ThoroughSim TS
  • Races
  • Tools
  • Import DRF
  • Calibration
  • Admin
  1. Calibration
Import DRF Add Race

Simulation Calibration

Factor weights learned from actual race results
31
Races Learned
0.145
Avg Rank Correlation (ρ)
1.0 = perfect · 0 = random
-0.9
Worst Race ρ
0.893
Best Race ρ

Current Factor Weights

Starts at 1.00 (unchanged). Increases when the factor successfully predicted better finishers; decreases when it led the model astray. Learning rate: 0.08 per race. Clamped between 0.15 and 4.00.

Factor Weight Visual Races Signal Description Last Updated
Best speed ceiling 0.9199 ↓
31 -1 Current avg BSR vs best Brisnet speed figure this year — horses near their proven ceiling get a boost May 21, 2026
Career form 0.7564 ↓
31 -3.04 Career win% quality + current-year momentum vs prior years May 21, 2026
Jockey local wins 0.7146 ↓
31 -3.57 Jockey win history in our own recorded races — proven winning jockeys at this level get a boost May 21, 2026
Jockey trainer combo 0.8631 ↓
31 -1.71 Jockey + Trainer partnership win rate — certain duos consistently outperform each individually (×1.5 coefficient) May 21, 2026
Local db wins 1.0000
31 0 Actual wins recorded in our own race results — strongest verified evidence (×2 coefficient) May 21, 2026
Morning line 1.0000
31 0 Track handicapper ML odds — soft secondary endorsement (±1.5 pts) May 21, 2026
Track affinity 0.6312 ↓
31 -4.61 Win% at this exact track vs 16% national avg (career stat blocks) May 21, 2026
Recalibrate = reset weights then replay every recorded result through the engine — use this to make all your results weigh in.
Reset = wipe weights only (no replay). Use after a major engine change.

Race Feedback History

Race Date Horses Rank ρ Top Factor Adjustments Analyzed
Laurel Park R6 May 16, 2026 6 0.371 form +0.746
jockey trainer combo +0.522
bsr trajectory +0.443
May 21 4:53pm
Laurel Park R7 May 16, 2026 7 0.893 beaten lengths +0.602
bsr trajectory +0.402
jockey trainer combo +0.348
May 21 4:53pm
Laurel Park R8 May 16, 2026 7 -0.214 freshness +0.737
bounce +0.478
post position +0.397
May 21 4:53pm
Laurel Park R9 May 16, 2026 7 0.286 weight +0.725
jockey +0.518
beaten lengths +0.229
May 21 4:53pm
Laurel Park R10 May 16, 2026 5 -0.6 best speed ceiling +0.395
bounce +0.395
bsr ability +0.142
May 21 4:53pm
Laurel Park R11 May 16, 2026 8 0.167 class change +0.304
trainer +0.207
career form +0.184
May 21 4:53pm
Laurel Park R12 May 16, 2026 5 0.1 form +0.879
trainer +0.846
bsr trajectory +0.59
May 21 4:53pm
Laurel Park R13 May 16, 2026 5 -0.9 form +0.912
bounce +0.79
trainer +0.669
May 21 4:53pm
Woodbine R4 May 3, 2026 8 0.048 jockey local wins +0.575
post position +0.49
distance +0.327
May 21 4:53pm
Woodbine R5 May 3, 2026 7 0.714 freshness +0.299
jockey local wins +0.282
best speed ceiling +0.27
May 21 4:53pm
Woodbine R6 May 3, 2026 8 -0.357 form +0.418
beaten lengths +0.311
pace scenario +0.067
May 21 4:53pm
Woodbine R7 May 3, 2026 6 0.829 jockey trainer combo +0.515
jockey +0.447
beaten lengths +0.417
May 21 4:53pm
Woodbine R8 May 3, 2026 6 0.486 form +0.861
class change +0.716
pace scenario +0.544
May 21 4:53pm
Woodbine R9 May 3, 2026 11 0.225 jockey +0.72
jockey local wins +0.497
career form +0.49
May 21 4:53pm
Laurel Park R1 May 16, 2026 10 0.103 bsr trajectory +0.477
jockey trainer combo +0.347
jockey local wins +0.297
May 21 4:53pm
Laurel Park R2 May 16, 2026 9 0.483 post position +0.636
surface +0.539
class change +0.326
May 21 4:53pm
Laurel Park R3 May 16, 2026 11 0.627 bsr trajectory +0.319
jockey local wins +0.265
jockey +0.168
May 21 4:53pm
Laurel Park R4 May 16, 2026 5 -0.5 best speed ceiling +0.612
weight +0.408
career form +0.271
May 21 4:53pm
Laurel Park R5 May 16, 2026 14 0.398 jockey local wins +0.331
class change +0.199
surface +0.172
May 21 4:53pm
Tampa Bay Downs R1 May 3, 2026 6 0.657 surface +0.446
weight +0.316
jockey trainer combo +0.316
May 21 4:53pm
Tampa Bay Downs R2 May 3, 2026 7 0.429 jockey trainer combo +0.74
bsr ability +0.174
best speed ceiling +0.125
May 21 4:53pm
Tampa Bay Downs R3 May 3, 2026 3 -0.5 post position +0.995
distance +0.995
freshness +0.585
May 21 4:53pm
Tampa Bay Downs R4 May 3, 2026 8 0.524 bounce +0.557
freshness +0.376
best speed ceiling +0.335
May 21 4:53pm
Tampa Bay Downs R5 May 3, 2026 9 -0.367 bsr trajectory +0.234
surface +0.083
freshness +0.05
May 21 4:53pm
Tampa Bay Downs R6 May 3, 2026 6 0.243 bounce +0.87
class change +0.618
form +0.399
May 21 4:53pm
Tampa Bay Downs R7 May 3, 2026 9 0.25 best speed ceiling +0.496
post position +0.375
jockey trainer combo +0.373
May 21 4:53pm
Tampa Bay Downs R8 May 3, 2026 5 -0.9 beaten lengths +0.401
post position +0.391
freshness +0.29
May 21 4:53pm
Tampa Bay Downs R9 May 3, 2026 9 0.3 bsr trajectory +0.366
career form +0.245
beaten lengths +0.228
May 21 4:53pm
Woodbine R1 May 3, 2026 7 -0.357 post position +0.559
freshness +0.248
jockey +0.222
May 21 4:53pm
Woodbine R2 May 3, 2026 7 0.464 weight +0.592
class change +0.43
jockey local wins +0.226
May 21 4:53pm
Woodbine R3 May 3, 2026 6 0.6 bsr trajectory +0.875
weight +0.463
class change +0.463
May 21 4:53pm

How the Learning Works

  1. You enter actual results on any race page after the race is run.
  2. The engine compares each horse's predicted rank (by simulation win%) to their actual finish position.
  3. For each scoring factor (BSR, trajectory, bounce, etc.) it computes the Pearson correlation between that factor's values across the field and the prediction errors.
  4. Factors that consistently pointed at the right horses have their weight increased ↑. Factors that led the model astray get decreased ↓.
  5. The Spearman rank correlation (ρ) measures how well the simulation ranked the field overall — closer to 1.0 = better predictions.
  6. After enough races, the weights converge to reflect what actually matters most in your data set.