Traders spend most of their research time on when to enter and exit. Yet two traders using identical signals can end with radically different outcomes purely because of how much they allocate to each trade. Position sizing is where strategy meets risk management.
Fixed fractional sizing
The simplest robust method risks a fixed percentage of current capital on each trade — often between 0.25% and 2% in practitioner literature. Because the amount scales with equity, losing streaks shrink position sizes automatically, slowing the rate of drawdown.
Volatility targeting
Volatility targeting sizes positions so each contributes a similar amount of expected volatility. A calm instrument receives a larger allocation than a volatile one. This approach is common in systematic futures and multi-asset strategies because it keeps portfolio risk more stable across changing market conditions.
The Kelly criterion and its fractions
The Kelly criterion computes the allocation that maximises long-run growth given a known edge and payoff. In practice the inputs are never known precisely, and full Kelly sizing produces severe drawdowns when estimates are even slightly optimistic. Practitioners who use it usually apply a fraction — half or a quarter of the Kelly amount.
Full Kelly assumes you know your edge. Nobody does.
Common mistakes
- Sizing from backtested win rates without discounting for overfitting.
- Ignoring correlation, so that several 'independent' positions are really one large bet.
- Increasing size after wins faster than the evidence justifies.
- Forgetting liquidity: a size that works on paper may move the market in practice.
A sensible starting framework
For most individual systematic traders, a conservative fixed-fractional or volatility-targeted approach, combined with a portfolio-level cap on total risk and correlation-aware grouping, is easier to reason about than optimisation-heavy methods. Platforms such as QuantConnect and libraries like Backtrader let you test sizing rules separately from signals — a step worth taking before any capital is committed.
Sizing at the portfolio level
Sizing individual trades is only half the problem. A portfolio of ten positions, each risking 1%, does not risk 10% in total unless the positions are perfectly correlated — and it may risk much more than expected if correlations rise during stress, as they often do. Practitioners therefore set a portfolio-level limit on total open risk and group positions by exposure: sector, asset class, or common driver such as interest rates or the US dollar.
A simple discipline is to estimate, before each new trade, what the portfolio would lose if every position in the same group moved against you simultaneously. If that figure exceeds your tolerance, the new position should be smaller or skipped.
Sizing and psychology
Sizing also shapes behaviour. Positions that are too large relative to a trader's comfort tend to be closed early or overridden, which breaks the systematic process the sizing was meant to support. A size that allows the strategy to run as designed through an ordinary losing streak is usually preferable to a theoretically optimal one that will not be followed in practice.
Revisiting sizes over time
Volatility, liquidity and correlations change. Review sizing parameters on a schedule — quarterly is common — rather than reacting to recent results, and document every change in the same experiment log used for strategy research.