Oct 8, 2026

What Is Positive EV Betting? The Formula, Break-Even Rates and Where Edges Disappear

The quick answer

Formula: EV per dollar staked = p x d - 1, where p is your win probability and d is the decimal price.
Break-even: a bet needs to win 1 / d of the time. At -110 that's 52.38%.
Worked example: at decimal 2.10, a 50% chance is +5% EV and a 46% chance is -3.4%, at the same price.
Variance: 100 bets at -110 with a 2% edge still finish in the red about 43% of the time.
Costs: vig and fees come straight off EV. A 52% estimate is -0.73% at -110 and +4.00% at a 50% price with no fee added, the cost structure of a pre-game straight trade on Novig.

Positive EV betting means only taking prices that pay more than your estimate of the outcome's chance justifies. It's a pricing question. A heavy favorite can be a bad bet and a longshot a good one, depending on what each pays against what it should.

Most of the work in the one-line formula goes into p, the probability you plug into it, and into the costs that eat a thin edge before it ever shows up in results.

The formula

For a simple win-or-lose bet, before fees:

EV per dollar staked = p x d - 1

Decimal price d is the total return per $1 if the bet wins, stake included. -110 is 1.9091, +150 is 2.50.

Take decimal 2.10. If the true chance is 50%, EV is 0.50 x 2.10 - 1 = +0.05, or +5% of every dollar staked. At 46%, EV is 0.46 x 2.10 - 1 = -0.034, or -3.4%.

Bets that can push or void need the calculation run across every settlement state, with the push returning the stake.

How we checked

Every number below is computed from the formula, with assumptions stated where a calculation depends on them, and the variance figures are exact binomial probabilities. The Kelly formula comes from Kelly's 1956 paper in the Bell System Technical Journal and Novig's fees from Novig's own fee page, all checked 7 October 2026.

What win rate do you need to break even?

Break-even probability is 1 / d. Beat it and the bet is +EV.

American odds	Decimal	Break-even win rate
-200	1.5000	66.67%
-150	1.6667	60.00%
-110	1.9091	52.38%
+100	2.0000	50.00%
+150	2.5000	40.00%
+200	3.0000	33.33%

At -110, a 52% hit rate on spreads loses money.

Where does p come from?

p comes from a reference market or a model.

Reference market: take a sharp, liquid market, remove its margin, and treat the result as the fair price. The +EV scanners covered below automate this method. If -150/+130 de-vigs proportionally to 57.98%/42.02%, a book offering the underdog at +145 (break-even 40.82%) is +EV against that reference. The estimate is only as good as the reference line, and several books often move on the same information, so agreement between them isn't independent evidence. A full walkthrough of de-vigging is in our companion piece on removing vig from odds.

Model: build your own probability from data. That gets you away from copying the market, and it adds new ways to be wrong: overfitting, information that wasn't available before the bet, and samples too small to test.

A scanner's "+EV" label rests on its reference line, so check which line that is.

How much estimate error erases an edge?

For every percentage point your p is off, EV moves by d percentage points.

At 2.10, each point of p is worth 2.1 points of EV. The 5% edge at a 50% estimate disappears at 47.62%, so an error of 2.38 points wipes it out.

At -110 it's tighter: a 2% edge needs p = 53.43% against a 52.38% break-even, so if your true p is 1.05 points lower than you think, the edge is gone.

An edge of about 2% leaves about a point of room, so the quality of the reference line decides whether the edge exists.

What does variance look like?

Assume 100 independent $1 bets, all at -110, each with a true 2% edge (p = 53.43%). Expected profit is +$2.00. The standard deviation is $9.52.

You need 53 wins to finish ahead. 52 wins loses $0.73. The exact binomial probability of 52 or fewer wins is 42.5%, so a bettor with a true 2% edge finishes those 100 bets down about 43% of the time.

Over 1,000 bets the expected profit is +$20.00, the standard deviation is $30.11, and the chance of finishing behind is still 24.7%.

A profitable month doesn't prove an edge, and a losing month doesn't disprove one. Closing line value (whether you're beating the price the market closes at) is a common diagnostic alongside profit, though it still measures you against a chosen benchmark and doesn't reveal the true probability.

How much should you stake?

The Kelly criterion, from John Kelly's 1956 paper, gives the bankroll fraction that maximizes long-run growth for a known edge. For a binary bet with net odds b = d - 1:

f* = (b x p - (1 - p)) / b

At 2.10 with p = 50%, b = 1.10 and f* = (0.55 - 0.50) / 1.10 = 4.55% of bankroll. At p = 46%, f* is negative: no bet. At -110 with a 2% edge, f* = 2.20%.

Kelly assumes you know p, and you don't. The fraction swings hard with small errors in the estimate, which makes it a theoretical upper bound. Fractional Kelly, betting some share of f*, is a common response, along with a cap on total exposure when several bets ride on the same game.

How do fees change EV?

Every cost comes straight off EV. The vig at a sportsbook is built into the price; a fee on a prediction market is charged on top. Either way, a thin edge has to clear it.

Here's a 52% estimate on an NFL side, bet three ways:

Where you bet	Price	Cost per $1 of payout	EV per dollar staked
Sportsbook	-110	$0.5238	-0.73%
Novig, pre-game straight, no fee	50%	$0.5000	+4.00%
Novig, live straight, 0.03 taker fee	50%	$0.5075	+2.46%

The same read is a losing bet at -110 and a +4% bet at a 50% price with nothing added. At 50%, the live taker fee is 0.03 x 0.50 x 0.50 = $0.0075 per contract, 1.5% of the price, which takes the edge from 4.00% to 2.46%.

The fee rows follow Novig's published schedule: no fee for maker or taker on pre-game straight trades and the 0.03 taker coefficient on live straights. Futures there carry 0.06 except golf and tennis, and parlays carry 0.10 built into the price.

The 50% price on a side you rate at 52% is an assumption for the example and comes from no venue's board; only the cost changes between rows.

+EV tools

Scanners such as OddsJam and RebelBetting flag prices that look better than the outcome's chances justify. RebelBetting builds its fair price from sharp bookmakers and sizes stakes with a Kelly-based system. You still have to check the reference line and the costs yourself.

What to look for

Price: convert to decimal and compare 1 / d with your estimate before anything else.
Reference: know which market your fair price came from and whether it was live and liquid when you checked.
Margin of error: work out how many points of p your edge can lose before it's gone.
Sample: judge the process over hundreds of bets, with closing line value alongside profit.
Costs: subtract the vig or fee on the exact bet type before calling anything +EV. A venue with no vig and no fee on pre-game straights leaves the estimate as calculated.

Positive EV FAQ

What is positive EV betting?

Taking prices that pay more than the outcome's probability justifies. EV per dollar is p x d - 1, and a bet is +EV when that's above zero.

How do you calculate EV on a bet?

Multiply your win probability by the decimal price and subtract 1. At 2.10 and 50%, that's +5%.

What win rate do you need at -110?

52.38%. A bettor hitting 52% at -110 loses money over time.

What is the best app for EV positive NFL betting?

The one with the lowest cost on the bet you're placing, because costs come straight off EV. On pre-game straight NFL bets, Novig charges no fee and builds no vig into the price, so nothing comes off the EV you calculate. At -110 you need to win 52.38% just to break even.