Methodology
Plain-English overview of how NTG Sports models work — and what they do not do.
What the models use
- Team win-probability models trained on historical game results, park factors, and rest.
- Player prop scoring from MLB game logs (L5, L10, season hit rates vs posted lines).
- Matchup context where available (e.g. opposing starter ERA for batter props).
- Live and cached sportsbook odds from The Odds API for EV and edge calculations.
What we ignore (for now)
- Injury news not yet in the data feed, steam moves, and sharp closing-line intelligence.
- Same-game parlay correlation beyond basic warnings.
- Guaranteed ROI or bankroll management — this is research tooling, not financial advice.
Edge & EV
For moneyline picks, edge is model win probability minus implied book probability (after vig). For props, we rank by recent form hit rate vs the posted line, with matchup adjustments. A pick marked actionable means the model side clears our minimum hit-rate threshold on offered odds.
Responsible use
NTG Sports is experimental analytics for adults 21+. It is not betting advice. Past model performance does not guarantee future results. See our performance log for tracked prop offers and outcomes.