Kelly Criterion for MLB Betting: Staking Plans That Work
Available in US
The moment I stopped sizing my bets by gut feeling and started using a formula, my annual return jumped by four percentage points. That formula was the Kelly criterion — a mathematical framework originally developed for information theory in the 1950s that tells you exactly how much of your bankroll to risk on any given wager, provided you can estimate two things: the probability of winning and the odds being offered. It sounds clinical, and it is. But in a sport that produces 2,430 regular-season games per year, clinical is exactly what you need.
Most punters I know fall into one of two camps. Some bet the same flat amount on every game regardless of confidence level, which is safe but leaves money on the table. Others size their bets based on how strongly they feel about a particular matchup, which introduces emotional bias that compounds over hundreds of wagers. The Kelly criterion sits between these approaches — it is a mathematically optimal staking plan that maximises long-term bankroll growth while managing the risk of ruin. The catch is that it requires honest inputs, and honesty about your own edge is the hardest skill in baseball betting.
The Kelly Criterion Formula Explained
Have you ever tried to work out how much of your bankroll a single bet is actually worth? The Kelly formula gives you a precise answer. It is expressed as: f = (bp – q) / b, where f is the fraction of your bankroll to wager, b is the decimal odds minus one, p is your estimated probability of winning, and q is the probability of losing (which is simply 1 minus p).
A worked example makes this concrete. Suppose you estimate a team has a 58% chance of winning a game, and the sportsbook offers decimal odds of 1.90. In Kelly terms, b = 0.90, p = 0.58, and q = 0.42. Plugging in: f = (0.90 x 0.58 – 0.42) / 0.90 = (0.522 – 0.42) / 0.90 = 0.102 / 0.90 = 0.113. The formula says to wager 11.3% of your bankroll on this bet.
That 11.3% is the “full Kelly” recommendation — the stake that theoretically maximises the long-term growth rate of your bankroll. In practice, almost nobody uses full Kelly. The reason is variance. Full Kelly produces aggressive stake sizes that lead to stomach-churning drawdowns, particularly in a sport where even strong favourites lose 40% of the time. A seven-game losing streak on full Kelly stakes can erase 30-40% of your bankroll, and recovering from that hole requires both time and nerve that most bettors do not have.
Full Kelly vs Fractional Kelly for Baseball
I learned the hard way that full Kelly is a theoretical ideal, not a practical strategy. During my second season, I ran full Kelly stakes across 180 MLB bets. The theoretical bankroll growth was spectacular on paper. The actual experience was unbearable — I endured a 35% drawdown in a single month that took six weeks to recover from. The maths eventually proved itself correct, but the emotional cost nearly made me quit.
Fractional Kelly — typically half-Kelly or quarter-Kelly — is what serious MLB bettors actually use. Half-Kelly simply divides the full Kelly recommendation by two. In the example above, instead of wagering 11.3% of your bankroll, you would wager 5.65%. Quarter-Kelly would be 2.83%. The trade-off is straightforward: fractional Kelly achieves roughly 75% of the theoretical growth rate of full Kelly (at half-Kelly) while reducing maximum drawdowns by roughly 50%. For a sport with 162 games per team, that smoother ride is worth the slightly lower ceiling.
Baseball’s structure makes fractional Kelly particularly well suited. With games running daily from April through October, you are placing hundreds of bets per season. The law of large numbers works in your favour over that volume — a half-Kelly approach across 300 bets produces a remarkably stable growth curve, whereas the same approach applied to 30 NFL bets would show far more noise. The volume is the reason Kelly works in baseball better than in almost any other betting market.
When Flat Staking Beats Kelly
Kelly is only as good as your probability estimates, and this is where most bettors overestimate their own abilities. The formula assumes you can accurately assess the true probability of each outcome. If your estimates are consistently off by even 3-5%, the Kelly recommendations become unreliable — and you end up over-staking losers and under-staking winners.
US sportsbooks averaged a 10.15% hold rate in 2025, which means the market’s collective pricing is already quite efficient. Beating the market requires an edge that exceeds the built-in margin, and accurately quantifying that edge is genuinely difficult. If you are not confident in your probability estimates — and early in your betting career, you should not be, flat staking is the safer choice. A flat stake of 1-2% of your bankroll per bet eliminates the risk of catastrophic over-staking on a misevaluated game.
I recommend flat staking for anyone in their first two seasons of MLB betting. During that period, your primary goal should be building a track record and developing your handicapping process, not optimising bankroll growth. Once you have 500-plus tracked bets and a demonstrable edge, measured by consistent closing-line value or a statistically significant positive ROI, transitioning to fractional Kelly makes sense. Before that threshold, flat staking protects you from the most dangerous input in the Kelly formula: overconfidence.
Tracking ROI Across an MLB Season
The question I hear most often from newer bettors is some variation of: “I am up 15 units after 80 bets, am I actually good, or just lucky?” The honest answer requires a larger sample than most people expect. Statistical significance in sports betting is notoriously elusive. A positive ROI over 80 bets tells you almost nothing, the standard deviation on an 80-bet sample is wide enough that a break-even bettor could easily show a 15-unit profit through variance alone.
The threshold I use is 400 bets at consistent stakes. At that volume, a 3% ROI or higher becomes statistically meaningful, unlikely to be the product of chance. Below 400 bets, I treat my results as provisional and avoid adjusting my staking plan based on short-term performance. This discipline is unglamorous, but it protects against the most common mistake in staking: increasing bet sizes after a hot streak and decreasing them after a cold one, which is the exact opposite of what the maths prescribes.
I track every bet in a spreadsheet with five columns: date, game, odds taken, closing odds, and outcome. From these five data points I calculate cumulative ROI, closing-line value, and the Kelly-recommended stake for each bet (retrospectively, using my pre-game probability estimate). At the end of each month, I compare the actual performance to what the Kelly model predicted. When the two diverge significantly, it tells me either my probability estimates need recalibration or I am experiencing a variance swing that will correct over time. Both pieces of information are useful, and neither is available without disciplined tracking.
Is the Kelly criterion practical for casual MLB bettors?
For purely recreational bettors placing a handful of bets per week, flat staking of 1-2% per bet is simpler and nearly as effective. Kelly becomes practical once you have a demonstrable edge over at least 400 tracked bets and confidence in your probability estimates. Without those inputs, the formula can do more harm than good.
What sample size do I need before trusting my MLB ROI figures?
Most statisticians and professional bettors consider 400 to 500 bets at consistent stakes the minimum threshold for meaningful results. Below that, short-term variance dominates. A 5% ROI over 100 bets is interesting but not conclusive; the same ROI over 500 bets is a much stronger signal of genuine skill.
Written by the editors at Online Betting mlb.
