Sports Betting Strategies: What the Math Says Works

A sports betting strategy that can win over time rests on two things you control: betting only when your estimate of a result beats the price, and staking a fixed, small share of your bankroll so a losing run does not end you. Everything else on this page, from line shopping to the Kelly criterion, is a way of doing one of those two things, and the staking systems that promise more (Martingale, Fibonacci, Labouchere) change when you win and lose, never the expected value of a bet.

Start with the price, because it hides the book’s cut

A price is a probability. At -110 the book is asking you to risk $110 to win $100, which implies a 52.38% chance (110 divided by 210). Post -110 on both sides of a game and the two add up to 104.76%. That extra 4.76% is the margin the book keeps, and it is why a bettor who picks winners exactly half the time at -110 loses about $4.55 for every $100 risked.

So the break-even line at -110 is 52.38%, not 50%. Our guide to how betting odds work has the conversion for every format.

Shop for the line on every bet

Two books can price the same side differently. Take a -105 instead of a -110 and your break-even drops from 52.38% to 51.22%. Over 500 bets that is the difference between needing 262 winners and needing 257.

Half a point on a spread works the same way. A -3 that becomes -2.5 turns a three-point win from a push into a winner. You can follow current NFL lines on our NFL predictions hub, and the point spread guide explains how spreads grade.

Bet only when the price is wrong in your favor

Value betting means one thing: your estimate of a result is higher than the probability the price implies. The expected value of a $100 bet is your chance of winning times the profit, minus your chance of losing times the stake.

A worked example. A team is +150, which implies 40.0%. If your own estimate is 45%, the bet is worth 0.45 x $150 minus 0.55 x $100, or +$12.50 per $100. If your estimate is 40%, it is worth nothing, and below that it loses. The hard part is the estimate, and a strategy that cannot say where its 45% came from is guessing.

Size every bet as a unit of your bankroll

A unit is a fixed share of your bankroll, set once. At 1%, a $1,000 bankroll bets $10 a unit, and a run of 10 straight losses costs $100, a tenth of the roll. At 10% a unit, the same run takes the whole thing. Our guide to betting units shows how to record results in units, so a $50 bettor and a $500 bettor can compare records.

Flat staking, the same unit on every bet, is the plain version. At $10 a unit and -110, no single loss costs more than $10 and every win pays $9.09, so one loss undoes one win and 91 cents more.

The Kelly criterion, if you trust your numbers

Kelly sizes each bet by your edge. The share of bankroll to bet is (b x p – q) / b, where b is the profit per $1 staked, p is your chance of winning and q is your chance of losing.

At +150 with a 45% estimate, b is 1.5, and Kelly says (1.5 x 0.45 – 0.55) / 1.5 = 8.3% of the bankroll. At -110 with a 55% estimate, it says 5.5%. Those are large bets for a number you estimated yourself, and half Kelly (4.2% and 2.75% here) halves each of them. If p is too high, full Kelly overbets, and an estimate at or below the break-even gives zero or a negative number: no bet.

Parlays pay less than their odds suggest

A two-leg parlay of -110 bets pays about +264. Two fair coin flips would pay +300, so the book’s margin compounds with every leg. The two-leg ticket implies 27.4% against a true 25% for two 50% legs. At three legs the parlay pays about +596 against a fair +700.

A parlay is not a bad bet when every leg carries an edge, since the edges compound too. Without an edge on each leg, it multiplies the margin. The teaser bet guide covers the related bet that moves the spread in exchange for a lower payout.

Arbitrage and hedging lock in a result

Arbitrage means backing both sides at different books when the two prices add up to less than 100%. Take +110 on one side (47.62%) and +105 on the other (48.78%): together 96.4%. Stake $100 on the +110 side, which returns $210, and $102.44 on the +105 side, which also returns $210. You have staked $202.44 and collect $210 either way, a profit of $7.56, or 3.7%. A price that moves before the second bet is placed turns the arb into an ordinary bet.

A hedge bets the other side of a position you already hold. Say you hold $100 on a team at +1000 to win a title, and it reaches the final with the opponent priced at -130. A bet of $621.74 on the opponent returns $1,100, so whoever wins, you finish $378.26 ahead. You give up the chance at $1,000 profit for a certain $378.26. Our hedge bet guide works through more cases, and matched betting applies the same arithmetic to sportsbook promotions.

Staking systems do not change the odds

Martingale doubles the stake after every loss. From $10, six straight losses cost $630 ($10, $20, $40, $80, $160 and $320), and the seventh bet must be $640 to win back $10. At 50% per bet, the chance of six straight losses is 1 in 64, or 1.6%. At -110 it is worse: a win after one loss nets $8.18, not the $10 the system promises, because the winning bet pays less than it risks.

The reverse systems press winners instead. The 1-3-2-6 system bets 1, 3, 2 and then 6 units after each win and resets after a loss. At even money, four straight wins make 12 units; a loss on the fourth bet leaves you at zero, and a loss on the first costs one unit.

Fibonacci and Labouchere are slower versions of Martingale. Fibonacci moves one step up the sequence 1, 1, 2, 3, 5, 8, 13 after a loss and two steps back after a win. Labouchere writes a line such as 1-2-3, bets the sum of the two ends (4 units), crosses both off after a win and adds the lost amount to the end after a loss; finishing the line wins its total, 6 units. None of the three changes the chance of any single bet. Each only rearranges when the losses arrive, and a long enough losing run beats all of them.

Handicap and spread bets are a price in points

A handicap gives one side a head start in points or goals so the market is closer to even. A favorite at -3.5 must win by four or more; the underdog at +3.5 covers with any loss by three or fewer, or a win. In soccer a quarter-ball line such as -0.25 splits the stake: half on -0.5 and half on 0, so a draw loses half and returns half.

The strategy question is the same as with any price: is your chance of covering higher than the price implies? Our guide to handicapping in betting covers the formats, and the beginners’ guide starts from the first bet. If you pay for picks, check the record first in our handicapper site reviews: a pick service at -110 has to hit 52.38% just to break even, before its fee.