How Many Units to Bet on Player Props

A player prop should usually get a smaller unit than a spread or a total: somewhere around 0.5% to 1% of your bankroll rather than the 1% to 2% that every bankroll guide quotes, and less again for alt lines, ladders and same-game parlays.

That isn't the usual advice. Search the question and you'll find a lot of pages saying the same thing: a unit is 1% to 2% of bankroll, never go past 5%, bet 0.5 units when you're unsure and 3 when you love it. All fine as far as it goes. None of them asks whether a Tuesday night rebounds over deserves the same stake as a Sunday side, and the numbers say it doesn't. Props carry more hold, rest on thinner samples, and tempt you into stacking bets on the same game, which is the quickest way to bet three units without noticing. This piece covers how to size a prop, what the unit is in dollars, how it should change by prop type, and how to keep it honest over a season.

Start with the price. Wizard of Odds puts the typical sportsbook hold on player props at 4% to 6% for major markets and 10% to 15% or more for niche and live markets. Even at a standard -110, you need to win 52.4% of the time just to break even. The same article states plainly that most player props offered by sportsbooks are negative EV for the average bettor. So the starting point is that the house takes more off the top here than almost anywhere else you can bet.

Then there's the estimate itself. A prop edge comes from your projection of what one player will do, and that projection usually rests on a small sample. Using the standard error formula Wizard of Odds gives, a 50% hit rate over 10 games carries a 95% confidence interval of about plus or minus 31 points. Over 50 games, a full chunk of an NBA season, it's still plus or minus 13.9. If you think a player clears his points line 58% of the time, the honest version of that sentence covers a range wide enough to include a losing bet.

Third, variance doesn't go away just because you have an edge. In Wizard of Odds' variance and bankroll breakdown for props, a bettor making 100 bets of $100 at -110 with a true 55% win rate expects $500 profit, but the standard deviation is $950. That leaves roughly a 30% chance of being behind after 100 bets with a real edge. Props are where a lot of bettors fire 10 or 15 bets a night, so that spread shows up fast.

Fourth, correlation. Two props from the same game aren't two separate bets. If a point guard has a big night, his points, his assists and his team's total tend to move together. So three "one unit" bets on that game add up to one position much bigger than one unit.

Odds Reference reaches the same place by a different route. Its bankroll guide gives sides and totals a unit of 1% to 3% with a 50 to 100 unit bankroll, and props and parlays a unit of 0.5% to 1% with a 100 to 200 unit bankroll. (Its own FAQ quotes 1% to 2% for sides, which contradicts its table, so treat the exact edges loosely.) The direction is clear, and I think it's right.

One gap needs flagging. Betting limits on props are often said to be lower than on sides, and that would be another reason to keep prop units small. The sources here don't give limit figures, though, so this piece makes its case on hold, variance and correlation alone.

A unit is a fixed share of your bankroll that counts as one standard bet. Then you think and record in units, not dollars, so a $20 bettor and a $200 bettor can compare results on the same scale. The generic standard is 1% to 2%, with 5% as the hard ceiling for any single bet. DeucesCracked defines the default as 1% of bankroll, with 2% as standard and 3% only for selective bettors with a proven long-term edge.

My suggestion is to run two units, not one. Keep your side/total unit wherever your main bankroll plan puts it, and set a prop unit at half of it. If you haven't set the base number yet, the step-by-step bankroll plan covers that first. Everything below adjusts it.

The ceiling column is there to show how far away it should stay. Wizard of Odds works through a $1,000 bankroll betting a fixed 5% at -110 with a 55% win rate: 20 straight losses wipe you out, and the risk of ruin comes to about 1.96%. That sounds small until you remember it assumes you really do win 55%, which most prop bettors can't show. Odds Reference gives a 3% edge a risk of ruin under 0.3% with 100 units, and about 22% with only 25. The size of the unit sets how many mistakes you can survive.

This is where the generic guides stop and props really start. "A prop" covers a main-line points over at -110, an alt line at +450, a leg buried in a same-game parlay, and a ladder of three alt lines on one player. Those carry very different risks, and each needs its own sizing rule.

The ranges below are this blog's recommendations, worked out from the figures in the sources. They're expressed as a share of the whole bankroll so they don't depend on which unit you picked.

Same-game parlays get their own row because the price itself is part of the risk. Covers worked through a correlated ticket of Lakers +7.5, Luka Doncic over 34.5 points and Stephen Curry under 3.5 made threes. Priced as independent legs it came to +673; FanDuel offered +496. On a $100 stake that's $773 back at the independent price versus $596 at FanDuel, so $173 of profit left on the table. DraftKings and bet365 both had +500, so even shopping around only got four points back. Sportsbooks know correlated legs hit together, and they price that in before you do. I'm not saying never bet an SGP. I'm saying size it like the long shot with extra hold that it is.

The standard confidence scale runs 0.5 units for a lean, 1 unit for a standard play, 2 units for a strong edge and 3 units for maximum conviction. SharpAi's version, on a $1,000 bankroll with a $20 unit, puts the top tier at $60, or 6% of bankroll. That's already above the 5% ceiling the same guides insist on.

For props, I'd use the scale with two changes. Apply it to the prop unit, not the side unit, and treat 3 units as almost never. In practice:

• 0.5 prop units: EV estimated at about +3%, the minimum Wizard of Odds suggests betting, or any niche market where the hold is wide.
• 1 prop unit: a main-line prop where your projection beats the no-vig price clearly and the sample behind it is reasonable.
• 2 prop units: a strong edge you could explain to a skeptic, at a price you've confirmed is the best available.
• 3 prop units: keep it for a known pricing error, like a stale line after confirmed injury news. OddCompare suggests no more than about 20% of a month's bets should sit in its top tier, and for props I'd go well below that.

The best argument against any of this comes from BettingInNJ, which says to flat-bet until 500 or more tracked bets show that your higher-confidence picks actually win more often. It's a fair point, and for most prop bettors I think it settles the question. A confidence scale only helps if your confidence predicts results, and until your log proves that, the tiers just decide how much you lose when you're wrong.

Kelly deserves a closer look, because it shows why props need restraint. The formula is f* = (bp − q) / b, where b is decimal odds minus 1, p is your true win probability and q is 1 − p. At -110 with a 55% chance, full Kelly says 5.5% of bankroll, $55 on $1,000. At a 52.5% chance the same formula says about 0.245%, or $1.70. So a 2.5-point mistake in your estimate, well within the error band of a 50-game sample, means betting over twenty times the right amount. And the Wizard of Odds Kelly explainer adds that betting double full Kelly takes long-run growth to zero, and anything beyond that turns it negative even with a positive edge. When your probability is a projection with wide error bars, fractional Kelly is the only kind that makes sense. If you want the full mechanics, the piece on sizing bets with Kelly covers them.

There's one more Kelly adjustment the generic guides skip. When you place several independent props at once, Wizard of Odds divides the single-bet Kelly by the square root of the number of bets: 0.5x Kelly each for 4 bets, 0.33x for 9, 0.25x for 16. Prop bettors are often the ones firing 16 bets a night.

You can run the same EV checks against your own prices in the prop edge calculator, which takes out the vig, turns a projection into a fair price and gives the Kelly stake.

The mistakes that break prop bankrolls aren't mainly math mistakes. They come from mixing up a feeling with a number.

Confidence isn't edge. Confidence is how sure you feel. Edge is the gap between your probability and the no-vig price, minus whatever your estimate could be off by. A player who's hit his over in eight of his last ten feels like a 2-unit play. The standard error on a 10-game sample says your real read could be off by 30 points either way. SharpAi sums up the 10% to 20% "lock" bet in four words: "There are no locks."

Unit creep after a hot run. A good fortnight makes your $20 prop unit feel timid, so it becomes $30, then $40, with no change in what you actually know. BettingInNJ's fix is to leave the unit alone after wins and withdraw profit back to the starting bankroll level. WhopReviews goes further and keeps the unit fixed for the whole season. Either rule works. What doesn't work is resizing based on how the last week felt.

Chasing losses by raising stakes. Each prop is settled on its own; last night's under has nothing to do with tonight's over. DeucesCracked notes a 55% bettor can still hit losing streaks of 5, 8 or even 10 bets. At a 2% unit a 10-bet streak costs 20% of the bankroll, and at 10% it costs everything, per BettingInNJ. Doubling up into a streak takes you from the first number to the second.

Treating a same-game stack as separate bets. Three props on one game at a unit each is one bet of up to three units on one outcome. The correlation guidance above exists because of this.