The Kelly Criterion: A Prediction Market Trader's Guide to Position Sizing

Every prediction market trader eventually faces the same question: how much should I put on this trade? Bet too little and you leave money on the table. Bet too much and a single bad outcome can wipe out weeks of gains.

The Kelly Criterion is a mathematical formula that answers this question precisely. Developed by John Kelly at Bell Labs in 1956, it was originally designed for information theory — optimizing the growth rate of a gambler with an edge. Decades later, it remains one of the most powerful tools in a trader's arsenal.

The Formula

For a binary outcome — which is exactly what prediction market contracts are — the Kelly Criterion reduces to:

f* = (p × b − q) / b

Where:

• f* = fraction of your bankroll to wager
• p = your estimated probability of winning
• q = probability of losing (1 − p)
• b = the net odds received (payout / wager)

On Kalshi, contracts settle at $1.00 for YES or $0.00 for NO. If a YES contract is trading at $0.40, you pay $0.40 to win $0.60 in profit. So b = 0.60 / 0.40 = 1.5.

A Practical Kalshi Example

Suppose you're looking at a Kalshi contract asking "Will the next jobs report show unemployment above 4.2%?" The YES contract is trading at $0.35. After your research — reading BLS data, economist surveys, leading indicators — you believe the true probability is 50%.

Plugging into the formula:

• p = 0.50 (your estimated probability)
• q = 0.50
• b = 0.65 / 0.35 = 1.857
• f* = (0.50 × 1.857 − 0.50) / 1.857 = 0.231

Kelly says to wager 23.1% of your bankroll. That's aggressive — and this is where most experienced traders make an important adjustment.

Why Fractional Kelly Is Better in Practice

Full Kelly assumes your probability estimate is exactly right. In reality, nobody's probability estimates are perfect. Off by a few percentage points and full Kelly can lead to devastating drawdowns.

Most professional traders use half-Kelly (also called fractional Kelly) — they wager half of what the formula recommends. In our example, that means betting ~11.5% instead of 23.1%.

The tradeoff is elegant: half-Kelly captures roughly 75% of the theoretical growth rate while cutting variance dramatically. Your worst drawdowns get much smaller, and the ride is far smoother.

Some traders go even further — quarter-Kelly or third-Kelly — especially when they're less confident in their edge or trading correlated contracts.

When Kelly Says Don't Bet

One of the formula's most useful features: if f* comes out negative, it means you have no edge. Don't trade.

Suppose that same unemployment contract is trading at $0.55 and you think the true probability is 50%. Kelly gives:

• f* = (0.50 × 0.818 − 0.50) / 0.818 = −0.111

Negative. The market is priced above your estimate. Either bet the other side or stay out entirely. This discipline — letting the math tell you when to sit — prevents the most common mistake in prediction markets: overtrading positions where you have no real edge.

Applying Kelly Across a Portfolio

If you're trading multiple Kalshi contracts — say, unemployment data, Fed rate decisions, and weather events — you need to think about portfolio-level Kelly. The key constraint: your total allocation across all positions shouldn't exceed what single-position Kelly math would suggest for your overall edge.

In practice, this means:

1. Calculate Kelly fraction for each individual trade
2. Apply your fractional Kelly discount (half or quarter)
3. Sum all positions — if total exceeds 50-60% of bankroll, scale everything down proportionally
4. Consider correlation: two bets on related economic data aren't independent, so treat them as partially the same bet

Common Mistakes

Overestimating your edge. The most dangerous input to Kelly is your probability estimate. If you think a contract is 60% to hit but the true probability is 45%, Kelly will tell you to overbet massively. Always ask: "What would have to be true for the market price to be right?"

Ignoring correlation. If you hold YES on "inflation above 3%" and YES on "Fed raises rates," those aren't independent bets. A single economic report can move both. Kelly assumes independent outcomes — correlated bets need manual adjustment.

Using full Kelly. Almost nobody should use full Kelly. The formula maximizes long-run growth rate but with brutal variance. Half-Kelly is the pragmatic standard.

Not having an edge at all. Kelly can't create an edge — it can only size one that already exists. If your research doesn't give you a better probability estimate than the market, Kelly won't help you.

The Bottom Line

The Kelly Criterion turns prediction market trading from gut feel into structured decision-making. It forces you to be explicit about two things: what do I think the probability is? and how much of my bankroll should that conviction command?

Start with half-Kelly. Be honest about your probability estimates. Let the math tell you when to stay out. Over time, this discipline compounds — not just returns, but judgment.