Find the mispriced bets
sportsbooks don't want you to see.
Fairline builds quantitative models for MLB, NHL, and the Premier League, converts them into fair odds, and flags every game where the sportsbook is giving you a mathematical edge. No insider tips. No tout picks. Just math.
Sportsbooks price every line with a hidden tax.
Every bet at a sportsbook comes with a built-in house edge, called the vig or juice. On a typical coin-flip line at -110/-110, the book takes about 4.5% off the top. Flip a fair coin all day at those odds and you still lose money.
The only way to beat that is to find bets where the book's price is wrong by more than their cut. That means building your own model of the game, calculating what the odds should be, and betting only when reality disagrees with the book enough to overcome the vig.
That's what Fairline does. Every day, for every game.
Three steps, every morning, every game.
Model the game
We pull box scores, rosters, lineups, weather, arenas, and travel data from free public APIs and rebuild each sport from first principles: Poisson goal rates for hockey and soccer, pace-adjusted efficiency for basketball, run environments for baseball.
Compute fair odds
The model outputs a win probability for every side of every market: moneylines, spreads, totals, team totals, alternates. We convert each probability into fair odds, the price at which that bet would break even forever.
Flag the value
We fetch live prices from Fanatics, DraftKings, and Kalshi, then pick the best price on each side and compare it against our fair odds. When the best book is worse than fair by more than 3%, we flag it as a value play and size it automatically based on edge.
From a one-line scan to the full model.
The dashboard lists every value play as a single scannable line. Open any game for the full projection behind it: fair odds and win probability for every side, the projected score, and a price-by-price comparison against each book we pull.
One line per edge.
Each row is a single bet the model likes, sorted by edge. Reading left to right: the edge, the bet and the book holding the price, the book's live odds, the model's fair odds, and the stake.
Edge %: how far the book's price beats fair. 3%+ gets flagged; brighter green means more.
Book: the live sportsbook price, tagged with the book offering it.
Fair: the model's break-even price. Worth a bet only when the book beats it.
Units: suggested stake, scaled by the size of the edge.
The full model behind the number.
Click any game and the projection opens in full: play-to fair odds and win probability for both sides, the projected score, and a grid of the model against every book, with the edge marked on each price.
Play-To Fair Odds
Sportsbooks vs Model
6:24 PM| MODEL | FAN | FD | DK | KAL | |
|---|---|---|---|---|---|
| Boston Bruins | -178 | -125 +8.4% / 1u | -130 +7.5% / 1u | -135 +6.6% / 0.5u | -140 +5.7% / 0.5u |
| Toronto Maple Leafs | +178 | +105 | +108 | +112 | +118 |
Green tags mark where a book's price beats the model's fair odds. The best price on Boston here is Fanatics at -125.
Six terms you need to read this site.
Skim these once and every number on the dashboard makes sense.
American odds
-110 +145The price format used by US sportsbooks. Negative numbers show how much you risk to win $100. Positive numbers show how much you win on a $100 bet. -110 means risk $110 to win $100. +145 means risk $100 to win $145.
Implied probability
-150 → 60%Every price converts to a win probability. A -150 favorite is saying there's a 60% chance of winning. Our model produces its own probability for the same outcome, and that's where disagreements show up.
Fair odds
52% → +92What the price would need to be for a bet to break even long-term. If our model says a team wins 52% of the time, fair odds are +92. Any price better than +92 is a positive-EV bet.
Play-to odds
stop at +115The worst American price we'll still take. Anything better than this is a value play. Anything worse and the edge is gone, so walk away. This is how you discipline yourself against chasing a line that has moved.
Edge %
+4.8%How much mathematical advantage you have over the book, in percentage points. 3% is the minimum we flag. 5%+ is a strong play. 10%+ is rare and usually a signal that something (injury, lineup) hasn't been priced in yet.
Units
1.5uA standardized bet size, usually 1% of your bankroll. We recommend 0.5u to 2u based on how much edge a play has. Talking in units instead of dollars keeps staking discipline separate from the size of your wallet.
Three sports. Three different models. One dashboard.
Major League Baseball
Pitcher-driven run environment with park factors, weather, and bullpen fatigue. Backtested at 56.6% accuracy with a positive Brier skill score across the 2025 season.
National Hockey League
Poisson goal-rate framework with starting goalies, arena altitude, rest days, and travel distance. Flags +alt lines when the main is a no-line value.
English Premier League
Dixon-Coles adjusted Poisson, the classic football model, with a tau correction that shifts draw probability by 2-3 points. Historically the highest-Sharpe edge in soccer betting.
What makes Fairline different.
We don't copy book lines.
Sportsbooks ban users who reverse-engineer their prices. We build the model independently, only comparing against books at the end. Every number is derived from public stats.
Every projection is backtested.
We replay the models against historical seasons before trusting them. If a methodology doesn't beat the vig on paper, it never ships to production.
You see the model, not a tipster.
We show you the fair odds, the edge, and our confidence, and we show you when the model has no edge, too. No paywalled picks, no Discord, no bro-speak.
Responsible by default.
Unit-based sizing by edge. No martingale. No chase-betting. If there's nothing to bet, the dashboard will tell you that, and that's an acceptable answer.
Ready to bet the math, not the vibes?
The dashboard is free to read. Sign in only when you want to log and track your own bets.