The most overlooked skill in sports betting
It's not bad picks - it's bad bet sizing. You can be a winning handicapper and still lose money if you size bets poorly. Conversely, proper sizing protects you during losing streaks and maximizes gains during hot runs.
A unit is your standard bet size, expressed as a percentage of your total bankroll. It provides a consistent framework for sizing all your bets.
If your bankroll is
and 1 unit = 2%
Then your unit size is
per standard bet
| Bankroll | 1% Unit | 2% Unit | 3% Unit |
|---|---|---|---|
| $500 | $5 | $10 | $15 |
| $1,000 | $10 | $20 | $30 |
| $2,500 | $25 | $50 | $75 |
| $5,000 | $50 | $100 | $150 |
| $10,000 | $100 | $200 | $300 |
Bet the same amount (1 unit) on every wager, regardless of confidence level.
Vary bet size based on your confidence level, using a 1-5 unit scale.
| Rating | Units | % of Bankroll | When to Use |
|---|---|---|---|
| 1 Star | 1 unit | 1-2% | Standard play, slight edge |
| 2 Star | 1.5 units | 2-3% | Above average confidence |
| 3 Star | 2 units | 3-4% | Strong edge, high confidence |
| 4 Star | 2.5 units | 4-5% | Rare - very strong edge |
| 5 Star | 3 units | 5% MAX | Extremely rare - max play |
A mathematical formula that calculates optimal bet size based on your edge and the odds offered.
f* = (bp - q) / b
You believe the Chiefs have a 58% chance to cover -3, and the odds are -110 (decimal 1.91):
Full Kelly suggests betting 11.8% of your bankroll - aggressive enough that almost nobody should follow it raw.
Most bettors use a fraction of the Kelly recommendation to account for uncertainty in their probability estimates:
Maximum growth but extreme variance. Not recommended.
Recommended. ~75% of optimal growth, much lower variance.
Conservative. For uncertain edges or risk-averse bettors.
Many successful bettors recalculate unit size weekly or monthly. Don't adjust mid-week based on short-term results - stick to your schedule.
Risking 10%+ per bet means a short losing streak can wipe you out. Even 5 losses in a row at 10% each leaves you down 40%.
Doubling after losses to "get even" is a recipe for disaster. This is the fastest way to go broke.
Betting $50 one game and $500 the next based on "feel" removes the benefits of proper bankroll management.
If you don't know your total bankroll, you can't calculate proper unit sizes. Define it first.
Money you can afford to lose entirely
Conservative starting point
1 unit per bet until profitable
Unit sizing is streak insurance, so price the streaks. For a bettor who genuinely wins 55% of bets (better than almost everyone), the chance that any given run of N bets is all losses is 0.45N — and across a 250-bet season you get roughly 250 overlapping chances for each streak length. Exact formula math:
| Streak | Chance in any given window (0.45N) | Bankroll hit at 2% units | At 5% units | At 10% units |
|---|---|---|---|---|
| 4 straight losses | 4.1% | −7.8% | −18.5% | −34.4% |
| 6 straight losses | 0.83% | −11.4% | −26.5% | −46.9% |
| 8 straight losses | 0.17% | −14.9% | −33.7% | −57.0% |
Exact formula math: bankroll hit = 1 − (1 − unit)N, betting a fixed percentage of the current bankroll each time. Streak chances are per window; over a season, expect multiple 4-loss streaks and treat a 6-loss streak as normal.
Read the 10% column and the “never exceed 5%” rule above stops being a slogan: a routine 6-loss streak costs nearly half the bankroll at 10% units, and recovering from −47% requires +88% just to get even. At 2% units the same streak is an 11% dent. The streaks are identical; only the sizing decides whether they are survivable.
The deeper justification for small units is how little one NFL game can tell you. From the bundled nflverse game log (7,276 completed games, 1999–2025): the final margin misses the closing spread by 10.3 points on average, with a standard deviation of 13.2, and only 21.5% of games land within 3 points of the line. Even a correct opinion — a spread genuinely half a point off — changes a single game's outcome only when the margin lands in that half-point window; the rest is noise you cannot forecast.
That is what “an edge shows up over hundreds of bets, not Sunday” means in numbers. A 55% bettor over 100 bets still finishes with a losing record about 13% of the time (normal approximation, σ ≈ 5 wins). Your unit size is what keeps you solvent long enough for 55% to look like 55%. Data-derived context computed by the author from the bundled game log; probability figures are exact formula math.
Keep reading: the Kelly calculator for edge-based sizing, and point spread accuracy for the noise numbers behind the second card.
Nothing here is betting advice, and no number on this page predicts any single game. Sports betting is legal only in some jurisdictions and only for adults (21+ in most U.S. states). If betting stops being entertainment, call or text 1-800-GAMBLER. Read our full disclaimer.