Kelly Criterion for Sports Betting Changed How I Manage Bankroll
I lost $3,400 in eight weeks betting flat $100 units on NHL underdogs before I learned the Kelly criterion for sports betting actually works. The math looked boring. Everyone in the forum kept saying just bet 2% or 3% of your roll. I ignored Kelly because I thought my edge was bigger than it was. My bankroll dropped 68% before I tracked my actual win rate versus what I thought my edge was. The Kelly formula told me I was betting nearly four times too large on plays with thin edges. Once I started using fractional Kelly at 25% of the full recommendation, my bankroll stopped hemorrhaging and actually grew by $1,850 over the next twelve weeks on the same types of bets.
What the Kelly Formula Actually Tells You
The Kelly criterion is not magic. It is a formula that tells you what percentage of your bankroll to risk based on your edge and the odds you are getting. The formula is: (bp – q) / b, where b is the decimal odds minus one, p is your probability of winning, and q is your probability of losing. If you think a team has a 55% chance of winning but you are getting +120 odds, Kelly tells you to bet 8.33% of your bankroll. That is the mathematically optimal size to maximize long-term growth without going broke.
The problem is nobody actually knows their true edge. I thought I had a 7% edge on those NHL dogs. Turns out over 240 tracked bets, my actual edge was 2.1%. The Kelly Calculator Sports tool showed me that at my real edge, I should have been betting 1.8% per play, not the 10% I was throwing around. The difference between perceived edge and actual edge will destroy your bankroll faster than bad beats.
My Actual Results Using Full Kelly vs Fractional Kelly
| Bet Sizing Method | Starting Bankroll | Bets Placed | Ending Bankroll | Max Drawdown |
|---|---|---|---|---|
| Flat $100 units | $5,000 | 82 | $1,600 | -68% |
| Full Kelly (overestimated edge) | $5,000 | 64 | $2,200 | -56% |
| Quarter Kelly (realistic edge) | $5,000 | 118 | $6,850 | -22% |
The quarter Kelly approach gave me room for error on my edge estimation. Even when my win rate dipped during cold streaks, the smaller bet sizes kept me alive. Full Kelly with an overestimated edge is a fast track to ruin. The math does not care about your confidence.
Why Most Bettors Fail at Kelly Implementation
Everyone screws up Kelly the same way. They plug in their gut feeling about win probability instead of actual tracked data. I kept a spreadsheet over six months of every NBA totals bet I made. My perceived edge on those bets was around 6% based on my model. My actual closing line value averaged 1.8%. When I recalculated my Kelly stakes using real performance data instead of model output, my recommended bet size dropped from 4.2% to 1.1% of bankroll.
The second mistake is not adjusting as your bankroll changes. Kelly is a percentage of your current roll, not your starting roll. If you start with $10,000 and drop to $7,500, your bet size needs to shrink. I watched a guy in our local group keep betting $300 per game because that was 3% of his starting $10,000. By the time he was down to $4,200, those $300 bets represented 7.1% of his remaining capital. He went broke three weeks later. Recalculate your stakes every week minimum.
Common Kelly Mistakes and Their Cost
| Mistake | Example Scenario | Cost Over 100 Bets |
|---|---|---|
| Using gut edge not tracked edge | Betting 5% thinking you have 8% edge when real edge is 2% | -$4,800 on $10k roll |
| Not adjusting for bankroll changes | Betting fixed dollar amounts as roll shrinks | -$3,200 additional vs proper sizing |
| Full Kelly instead of fractional | Betting 6% per play with 3% true edge | -45% drawdown vs -18% with half Kelly |
| Ignoring correlation between bets | Betting 4% on three same-game parlays | Effective 12% exposure, violates Kelly |
That last one killed me during NFL season. I would find three player props in the same game, run Kelly on each separately, and bet all three. The bets were correlated. If the game went under, all three props usually missed. I was risking 9% of my roll on outcomes that moved together. Actual Kelly says treat correlated bets as a single wager.
The Math Behind Kelly Sizing Nobody Explains
Kelly maximizes the logarithm of your bankroll growth. That sounds academic but it matters because log growth means you care more about avoiding ruin than hitting home runs. A 50% loss requires a 100% gain to recover. Kelly inherently protects against that by sizing bets smaller as your edge gets thinner. When I started using an EV Calculator to verify my actual edge on each bet type, I found my MLB first inning bets had a 4.2% edge while my NHL puck line bets had just 0.8% edge. Kelly told me to bet five times more on the MLB plays. Before Kelly, I was betting them equally.
The formula gets criticized because it assumes you know your exact win probability. You don’t. Nobody does. That is why fractional Kelly exists. Betting half Kelly means you are betting as if your edge is half what you calculated. Quarter Kelly assumes your edge is one-fourth. Over a twelve-week period I tracked quarter Kelly versus half Kelly on 340 MLB bets. Quarter Kelly had a max drawdown of 19%. Half Kelly hit 34%. Both made money but the ride was completely different. For context, check out Betting Data Lab for real tracking data on edge estimation accuracy across different sports.
Kelly Sizing at Different Edge Levels
| Your Actual Edge | Odds (+110) | Full Kelly % | Half Kelly % | Quarter Kelly % |
|---|---|---|---|---|
| 1% | +110 | 0.5% | 0.25% | 0.125% |
| 3% | +110 | 1.4% | 0.7% | 0.35% |
| 5% | +110 | 2.4% | 1.2% | 0.6% |
| 8% | +110 | 3.8% | 1.9% | 0.95% |
Look at how small those numbers get. Even with an 8% edge, which is massive in sports betting, quarter Kelly only puts down less than 1% of your roll. That feels too small when you are confident. The math does not care about your feelings. I learned this when I had a hot streak on college basketball and started bumping my stakes because I felt dialed in. Went 11-3 over eight days, got cocky, increased bet sizes to half Kelly from quarter Kelly. The next 22 bets went 9-13. The bigger sizing during the cold streak cost me $1,940 more than if I had stayed at quarter Kelly.
Where Kelly Criterion Fails in Sports Betting
Kelly assumes unlimited betting opportunities and perfect knowledge of your edge. Sports betting gives you neither. You get maybe 10-15 bets per week that meet your criteria if you are selective. Kelly also breaks down when bookmaker limits kick in. I had a solid edge on MLS totals for about nine weeks. My Kelly stake climbed to $420 per bet as my roll grew. The book limited me to $85 max. At that point Kelly is useless because you cannot scale with your bankroll.
The other failure point is psychological. Kelly will tell you to bet $680 on a single play if your edge is big enough and your roll is large enough. Most bettors cannot pull the trigger on a bet that size even when the math says to. I have had Kelly recommend 6.2% of my roll on a bet with a massive line error. I bet 2% because I could not handle the mental weight. Kelly is mathematically optimal but emotionally brutal. There is no shame in betting smaller than Kelly recommends. There is massive shame in betting larger.
Simulated Results Over 1,000 Bets
| Strategy | Starting Roll | Median Final Roll | Risk of Ruin | 95th Percentile Drawdown |
|---|---|---|---|---|
| Flat 2% betting | $10,000 | $14,200 | 8% | -32% |
| Full Kelly (perfect edge) | $10,000 | $28,400 | 12% | -61% |
| Half Kelly | $10,000 | $19,800 | 4% | -38% |
| Quarter Kelly | $10,000 | $16,100 | 1.2% | -24% |
That simulation assumed a 3% edge at -110 odds with perfect edge knowledge. Full Kelly had the highest median outcome but also a 12% chance of going completely broke and insane drawdowns. Quarter Kelly sacrificed some upside for dramatically lower ruin risk. In real betting where your edge estimate is wrong half the time, full Kelly is suicide. The drawdowns will break you mentally before they break you financially.
How to Actually Implement Kelly Without Blowing Up
Start by tracking every single bet for at least 200 plays. Record the odds, your perceived edge, the outcome, and whether you beat the closing line. Use that data to calculate your real edge by bet type. I found my player prop edge was 2.8% but my spread edge was only 1.1%. Kelly sizing should differ for each category. The ROI Calculator helped me separate which bet types actually had positive expectation versus which ones I was fooling myself about.
Use quarter Kelly or even smaller. I settled on one-fifth Kelly after testing different fractions over eight months. My edge estimation is not precise enough for anything more aggressive. One-fifth Kelly with a 3% true edge means betting 0.48% of your roll per play at -110 odds. On a $10,000 bankroll that is $48 per bet. It feels tiny but it survives cold streaks. I went through a 12-26 stretch on NFL totals and only lost $890 of a $12,400 roll because my bet sizing was conservative.
Recalculate your bankroll and bet sizes weekly. Every Sunday night I update my spreadsheet with my current roll and adjust all my bet sizes. When my roll grows, stakes grow. When it shrinks, stakes shrink. This compounding works both ways and it is the core of what makes Kelly powerful over time. The bettors who go broke are the ones who keep firing $500 bets when their roll is down to $3,000 because that is what they started with.
FAQ
Should I use full Kelly or fractional Kelly for sports betting?
Use fractional Kelly unless you have thousands of tracked bets proving your exact edge. Quarter Kelly or one-fifth Kelly gives you room for error when your edge estimate is wrong. Full Kelly will destroy your bankroll during inevitable cold streaks when your real edge is lower than you think.
How do I calculate my actual edge for the Kelly formula?
Track at least 200 bets of the same type and calculate your win rate versus the implied odds. If you win 54% of -110 bets, your edge is roughly 2.7%. Compare your opening line to the closing line to see if you are consistently getting value. Without tracked data you are guessing.
What bankroll size do I need to use Kelly criterion effectively?
You need enough that your minimum bet size is at least $10-$20 when using fractional Kelly. With quarter Kelly and a 2% edge at -110 odds, you need around $5,000 to make $25 bets. Smaller rolls force you into bets that are too large relative to the math.
Explore more strategies in our Football First Half Betting Strategy: How I Found Value and Lost $3,400 Before Fixing My Approach.