Finished
RosentWinternet_311d *diff8.0+0.08LLR: -2.95 (-2.94, 2.94) [0.00, 5.00]
Games: 20226 W: 5005 L: 5044 D: 10177
Ptnml(0-2): 475, 2478, 4222, 2487, 451
Reduced the weighting of the draw prediction loss component.
RosentWinternet_311d *diff8.0+0.08LLR: -2.95 (-2.94, 2.94) [0.00, 5.00]
Games: 11346 W: 2730 L: 2809 D: 5807
Ptnml(0-2): 236, 1434, 2411, 1357, 235
Draw calibration handled with MSE loss directly on the draw probability should be less sensitive to outliers that are common in the high entropy STC time controls the training data stems from.
RosentWinternet_311 *diff40.0+0.40LLR: 2.96 (-2.94, 2.94) [0.00, 5.00]
Games: 3948 W: 860 L: 731 D: 2357
Ptnml(0-2): 32, 395, 1004, 498, 45
Fully trained model with identical hyperparameters to prior gen, but extra data. This should clearly beat the master branch.
RosentWinternet_311 *diff8.0+0.08LLR: 2.96 (-2.94, 2.94) [0.00, 5.00]
Games: 8158 W: 2101 L: 1925 D: 4132
Ptnml(0-2): 193, 922, 1694, 1056, 214
Testing net with same hyperparameters but extra self play data a bit prematurely to get a sense of where things lie.
RosentWinternet_test *diff40.0+0.40LLR: 2.95 (-2.94, 2.94) [0.00, 5.00]
Games: 3618 W: 807 L: 675 D: 2136
Ptnml(0-2): 37, 366, 880, 480, 46
Trained with new data. The new training data has a very different distribution and has neutral effect on validation metrics.
RosentWinternet_test *diff8.0+0.08LLR: 2.97 (-2.94, 2.94) [-2.00, 3.00]
Games: 7204 W: 1851 L: 1710 D: 3643
Ptnml(0-2): 160, 820, 1495, 973, 154
Trained with new data. The new training data has a very different distribution and has neutral effect on validation metrics.
RosentWinternet_test *diff40.0+0.40LLR: 1.77 (-2.94, 2.94) [0.00, 5.00]
Games: 4820 W: 1009 L: 922 D: 2889
Ptnml(0-2): 42, 513, 1225, 576, 54
Net trained with some extra data I had laying around and not yet utilized.
RosentWinternet_test *diff8.0+0.08LLR: 2.95 (-2.94, 2.94) [0.00, 5.00]
Games: 4334 W: 1208 L: 1050 D: 2076
Ptnml(0-2): 92, 482, 903, 556, 134
Net trained with some extra data I had laying around and not yet utilized.
RosentWinternet_test *diff40.0+0.40LLR: 2.17 (-2.94, 2.94) [0.00, 5.00]
Games: 8822 W: 1791 L: 1673 D: 5358
Ptnml(0-2): 80, 938, 2276, 1018, 99
Confirmed 2 year old net reproduced and slightly improved
RosentWinternet_test *diff8.0+0.08LLR: 3.03 (-2.94, 2.94) [0.00, 5.00]
Games: 27442 W: 6402 L: 6150 D: 14890
Ptnml(0-2): 464, 3213, 6148, 3399, 497
RosentWinternet_test *diff8.0+0.08LLR: -3.11 (-2.94, 2.94) [0.00, 5.00]
Games: 4948 W: 1188 L: 1304 D: 2456
Ptnml(0-2): 109, 648, 1070, 544, 103
RosentWinternet_test *diff8.0+0.08LLR: -3.06 (-2.94, 2.94) [0.00, 5.00]
Games: 7424 W: 1778 L: 1880 D: 3766
Ptnml(0-2): 163, 943, 1583, 879, 144
RosentWinternet_test *diff8.0+0.08LLR: -3.05 (-2.94, 2.94) [0.00, 5.00]
Games: 4774 W: 1109 L: 1224 D: 2441
Ptnml(0-2): 124, 593, 1040, 534, 96
RosentWinternet_test *diff8.0+0.08LLR: -3.02 (-2.94, 2.94) [0.00, 5.00]
Games: 6534 W: 1565 L: 1671 D: 3298
Ptnml(0-2): 155, 823, 1413, 725, 151
RosentWinterfable_bugs *diff8.0+0.08LLR: -2.97 (-2.94, 2.94) [-5.00, 0.00]
Games: 48452 W: 11655 L: 12015 D: 24782
Ptnml(0-2): 1111, 6015, 10255, 5813, 1032
RosentWinterminor_performance_improvementsdiff40.0+0.40Elo: 77.61 +- 5.83 (95%) [N=4000]
Games: 4000 W: 1121 L: 242 D: 2637
Ptnml(0-2): 2, 142, 963, 761, 132
RosentWinterminor_performance_improvements *diff8.0+0.08LLR: 3.00 (-2.94, 2.94) [0.00, 5.00]
Games: 1760 W: 516 L: 372 D: 872
Ptnml(0-2): 23, 183, 351, 273, 50
RosentWinterstable_softmax *diff40.0+0.40LLR: 2.95 (-2.94, 2.94) [-5.00, 0.00]
Games: 7554 W: 1511 L: 1429 D: 4614
Ptnml(0-2): 54, 843, 1917, 893, 70
RosentWinterstable_softmax *diff8.0+0.08LLR: 2.99 (-2.94, 2.94) [-5.00, 0.00]
Games: 10990 W: 2702 L: 2619 D: 5669
Ptnml(0-2): 237, 1274, 2400, 1337, 247
RosentWinterstable_softmax *diff8.0+0.08LLR: 0.05 (-2.94, 2.94) [0.00, 5.00]
Games: 348 W: 83 L: 79 D: 186
Ptnml(0-2): 6, 48, 66, 44, 10
RosentWinterno_speculative_until_not_mated *diff40.0+0.40LLR: 1.72 (-2.94, 2.94) [-5.00, 0.00]
Games: 11782 W: 2289 L: 2270 D: 7223
Ptnml(0-2): 117, 1277, 3087, 1290, 120
RosentWinterasymmetric_aspiration_windows *diff40.0+0.40LLR: -0.52 (-2.94, 2.94) [0.00, 5.00]
Games: 3990 W: 783 L: 787 D: 2420
Ptnml(0-2): 36, 438, 1055, 426, 40
RosentWinterno_speculative_until_not_mated *diff8.0+0.08LLR: 2.95 (-2.94, 2.94) [-5.00, 0.00]
Games: 24150 W: 5894 L: 5874 D: 12382
Ptnml(0-2): 518, 2950, 5099, 3010, 498
RosentWinteraspiration_rewrite *diff40.0+0.40LLR: -1.27 (-2.94, 2.94) [-5.00, 0.00]
Games: 1840 W: 337 L: 394 D: 1109
Ptnml(0-2): 31, 216, 466, 193, 14
RosentWinteraspiration_rewrite *diff8.0+0.08LLR: -0.26 (-2.94, 2.94) [-5.00, 0.00]
Games: 9526 W: 2341 L: 2398 D: 4787
Ptnml(0-2): 242, 1152, 1997, 1165, 207