Position Sizing: Fixed, Volatility-Targeted, and Kelly

The same edge can produce wildly different equity curves depending on how you size positions. Position sizing is the second-most important decision in any strategy after the signal itself — and it is the decision most often made by accident.

Three Common Approaches

Fixed Notional (Equal Weight)

Every trade gets the same dollar amount — say, $10,000. Simple to reason about, easy to execute. Treats all opportunities equally regardless of how risky they are individually.

When to use: simple equity strategies on liquid names with similar volatility profiles. Fine as a baseline; rarely optimal.

Volatility-Targeted

Size each position so that its expected daily volatility (in dollar terms) is the same. A 10%-vol stock gets twice the dollar size of a 20%-vol stock. The portfolio has approximately constant total volatility regardless of which names are in it.

When to use: multi-asset strategies, or any strategy where the universe spans wide volatility ranges. The default for institutional systematic trading and a strong default for retail.

Kelly Criterion (and its variants)

Size each trade in proportion to the strategy's edge and the inverse of variance. Mathematically optimal for log-wealth growth — but only if you know your edge precisely, which you don't.

When to use: as an upper bound on aggressive sizing, with substantial discounting (Half-Kelly or Quarter-Kelly) to survive parameter mis-estimation.

The Edge-Sizing Symmetry

A strategy with a true Sharpe of 1.0 sized at Half-Kelly will have a 50% chance of drawing down 25% before reaching new highs. The same strategy at Full-Kelly will have a 50% chance of drawing down 50%. Same edge, twice the pain, only marginally more return. Half-Kelly is usually right.

The Common Mistakes

  • Using fixed notional with mixed-volatility universes. Treating a $10K position in TSLA the same as $10K in JNJ means TSLA dominates portfolio risk by an order of magnitude.
  • Setting position size by % of equity without considering volatility. "10% per trade" sounds disciplined but produces wildly different risk profiles depending on what you are trading.
  • Increasing size after wins, decreasing after losses. Path-dependent sizing creates path-dependent results that are not what you backtested.
  • Trading Full Kelly. Kelly is optimal only if you know the true edge. You don't. Full Kelly with mis-estimated edge can be ruin-optimal instead.

QuanterLab's Sizing Modes

Most builder modules expose at least three sizing options in the Risk Management panel:

  • Full Equity: deploy all available capital per trade. Simple, aggressive, ignores volatility.
  • Fixed: a constant dollar size per trade.
  • Volatility-Targeted: size to a target portfolio volatility, normalized by each name's recent realized volatility.
  • Kelly% (advanced): exposed in some modules; usually output as a ceiling rather than a directly-tradable size.

Run the same strategy across all sizing modes and compare equity curves. Same edge, dramatically different ride. Pick the one whose worst day you can actually live through.

The Bottom Line

Position sizing is not a free parameter to optimize — it is a risk preference. Pick a sizing scheme that matches what you can psychologically and financially survive, then leave it alone. Optimizing position sizing on the same data you optimized signals on is a quiet way to overfit twice.

Further Reading

Foundational papers

  • Kelly, J. L. (1956). A New Interpretation of Information Rate. Bell System Technical Journal, 35(4), 917–926.
  • Thorp, E. O. (2006). The Kelly Criterion in Blackjack, Sports Betting, and the Stock Market. In: Handbook of Asset and Liability Management, Vol. 1.

Textbook references

  • MacLean, L. C., Thorp, E. O. & Ziemba, W. T. (2010). The Kelly Capital Growth Investment Criterion. World Scientific.
  • Chan, E. P. (2013). Algorithmic Trading: Winning Strategies and Their Rationale. Wiley.

Related QuanterLab articles

Try it in QuanterLab

In any builder's Risk Management section, switch position sizing modes (Fixed, Volatility-Targeted, Kelly) and observe how max drawdown and Sharpe change. Same edge, different sizing — different ride.

Back to QuanterLab
Report
Loading report...
Article
Loading article...