SirStone
1ed7797cb6
- New guns: guess-factor (GF histogram), pattern-matcher (movement tape replay) - New modules: minimum-risk melee movement, spinning melee radar - New test bots: PatternMover, RandomMover, WaveSurfer - Fixed: FeedbackEvent now carries actualX/actualY for proper GF learning - Fixed: TM gun warmup gating + directional residuals - Fixed: circular gun integrated formula + multi-bin omega cache - Fixed: oscillator wall-bounce lockout - Fixed: phantom meteor perpendicular body orientation - 6/6 battle wins across all enemy types
110 lines
5.0 KiB
Plaintext
110 lines
5.0 KiB
Plaintext
========================================================================
|
|
WiSARD / WNN BACKTEST REPORT
|
|
Rows: 277
|
|
========================================================================
|
|
|
|
--- BASELINE: Pure Linear Extrapolation (P1) ---
|
|
|
|
Power MAE RMSE Median Max %<5u
|
|
----------------------------------------------------
|
|
p0.10 8.10 11.93 5.00 61.09 49.8%
|
|
p0.42 8.68 12.70 5.33 64.35 49.1%
|
|
p0.74 9.37 13.59 5.75 67.94 46.6%
|
|
p1.07 10.18 14.62 6.24 72.04 40.1%
|
|
p1.39 11.03 15.72 7.26 76.48 38.6%
|
|
p1.71 11.97 16.96 8.38 81.92 35.7%
|
|
p2.03 13.22 18.49 9.62 87.57 32.5%
|
|
p2.36 14.64 20.24 11.00 94.46 29.2%
|
|
p2.68 16.46 22.43 12.77 102.17 26.4%
|
|
p3.00 18.70 25.16 14.00 111.15 23.5%
|
|
|
|
--- WiSARD: K=8, 276 bits (no XOR) | 35 nodes ---
|
|
|
|
Power MAE RMSE Median Max %<5u vs P1
|
|
--------------------------------------------------------------
|
|
p0.10 6.07 9.22 4.07 42.89 56.7% -2.03
|
|
p0.42 6.61 9.87 4.25 46.12 53.8% -2.07
|
|
p0.74 7.27 10.67 5.06 49.69 49.8% -2.10
|
|
p1.07 8.04 11.62 5.90 53.78 46.2% -2.14
|
|
p1.39 8.88 12.65 6.48 58.22 43.0% -2.15
|
|
p1.71 9.88 13.87 7.60 63.58 39.4% -2.09
|
|
p2.03 11.13 15.40 8.65 69.24 33.6% -2.10
|
|
p2.36 12.57 17.15 9.82 76.11 30.0% -2.07
|
|
p2.68 14.45 19.39 11.21 83.82 27.4% -2.01
|
|
p3.00 16.75 22.20 13.13 92.80 23.5% -1.95
|
|
|
|
Learning curve (p1.07): first-50 MAE=14.99 last-50 MAE=5.55
|
|
LUT fill: 1405 / 8960 entries (15.681%)
|
|
|
|
--- WiSARD: K=8, 345 bits (with XOR) | 44 nodes ---
|
|
|
|
Power MAE RMSE Median Max %<5u vs P1
|
|
--------------------------------------------------------------
|
|
p0.10 6.19 9.17 4.11 41.57 59.2% -1.92
|
|
p0.42 6.73 9.84 4.71 44.80 52.3% -1.95
|
|
p0.74 7.39 10.65 5.04 48.37 49.8% -1.98
|
|
p1.07 8.17 11.63 5.80 52.45 45.8% -2.01
|
|
p1.39 9.00 12.67 6.60 56.89 42.6% -2.02
|
|
p1.71 9.98 13.89 7.52 62.26 37.9% -2.00
|
|
p2.03 11.21 15.42 8.68 67.92 32.9% -2.01
|
|
p2.36 12.65 17.18 9.73 74.80 28.9% -1.99
|
|
p2.68 14.51 19.41 11.26 82.50 26.4% -1.95
|
|
p3.00 16.79 22.22 13.07 91.48 23.8% -1.90
|
|
|
|
Learning curve (p1.07): first-50 MAE=15.17 last-50 MAE=5.78
|
|
LUT fill: 1520 / 11264 entries (13.494%)
|
|
|
|
--- WiSARD: K=12, 276 bits (no XOR) | 23 nodes ---
|
|
|
|
Power MAE RMSE Median Max %<5u vs P1
|
|
--------------------------------------------------------------
|
|
p0.10 5.85 9.07 3.70 42.14 57.8% -2.26
|
|
p0.42 6.38 9.69 4.28 45.36 54.2% -2.30
|
|
p0.74 7.02 10.44 5.17 48.93 49.5% -2.35
|
|
p1.07 7.81 11.37 5.91 53.02 44.4% -2.37
|
|
p1.39 8.63 12.37 6.43 57.45 43.0% -2.40
|
|
p1.71 9.63 13.57 7.38 62.81 39.0% -2.34
|
|
p2.03 10.86 15.09 8.20 68.47 33.2% -2.36
|
|
p2.36 12.32 16.84 9.55 75.34 30.0% -2.31
|
|
p2.68 14.23 19.08 10.76 83.05 24.9% -2.23
|
|
p3.00 16.54 21.90 12.54 92.03 23.8% -2.16
|
|
|
|
Learning curve (p1.07): first-50 MAE=14.43 last-50 MAE=5.13
|
|
LUT fill: 1969 / 94208 entries (2.090%)
|
|
|
|
--- WiSARD: K=12, 345 bits (with XOR) | 29 nodes ---
|
|
|
|
Power MAE RMSE Median Max %<5u vs P1
|
|
--------------------------------------------------------------
|
|
p0.10 5.96 9.15 3.89 43.90 58.5% -2.15
|
|
p0.42 6.49 9.78 4.45 47.12 55.2% -2.19
|
|
p0.74 7.14 10.55 4.88 50.69 50.9% -2.23
|
|
p1.07 7.92 11.50 5.75 54.77 46.9% -2.26
|
|
p1.39 8.77 12.52 6.21 59.21 41.2% -2.25
|
|
p1.71 9.74 13.73 7.09 64.56 38.3% -2.23
|
|
p2.03 11.02 15.26 8.29 70.22 32.5% -2.21
|
|
p2.36 12.48 17.02 9.51 77.09 28.9% -2.16
|
|
p2.68 14.34 19.26 10.74 84.80 24.9% -2.11
|
|
p3.00 16.64 22.08 12.52 93.78 22.4% -2.06
|
|
|
|
Learning curve (p1.07): first-50 MAE=14.85 last-50 MAE=5.27
|
|
LUT fill: 2167 / 118784 entries (1.824%)
|
|
|
|
|
|
========================================================================
|
|
SUMMARY — Average MAE across all power levels
|
|
========================================================================
|
|
|
|
Config Avg MAE vs P1
|
|
----------------------------------------------------
|
|
P1 (linear baseline) 12.235
|
|
K=8, 276 bits (no XOR) 10.165 -2.071
|
|
K=8, 345 bits (with XOR) 10.263 -1.973
|
|
K=12, 276 bits (no XOR) 9.927 -2.308
|
|
K=12, 345 bits (with XOR) 10.051 -2.184
|
|
|
|
|
|
Note: negative 'vs P1' = improvement; positive = worse than linear baseline.
|
|
Learning is online: each row trains BEFORE the next prediction.
|
|
WiSARD learns the residual correction on top of linear extrapolation.
|