How EV Sports Betting Works

For a comprehensive overview, see our full EV betting guide.

What Is +EV Betting?

A positive expected value (+EV) bet is one where the odds offered are better than the true probability of the outcome. Over hundreds of bets, +EV bets are mathematically expected to profit regardless of any individual result.

The key insight: you don't need to predict who will win. You just need to find where a sportsbook is offering better odds than they should be.

How We Calculate Fair Odds

Every sportsbook builds a profit margin (called the vig or juice) into their odds. A "fair" line is what the odds would be with zero vig — the true implied probability.

Here's how we find it:

Collect odds from multiple books — We track FanDuel, DraftKings, BetMGM, Caesars, BetRivers, and ESPN BET for every NHL and NBA game.
Remove the vig from each book — Convert each book's moneyline to implied probabilities, then normalize so they sum to 100%.
Average across books — The mean of all no-vig probabilities gives us a consensus fair price. This is our benchmark.

For example, if five books price a team between 38-42% after removing vig, our consensus fair probability might be 40%, which translates to fair odds of +150.

Why Sportsbooks Disagree

Sportsbooks set lines independently based on their own models, the action they're seeing, and their risk tolerance. This means:

FanDuel might have a team at +150
DraftKings has them at +135
BetMGM has them at +140

These differences create opportunities. If the consensus fair price is +135 and FanDuel is offering +150, that's a +EV bet at FanDuel — you're getting paid more than the true odds suggest.

A Concrete Example

Say the consensus fair probability for the Sacramento Kings to win is 31.3% (fair odds: +220). FanDuel is offering Kings at +250.

Fair implied probability: 31.3%
FanDuel implied probability: 28.6% (from +250 odds)
EV calculation: (0.313 × 3.50) - 1 = +9.6% EV

That's a 9.6% edge. The Kelly Criterion then sizes the bet proportional to this edge — maybe 1-2% of your bankroll. The Kings don't need to win this specific game — the math works over volume.

Leave-One-Out Consensus

We use a technique called leave-one-out (LOO) consensus. When evaluating whether FanDuel's line is +EV, we build the consensus from all books except FanDuel. This prevents circular logic — a book can't create its own edge by being different from a consensus it's part of.

Kelly Criterion Sizing

Not all +EV bets deserve the same stake. We use the Kelly Criterion (specifically half-Kelly for safety) to size each bet proportional to the edge:

Big edge (8%+ EV): Larger bet, around 2-4% of bankroll
Medium edge (4-8% EV): Moderate bet, around 1-2% of bankroll
Small edge (2-4% EV): Smaller bet, under 1% of bankroll

This ensures you risk more when the edge is bigger and less when it's marginal.

Our Backtested Results

We backtested this approach on the 2025-26 season using historical multi-book odds data:

NHL: 244 picks, 132-112 record, +21.3% ROI (p=0.003)
NBA: 444 picks, 175-269 record, +33.7% ROI (p=0.002)

Note the NBA win rate is only 39.5% — yet the ROI is +33.7%. This is because we target underdogs at plus-money odds. You don't need to win often when the payouts are large enough.

Both results are statistically significant and held up in out-of-sample testing (training on earlier months, testing on later months).

Why This Works Long-Term

This approach works because:

It's math, not prediction — We're not claiming to know who will win. We're finding where the price is wrong.
Market efficiency is imperfect — Sportsbooks sometimes misprice lines, especially on less popular markets.
Volume smooths variance — Any single bet can lose, but over hundreds of +EV bets, the expected value shows up in the results.
It's verifiable — We track every pick publicly. Closing Line Value (CLV) tells us whether lines moved toward our picks after we posted them.

What We Don't Claim

We don't guarantee profits on any individual bet
We don't predict game outcomes
We don't claim to beat the market with AI or models
Short-term losing streaks are normal and expected

Our edge is systematic and math-based. Over time, the results should converge toward the expected value — but variance is real, and responsible bankroll management is essential.

See today's +EV opportunities on our homepage or check our track record for live results.