Published July 16, 2026 by Statz NFL.
The maths of not going broke — flat units, fractional Kelly, and why bet sizing beats bet picking.
The unit
Pick a bankroll you could lose without consequence, and define one unit as 1–2% of it. Every result is then measured in units, which does two jobs: it keeps any single loss survivable, and it makes your record honest — up 8.4 units means something; up $300 means nothing without the stakes.
Variance is bigger than you think
A genuinely good bettor hitting 55% against the spread still loses 45 of every 100 bets, and any given run of five goes all wrong about once in fifty-four — over a few hundred bets, a five-game losing streak is close to certain. Sizing exists so that inevitable stretch is an annoyance rather than an ending.
Kelly, and why to shrink it
The Kelly criterion sizes a bet by its expected value: stake fraction = EV per unit staked / (decimal odds − 1). It maximises long-run growth but assumes your edge estimate is exact — it never is. Betting a quarter or half of Kelly keeps most of the growth with a fraction of the drawdowns, which is why every practical bettor shades down.
Never chase
The only unbreakable rule: the size of your next bet must not depend on the result of your last one. Doubling after losses feels like discipline and is mathematically the opposite — it concentrates your bankroll on the exact nights variance runs against you.
Put it to work
- Best Bets Edges and EV to size against.
- Best Bets Tracker Every published pick, graded at its stake — the running profit line.
More NFL betting guides
- What Is Expected Value in NFL Betting?
- Model Edges and Star Ratings, Explained
- How to Read NFL Odds
- Same-Game Parlays: Why the Quoted Price Lies
- A Player-Props Research Workflow
- AI Anytime Touchdown Predictions: NFL TD Scorer Picks
- AI NFL Player Props: Prop Picks, Projections and Hit Rates
- AI NFL Stat Predictions: Player and Team Stat Projections