Analysis
Model v2.4 Performance Update: Why February Has Been Our Best Month
Published February 09, 2026 · 2 days ago
Key Takeaway
Model v2.4 delivering 75% accuracy and +22.1% ROI in February, driven by improved fatigue modeling, injury nuance, and referee tendency analysis.
Our v2.4 model update, deployed on January 27th, has produced our best 2-week stretch of the season: 75% accuracy on 48 predictions with a +22.1% ROI.
The key improvements:
1. Enhanced fatigue modeling: v2.4 incorporates travel distance, time zone changes, and back-to-back scheduling into our fatigue coefficient. This has particularly improved NBA predictions, where our B2B predictions went from 61% to 78% accuracy.
2. Real-time injury impact: Instead of binary in/out status, we now model partial injury effects — a player listed as questionable with a knee issue plays, but at 85% effectiveness. This nuance has been worth 3-4 percentage points of accuracy.
3. Referee tendency integration: Different referee crews call games differently. Our model now adjusts over/under predictions based on the assigned crew's historical foul rate and pace-of-play data.
Looking ahead, we're testing v2.5 features including weather impact modeling for outdoor sports and crowd noise analysis for home advantage calculations.
The key improvements:
1. Enhanced fatigue modeling: v2.4 incorporates travel distance, time zone changes, and back-to-back scheduling into our fatigue coefficient. This has particularly improved NBA predictions, where our B2B predictions went from 61% to 78% accuracy.
2. Real-time injury impact: Instead of binary in/out status, we now model partial injury effects — a player listed as questionable with a knee issue plays, but at 85% effectiveness. This nuance has been worth 3-4 percentage points of accuracy.
3. Referee tendency integration: Different referee crews call games differently. Our model now adjusts over/under predictions based on the assigned crew's historical foul rate and pace-of-play data.
Looking ahead, we're testing v2.5 features including weather impact modeling for outdoor sports and crowd noise analysis for home advantage calculations.