Losing bettors hunt for winners. Winning bettors track something else entirely: whether their odds beat the market’s final price, and whether their stake sizes match the size of their edge. This Sports CLV and Kelly workflow devigs closing odds to expose your real probability of winning, then sizes each bet with a fractional Kelly approach (run the numbers themselves in our Kelly Calculator).
Sports CLV + Kelly
Why Closing Line Value is the metric that separates pros from fish
The closing line is the last odds a sharp bookmaker publishes before kickoff — books such as Pinnacle or Bookmaker.eu are the usual references. By that point, the line has absorbed team news, injury reports, syndicate money, and thousands of individual wagers. No single bettor knows more about the game than that number does.
Beat it regularly and you own a demonstrable, statistical edge (**Closing Line Value**, or **CLV**). Bet worse prices than the close, time after time, and the vig grinds your bankroll down no matter how good your picks feel.
Two calculations: removing the vig, then sizing the stake
The process splits into independent steps. First, devig the closing odds to recover true probabilities. Second, feed those probabilities through the Kelly Criterion to get an exact betting fraction.
1. Devigging closing odds
Start with the total implied probability across both outcomes. Thanks to the vig, this always overshoots 100%:
Market Overround = (1 / Odds_A) + (1 / Odds_B)
Using the standard **multiplicative devigging method**, divide each side’s implied share by that total to get the fair probability. For outcome A:
True_Probability (p) = (1 / Odds_A) / Market Overround
A worked pair: sharp closing odds of 1.85 on Team A against 2.05 on Team B.
- Market Overround =
(1 / 1.85) + (1 / 2.05) = 0.5405 + 0.4878 = 1.0283(a 2.83% hold) - True_Probability Team A =
0.5405 / 1.0283 = 52.56%
2. Edge and expected value (EV)
Fair odds equal 1 divided by the true probability. If you managed to lock in a price above that level, the wager carries positive EV:
Edge = (True_Probability * Odds_Placed) - 1
3. The Kelly Criterion equation
The Kelly Criterion answers a narrower question than most bettors realize: what fraction of your bankroll maximizes long-run exponential growth on a positive-EV bet? Answering more than that — or less — costs you. The formula:
f* = Edge / (Odds_Placed - 1)
Here f* is full Kelly. Few professionals actually bet it. A cold stretch at full stakes can crater your balance fast, so the common practice is **Fractional Kelly** — one-quarter or one-half of the recommended fraction — trading some growth speed for survivable drawdowns.
Data Sandwich: Devigging a real-world underdog hedge
Concrete example. On Tuesday, you took Team A at 2.20 while the line was soft.
By kickoff, syndicate money has hammered the other side. The sharp book closes with:
- Team A: 1.95
- Team B: 1.95
Now audit your original ticket:
- Market Overround =
(1 / 1.95) + (1 / 1.95) = 0.5128 + 0.5128 = 1.0256 - True Probability Team A =
0.5128 / 1.0256 = 50.0% - Your Edge =
(0.50 * 2.20) - 1 = +10.0% - Full Kelly Stake =
0.10 / (2.20 - 1) = 0.10 / 1.20 = 8.33%of bankroll. - Quarter Kelly Stake (Recommended) =
8.33% * 0.25 = 2.08%of bankroll.
Beating the close by enough to devig into a 10.0% edge is exactly the kind of fill CLV measures. Staking just 2.08% per play keeps a normal losing streak from ending your season while still compounding the edge over hundreds of wagers.
Frequently asked questions
How do sharp sportsbooks differ from recreational ones?
Sharp operators like Pinnacle accept large bets and tolerate winning players; their enormous volume produces the most accurate closing lines in the industry. Recreational books — DraftKings and similar local sites — shadow those lines, pad the vig, and limit or ban anyone who beats them consistently.
Why is Full Kelly betting considered dangerous? (It can yield negative Expected Value (EV) on under-hedges)
Full Kelly compounds fastest over the very long run — but only if you survive the middle part. A short run of losses at full stakes routinely erases 50% to 70% of a bankroll. Scaling back to a fractional multiplier (0.25x or 0.50x is typical) cuts drawdown risk sharply while keeping most of the growth rate.
Can I have a negative CLV but still win a bet?
Of course. One bad-price bet can cash; one great-price bet can lose — variance does not read your tickets. Over roughly 1,000 wagers, though, the correlation takes over: persistent negative CLV means structural losses, persistent positive CLV means profit. The math has no exceptions clause.