Any single bet can lose. Expected value is what you would earn on average if you could place the same bet thousands of times. Betting only when EV is positive is the mathematical definition of an edge: you will still lose plenty of individual wagers, but the average outcome tilts in your favor, and over a large sample the results converge on that average. It is the same principle a sportsbook uses against its customers, run in the other direction.
Suppose the fair probability of a side, taken from a devigged sharp market, is 52%. Your book is offering that same side at +105. Is the bet +EV?
# +105 in decimal odds = 2.05 (return $2.05 per $1 staked, incl. stake) fair_win_prob = 0.52 decimal_odds = 2.05 EV per $1 = (fair_win_prob * decimal_odds) - 1 = (0.52 * 2.05) - 1 = 1.066 - 1 = +0.066 # +6.6% expected return
The bet returns about 6.6 cents of profit per dollar on average. It is +EV because you are being paid +105 for something that fairly deserves roughly -108. Note the dependency: the whole calculation rests on the fair probability, so a +EV read is only as trustworthy as the devig behind it.
ParlayAPI scans for these spots directly. The /v1/sports/{sport_key}/ev endpoint compares each book's price to a devigged fair value and returns the outcomes trading above fair, with the edge percentage attached. The fair value is built from consensus across 45+ sources (as of 2026-08), and to keep phantom edges out, ParlayAPI guards the fair anchor against lopsided exchange asks and stale outliers before crowning any bet +EV. Live coverage spans 90+ sports and leagues; see /v1/sports for the current list.