Kelly Criterion in Horse Racing: How to Size Your Bets to Maximize Long-Term Profit

# Kelly Criterion in Horse Racing: How to Size Your Bets to Maximize Long-Term Profit **The Kelly Criterion is a mathematical formula developed by physicist John Kelly Jr. in 1956 that calculates the optimal percentage of your bankroll to wager on a bet with a known positive expected value. Applied correctly to horse racing, it prevents you from both overbetting — which leads to ruin even with a genuine edge — and underbetting — which leaves long-term profit on the table. It is the most mathematically rigorous bankroll management framework available to horse racing bettors, and the vast majority of bettors have never heard of it.** Most bettors manage their bankroll through intuition: bet more when confident, bet less when uncertain, adjust based on how recent sessions have gone. This approach produces inconsistent results not because the underlying analysis is wrong, but because the bet sizing is arbitrary. The Kelly Criterion replaces arbitrary with optimal — and the difference over hundreds of bets is substantial. ## The Kelly Formula The formula itself is straightforward: **Kelly % = (Edge × Odds + Edge) ÷ Odds** Or equivalently: **Kelly % = (bp − q) ÷ b** Where: - **b** = the decimal odds minus 1 (the net profit per unit staked) - **p** = your estimated probability of winning - **q** = your estimated probability of losing (1 − p) Let us apply this to a concrete horse racing example. You assess a horse has a 30% probability of winning. The live odds are 4-1 (decimal: 5.0, so b = 4). Your estimated edge is positive. Kelly % = (4 × 0.30 − 0.70) ÷ 4 = (1.20 − 0.70) ÷ 4 = 0.50 ÷ 4 = 12.5% The formula tells you to bet 12.5% of your current bankroll on this horse. If your bankroll is $1,000, the optimal bet is $125. This may seem aggressive. It is — and that is one of the reasons practitioners use fractional Kelly rather than full Kelly in most real-world applications. ## Why Full Kelly Is Often Too Aggressive The Kelly Criterion is mathematically optimal under specific assumptions: that your probability estimates are accurate, that you will bet consistently according to the formula across a large number of independent wagers, and that you are maximizing the long-term expected logarithm of your wealth. In practice, horse racing probability estimates are always uncertain. A 30% probability estimate carries estimation error — the true probability might be 26% or 34%, and you cannot know which with certainty. When the true probability is below your estimate, full Kelly overbets. Over a long series of such errors, full Kelly produces significant drawdowns that can be psychologically difficult to sustain. The most widely adopted solution is **fractional Kelly** — typically half Kelly or quarter Kelly. Using half Kelly, the 12.5% bet in the example above becomes 6.25% of bankroll. Using quarter Kelly, it becomes 3.125%. Fractional Kelly sacrifices some of the theoretical growth rate of full Kelly in exchange for significantly reduced variance and smaller drawdowns. For most practical bankroll management in horse racing, half Kelly is the standard recommendation: it provides most of the mathematical advantage of the formula while keeping drawdowns manageable. ## Setting Your Bankroll and Unit Size Before applying Kelly, you need to establish a dedicated betting bankroll — a sum of money that you can afford to lose entirely without financial hardship, separate from your regular finances, and held specifically for betting purposes. This segregation is not just psychological comfort — it is a practical requirement for Kelly to function correctly. If you commingle your betting bankroll with your regular spending account, you will not make rational Kelly-sized bets. When you are up, you will overbet. When you are down, you will either underbet (fearful of losing more) or overbet (trying to recover). Both behaviors destroy the mathematical advantage that the formula provides. A realistic starting bankroll for applying Kelly to horse racing depends on the number of races you plan to bet and the average edge you expect to generate. As a general principle, a starting bankroll should be large enough that the minimum trackside bet ($2 at most American venues) represents no more than 1% of the total. A $200 starting bankroll means a $2 minimum bet is already 1%, leaving little room for Kelly-sized bets at low edge levels. A $1,000 or larger bankroll gives the formula meaningful room to operate. ## Practical Application: Kelly at the Track The Kelly formula requires you to know two things before every bet: your probability estimate for the horse winning, and the current live odds. The live odds are observable. The probability estimate is the output of your handicapping process. Most bettors express their handicapping in terms of confident versus uncertain rather than specific probability numbers. Learning to assign rough probability ranges — 20-25%, 30-35%, 40-45% — is a skill that develops with practice. You do not need precision to the nearest decimal point. A rough probability range combined with the Kelly formula is significantly more rigorous than arbitrary bet sizing even if the estimate is imprecise. For a bettor using StrideOdds, the confidence score and edge in basis points provide the inputs needed to calculate a Kelly-sized bet. A horse with a confidence score of 88% and a +41 bps edge represents a specific implied probability relationship between the algorithm's assessment and the live market price — and that relationship is the input the Kelly formula requires. ## The Most Common Kelly Mistakes **Applying Kelly to negative expected value bets.** The formula assumes positive EV. If your probability estimate does not produce a positive Kelly percentage — meaning the bet has negative expected value — the formula says not to bet. The result of the Kelly calculation is zero or negative, which means pass on the race. This is one of the formula's most valuable features: it enforces discipline by giving you a mathematical reason to decline a wager. **Recalculating Kelly after each win or loss within the same race card.** Kelly is designed for sequential bet sizing over time. Updating your bankroll figure and recalculating Kelly for every bet on a multi-race card is theoretically correct but practically cumbersome. Most practitioners update their bankroll at the end of each session rather than mid-card. **Treating Kelly as a system rather than a tool.** The Kelly Criterion does not tell you which horses to bet. It only tells you how much to bet on the ones you have already assessed as positive expected value. The quality of your probability estimates determines whether the formula produces profitable results. Kelly amplifies good analysis and amplifies bad analysis in equal measure — with good analysis, the results compound favorably; with bad analysis, they compound unfavorably at a rate faster than flat betting would produce. ## Kelly and Long-Shot Betting One important insight from the Kelly Criterion is its treatment of longshots. When a horse has a high positive expected value — a large gap between your probability estimate and the implied probability from long odds — the Kelly formula recommends a larger percentage of bankroll than intuition would typically suggest. Many bettors are uncomfortable betting a substantial percentage of bankroll on a 20-1 shot, even one they believe to be a genuine overlay. The Kelly formula, correctly applied with an accurate probability estimate, says that discomfort is costing them money. The formula is indifferent to the absolute level of odds — it cares only about the relationship between your probability estimate and the market price. A 20-1 horse you assess at 12% probability is a much better Kelly bet than a 2-1 horse you assess at 33% probability, and the formula will tell you exactly how much better. Bankroll management in horse racing is the unglamorous half of profitable betting. It does not produce winners. It does not surface overlays. It does not tell you which horse to back in the Kentucky Derby. What it does is ensure that your edge — however small — compounds over time into meaningful profit rather than being destroyed by the natural variance of individual results. The Kelly Criterion is the most mathematically rigorous tool available for that purpose. --- *StrideOdds provides confidence scores and edge values in basis points for every race — the inputs you need to calculate Kelly-optimal bet sizes and manage your bankroll systematically. Join the waitlist at strideodds.ai.*