.00, 5.00] Games: 12978 W: 3897 L: 3943 D: 5138 Ptnml(0-2): 496, 1499, 2509, 1525, 460
Use different value
MasterStruct	Eunoia	feat/iir #8878 Finished	diff ↗	8.0+0.08	LLR: -0.90 (-2.25, 2.89) [0.00, 5.00] Games: 7868 W: 2378 L: 2382 D: 3108 Ptnml(0-2): 283, 887, 1560, 959, 245
Try different value
MasterStruct	Eunoia	feat/iir #8877 Finished	diff ↗	40.0+0.40	LLR: -0.19 (-2.25, 2.89) [0.00, 5.00] Games: 1206 W: 344 L: 347 D: 515 Ptnml(0-2): 38, 151, 232, 140, 42
Add Internal Iterative Reductions
MasterStruct	Eunoia	feat/iir #8876 Finished	diff ↗	8.0+0.08	LLR: 0.54 (-2.25, 2.89) [0.00, 5.00] Games: 18314 W: 5616 L: 5493 D: 7205 Ptnml(0-2): 650, 2136, 3498, 2187, 686
Add Internal Iterative Reductions
MasterStruct	Eunoia	feat/killer-heuristic #8872 Finished	diff ↗	8.0+0.08	LLR: -1.22 (-2.25, 2.89) [0.00, 5.00] Games: 19254 W: 5902 L: 5863 D: 7489 Ptnml(0-2): 701, 2228, 3708, 2311, 679
Clear killer table before searching
nocturn9x	Heimdall	qs-pv-fixes #7935 Finished Error	diff ↗	8.0+0.08	LLR: -0.61 (-2.25, 2.89) [-5.00, 0.00] Games: 3466 W: 833 L: 869 D: 1764 Ptnml(0-2): 13, 450, 850, 400, 20
Make qsearch keep track of PVs, avoid qsearching with invalid windows and don't cutoff QS in PV nodes (bench 5515662)
MasterStruct	Eunoia	feat/killer-heuristic #8871 Finished	diff ↗	8.0+0.08	LLR: -0.98 (-2.25, 2.89) [0.00, 5.00] Games: 3240 W: 970 L: 1002 D: 1268 Ptnml(0-2): 119, 371, 660, 363, 107
Add killer move heuristic
nocturn9x	Heimdall	[Potential] main #8870 Finished	diff ↗	40.0+0.40	Elo: 32.17 +- 16.65 (95%) [N=1000] Games: 444 W: 140 L: 99 D: 205 Ptnml(0-2): 2, 34, 107, 79, 0
so eren will stop sending heimdall blunder positions
Tecci	WakWak	frac-tunables #8856 Finished Error Duck960	diff ↗	40.0+0.40	Tuning 175 Parameters 5000/5000 Iterations 20000/20000 Games Played
paul	Byte-Knight	fix/needless-metadata-compute #8869 Passed	diff ↗	8.0+0.08	LLR: 2.97 (-2.94, 2.94) [-5.00, 0.00] Games: 9302 W: 2595 L: 2515 D: 4192 Ptnml(0-2): 134, 924, 2483, 948, 162
fix: don't recompute metadata if already set
Drex	Blini	q-search #8868 Passed Error	diff ↗	8.0+0.08	LLR: 2.93 (-2.25, 2.89) [0.00, 5.00] Games: 628 W: 218 L: 66 D: 344 Ptnml(0-2): 3, 33, 122, 121, 35
near checkmate scoring
MasterStruct	Eunoia	fix/relax-lmp-mate-score #8867 Finished	diff ↗	8.0+0.08	LLR: 1.69 (-2.25, 2.89) [-5.00, 0.00] Games: 15464 W: 4699 L: 4695 D: 6070 Ptnml(0-2): 391, 1629, 3699, 1611, 402
Relax LMP mate score
MasterStruct	Eunoia	main #8865 Passed	diff ↗	N=25000	LLR: 3.15 (-2.25, 2.89) [-5.00, 0.00] Games: 33150 W: 12066 L: 12079 D: 9005 Ptnml(0-2): 1187, 2734, 8678, 2857, 1119
Fix root position repetition detection
MasterStruct	Eunoia	main #8864 Finished	diff ↗	40.0+0.40	LLR: 1.76 (-2.25, 2.89) [-5.00, 0.00] Games: 10636 W: 3065 L: 3034 D: 4537 Ptnml(0-2): 295, 1214, 2274, 1235, 300
Fix root position repetition detection oops pushed to main
MasterStruct	Eunoia	main #8863 Passed	diff ↗	8.0+0.08	LLR: 2.90 (-2.25, 2.89) [-6.70, 0.00] Games: 11452 W: 3466 L: 3438 D: 4548 Ptnml(0-2): 377, 1277, 2385, 1315, 372
Fix root position repetition detection oops pushed to main
MasterStruct	Eunoia	refactor/repetition #8862 Passed	diff ↗	8.0+0.08	LLR: 3.07 (-2.25, 2.89) [-10.00, 0.00] Games: 1846 W: 592 L: 539 D: 715 Ptnml(0-2): 49, 180, 411, 235, 48
Refactor repetition Keys
MasterStruct	Eunoia	feat/mate-distance-pruning #8851 Finished	diff ↗	8.0+0.08	LLR: 0.60 (-2.25, 2.89) [-5.00, 0.00] Games: 25004 W: 7491 L: 7581 D: 9932 Ptnml(0-2): 666, 2689, 5873, 2617, 657
Add mate distance pruning
MasterStruct	Eunoia	refactor/search-loop #8849 Finished	diff ↗	8.0+0.08	LLR: 0.74 (-2.25, 2.89) [-5.00, 0.00] Games: 10922 W: 3291 L: 3308 D: 4323 Ptnml(0-2): 278, 1076, 2747, 1105, 255
Revert PV implementation
nocturn9x	Heimdall	tyrfing-12ksn #8860 Finished	diff ↗	40.0+0.40	LLR: 0.10 (-2.25, 2.89) [0.00, 3.00] Games: 448 W: 116 L: 109 D: 223 Ptnml(0-2): 2, 47, 118, 56, 1
Switch to tyrfing-12ksn (bench 4979204) Select tyrfing-12ksn network 77F0596E and set the multilayer scale to 288. Update the documented default scale. At scale 305, the 7,153,653-position Lichess comparison gives absolute means 927.90 for master and 981.15 for the new network; getScale yields 288.446721, rounded to 288. Validation: make bench with 77F0596E at scale 288 gives 4979204 nodes at depth 13. Confirmed the network ID and compiled scale; the 10,000-position absolute mean is 911.04. Runtime-check build, full-book evaluation and Python regressions passed (5 skipped). Network weights remain local pending separate publication. Co-authored-by: GPT 6.1 Sol <codex@openai.com>
nocturn9x	Heimdall	tyrfing-12ksn #8859 Finished	diff ↗	8.0+0.08	LLR: -0.05 (-2.25, 2.89) [0.00, 3.00] Games: 4238 W: 1130 L: 1123 D: 1985 Ptnml(0-2): 52, 497, 1008, 516, 46
Switch to tyrfing-12ksn (bench 4979204) Select tyrfing-12ksn network 77F0596E and set the multilayer scale to 288. Update the documented default scale. At scale 305, the 7,153,653-position Lichess comparison gives absolute means 927.90 for master and 981.15 for the new network; getScale yields 288.446721, rounded to 288. Validation: make bench with 77F0596E at scale 288 gives 4979204 nodes at depth 13. Confirmed the network ID and compiled scale; the 10,000-position absolute mean is 911.04. Runtime-check build, full-book evaluation and Python regressions passed (5 skipped). Network weights remain local pending separate publication. Co-authored-by: GPT 6.1 Sol <codex@openai.com>
dan	Hobbes	feature/re-corr #8857 Finished Error	diff ↗	8.0+0.08	LLR: 0.02 (-2.25, 2.89) [0.00, 3.00] Games: 70 W: 16 L: 15 D: 39 Ptnml(0-2): 0, 5, 24, 6, 0
nocturn9x	Heimdall	tyrfing-12ksn #8855 Finished	diff ↗	8.0+0.08	LLR: -0.58 (-2.25, 2.89) [0.00, 3.00] Games: 436 W: 96 L: 132 D: 208 Ptnml(0-2): 6, 66, 105, 40, 1
Switch to tyrfing-12ksn (bench 5228946) Select the completed SB 1100 tyrfing-12ksn network and set the multilayer evaluation scale to 274. Update the documented scale. The 7,153,653-position Lichess comparison gives absolute means 927.90 for master and 1032.37 for the new network at scale 305; getScale yields 274.135727, rounded to 274. Validation: make bench with the selected network at scale 274 gives 5228946 nodes at depth 13. Confirmed the built-in network ID and scale. Runtime-check build and Python regressions passed (5 skipped). The network weights remain local pending separate publication. Co-authored-by: GPT 6.1 Sol <codex@openai.com>
sp00ph	WakWak	matzo2 #8836 Finished Error Duck960	diff ↗	8.0+0.08	LLR: -1.82 (-2.25, 2.89) [0.00, 4.20] Games: 7884 W: 2921 L: 2993 D: 1970 Ptnml(0-2): 306, 717, 1978, 625, 316