Lab: free tools · size first · stock scanner · crypto scanner · journal
Kelly Criterion for Crypto Position Sizing
The Kelly Criterion is a mathematical formula for optimal bet sizing that maximizes the long-run growth rate of a portfolio. It calculates the ideal percentage of capital to risk on each trade based on win probability and win/loss ratio. In crypto trading, the Kelly Criterion helps traders avoid both under-betting (leaving returns on the table) and over-betting (risking ruin from drawdowns).
The Kelly criterion calculates the bet size that maximizes long-term growth, given your edge and the odds. In trading, the inputs are guesses, and overestimating your edge leads to oversized bets that can wipe out an account. Many traders who use Kelly use a fraction of it, often half or less. It is a better reminder of why sizing matters than a formula to apply blindly.
The formula
John Kelly published the idea in 1956 in the Bell System Technical Journal, in a paper on information theory. For a simple bet that wins b times your stake with probability p, the Kelly fraction is p minus the losing probability divided by b. Say a setup wins 55% of the time and pays the same as it risks. Kelly says bet 0.55 minus 0.45, or 10% of the account. If the same 55% setup paid 2 to 1, the answer is 0.55 minus 0.45 divided by 2, about 32.5%.
Why betting the whole amount hurts
If your real edge is half what you think, betting the Kelly amount for the edge you imagined puts you at about twice true Kelly, and at twice Kelly the expected growth rate of the account falls to roughly zero while the swings are huge. Crypto adds fat tails: a gap through a stop or an exchange failure can lose far more than the planned risk. Leverage makes this worse, because the loss on a bad trade isn't capped at the stop.
Using a fraction of it
Half Kelly keeps about three quarters of the long run growth rate and roughly halves the swings in the account. In the 55% even money example, that means risking 5% a trade.
Getting the inputs
Use a large sample of real trades, recorded the same way, and include fees and slippage. Estimate the win rate and payoff for each setup separately, since a blended figure hides weak ones. Recheck as markets change; an edge measured in a bull run may disappear in a range. If the formula ever tells you to bet a large share of the account, assume your inputs are wrong.