Built an EPL/NBA value model — looking for feedback on methodology
Hey all — been building a statistical model for EPL/NBA value bets and wanted feedback on the approach. Not selling anything, just genuinely curious if this makes sense or if I'm missing something.
What it does
- Estimates true probabilities from recent form, injuries, and market movement
- Compares to bookmaker implied odds to find edges - Outputs 1U/2U picks based on edge strength
- Logs every pick and result publicly, win or lose
What I'm unsure about:
- Data freshness: built a guardrail that blocks picks when team/player data is stale, not sure if the threshold is right
- Sample size: ~29 logged bets so far (EPL 5 gameweeks + NBA testing) — feels small but don't want to force volume
- Market efficiency: any markets you've found consistently show more exploitable inefficiency (EPL totals vs ML, NBA props vs spreads, etc.)?
Tech stack:
- Python for data/model logic - CSV logging for now (planning to move to a proper DB)
- Manual odds capture (no real-time API yet)
Happy to share more on the probability estimation method or hear where this might be overfitting. Thanks in advance