Practitioner course in algorithmic trading
Position Sizing and the Kelly Criterion
How much to bet: the Kelly criterion and its derivation, why full Kelly is dangerous with estimated edges, fractional Kelly, volatility targeting, and sizing across several strategies.
The Kelly criterion answers a narrow question exactly: what fraction of capital to risk on a bet with known odds so that wealth grows as fast as possible over the long run. Traders do not know their odds, which is why the interesting part of this course is what to do instead.
What does the Kelly criterion say?
For a bet that wins with probability p and pays b to one, the Kelly fraction is p minus the losing probability divided by b. For a strategy with expected excess return and variance, the Kelly leverage is the mean divided by the variance. Bet more than this and growth falls; bet twice as much and long-run growth above the risk-free rate falls to zero.
Why do traders use fractional Kelly?
Because the edge is estimated, not known, and estimates of the mean return are noisy. Overestimating the edge pushes you past the Kelly point, where risk rises and growth falls. A fraction of Kelly gives up a little growth for a large reduction in drawdown. The course quantifies the trade-off by simulation.
Who is this course for?
Traders with a tested strategy who need to decide its size, and anyone allocating capital across strategies. It counts toward the Certificate in Algorithmic Trading and the Professional Diploma in Algorithmic Trading.
What you will be able to do
- Derive the Kelly fraction for a simple bet and for continuous returns.
- Show how estimation error in the edge changes the right fraction.
- Size positions by volatility target and by fractional Kelly, and compare the results.
- Set drawdown limits and explain their effect on long-run growth.
- Allocate capital across several strategies with correlated returns.
Syllabus, module by module
Why sizing matters
The same strategy at different sizes: growth, ruin and the path between.
The Kelly criterion
Kelly's 1956 result, log utility and the derivation for a binary bet.
Continuous returns
The Kelly leverage for a strategy with a mean and variance, and its link to the Sharpe ratio.
Estimation error
What happens when the edge is overestimated, and why practitioners use a fraction of Kelly.
Volatility targeting
Sizing by forecast risk, rebalancing and the behaviour in volatility spikes.
Several strategies
Correlation, the multi-asset Kelly solution and practical constraints.
How the certificate is earned
An assessed project: you size an existing strategy under three methods, simulate the range of outcomes including estimation error, and submit the code with a recommendation and its reasoning. On a pass you receive the Course certificate in Position Sizing and the Kelly Criterion, issued by the Institute with a verification code. Read how certification works.
Before you start
Probability to expectation and variance, and working Python.
This course counts toward
- Certificate in Algorithmic TradingSix courses and a capstone that take a strategy from idea to a costed, sized and validated system, with performance reported the way a trading desk would read it.
- Professional Diploma in Algorithmic TradingThe top credential on the practitioner path. Eight courses and a 40-hour capstone covering everything between a trading idea and a book you could defend to a risk committee.
Questions applicants ask
Should I use the full Kelly criterion?
Almost never in trading. Full Kelly assumes you know your edge exactly. With an estimated edge, betting full Kelly often means betting more than Kelly, which lowers growth and produces deep drawdowns. Fractions such as a half or a quarter are common.
Who invented the Kelly criterion?
John L. Kelly Jr., a researcher at Bell Labs, published it in 1956. Edward Thorp later applied it to blackjack and to investing.