📅 Select Date Range
Quick Select:
📊 Showing: Beginning to Present
📊 Overall Model Performance
Total Predictions
47454
35078 with actual results
Avg BA Error
0.205
Predicted 0.215 / Actual 0.223
Avg Rank Correlation (ρ)
0.111
DFS rank signal (>0.3 = meaningful)
📈 Predicted vs. Actual Batting Average Over Time
When the two lines track closely, the model is well-calibrated. A persistent gap means the model is systematically over- or under-projecting.
🏆 DFS Rank Correlation (Spearman ρ)
0.111
avg ρ over 134 days
WEAK SIGNAL
ρ ≥ 0.30 Strong — model adds clear DFS value
ρ 0.15–0.29 Moderate — some signal above noise
ρ 0.00–0.14 Weak — marginal improvement over random
ρ < 0.00 Negative — model is hurting rankings
Spearman ρ measures whether the model's rankings match actual outcomes — the core DFS question. Computed per game day (min 5 batters with results), then averaged across all days.
🎯 Hit Rate by Confidence Tier
Tier Description Predictions % of Total Avg Predicted BA Avg Actual BA Avg Error
High Confidence Predicted BA ≥ .300 2327 7.0% 0.341 0.267 0.222
Medium Confidence Predicted BA .250-.299 5416 16.2% 0.271 0.255 0.204
Low Confidence Predicted BA < .250 25596 76.8% 0.193 0.227 0.205
A well-calibrated model shows clear separation across tiers and lower error in High Confidence picks. If errors are similar across all tiers, the confidence signal isn't adding value.
🎰 Game Prediction Accuracy by Confidence Band
Min Confidence Games Correct Accuracy vs. 50% Baseline
≥ 50% 1653 866 52.4% +2.4%
≥ 55% 1182 634 53.6% +3.6%
≥ 60% 752 415 55.2% +5.2%
≥ 65% 441 242 54.9% +4.9%
≥ 70% 195 104 53.3% +3.3%
Higher-confidence predictions should be more accurate. If accuracy drops at higher thresholds, the model may be overconfident. Target: ≥55% accuracy beats a coin flip.
💥 Home Run Prediction Calibration
Expected HRs
2753.2
Actual HRs
3627
Calibration Ratio
0.76x
Target: 0.8–1.2x
Avg HR Probability
7.8%
📐 Brier Score
Brier Score
0.0656
Lower is better (0 = perfect)
Brier Skill Score
-2.21%
Improvement over always predicting .250 baseline
Sample Size
33339
Predictions analyzed
📊 Calibration by Predicted BA Range
Predicted Range Count Avg Predicted Avg Actual Calibration Error
0.00-0.20 14215 0.169 0.220 0.052
0.20-0.25 11381 0.223 0.236 0.013 ✓
0.25-0.30 5416 0.271 0.255 0.016 ✓
0.30-0.35 1671 0.319 0.265 0.055
0.35+ 656 0.397 0.273 0.124
Mean Absolute Calibration Error: 0.052
🎯 Game Outcome Summary
Total Games
1653
With results
Correct
866
52.4% win rate
Brier Score
0.2582
Probabilistic accuracy
Avg Confidence When Correct
60.6%
Avg Confidence When Wrong
60.0%