Home/Systematic Trading/Expectancy and Position Sizing

Expectancy and Position Sizing

model-buildingLevel 2 — Intermediate

Prerequisites

Edge IdentificationBacktesting Discipline

What It Is

Expectancy and position sizing is the quantitative framework for translating a trading edge into risk-adjusted returns — balancing win rate, payoff ratio, and frequency of opportunities to optimize for expected value while managing risk of ruin.

Correct Execution

  • Measure edge in terms of expectancy: E = (win rate × avg win) − (loss rate × avg loss); positive expectancy is necessary but not sufficient for profitability
  • Size positions to maximize long-run growth, not single-trade expectancy — Kelly criterion provides the theoretical optimum but practical trading uses fractional Kelly
  • Optimize frequency of occurrence alongside edge magnitude — a small edge applied thousands of times can outperform a large edge applied rarely
  • Apply rigid loss mitigation: high win rate with unlimited losses destroys the edge; high loss rate with large wins requires extraordinary payoff ratio to be viable
  • Concentrate on exploiting your strengths at maximum scale; do not dilute edge by applying it where it barely exists

Progression Levels

Diagnostic Tree

Coaching Cues

  • "Expectancy is a function of three things: win rate, win size, and loss size. Everyone focuses on win rate. That's the mistake." — David Sun, FWM S5E5
  • "You should exploit your strengths, not try to fix your weaknesses. If you have an edge, press it to maximum scale while it's there." — Euan Sinclair, 2020-09-25
  • "Rolling a loaded die thousands of times is how you actually express a probabilistic edge. The central limit theorem needs occurrences to work." — David Sun, FWM S5E5

Common Errors

  1. Focusing exclusively on win rate: Most traders want high win rate because losses feel bad; but expected value is what matters → Evaluate strategies on expectancy (E = WR × avg win − LR × avg loss), not on win rate alone
  2. Not engineering frequency: A strategy with 10x better frequency at the same expectancy earns 10x the annual edge → Actively seek ways to increase the number of independent occurrences for a given strategy concept
  3. Full Kelly sizing: Full Kelly maximizes long-run growth in theory but is too aggressive in practice — parameter estimation error means true edge is always smaller than estimated → Use half-Kelly or quarter-Kelly as the default; full Kelly should require overwhelming evidence of edge

Edges

🔑 Hidden Causal Lever

Frequency Is Equally Important as Edge Magnitude — and Far More Engineerable

model-buildingexpectancy-and-position-sizing →

Expected annual return is the product of three components: edge per trade × win rate × number of independent occurrences. Most systematic traders obsess over improving edge (better signal, better entry) while ignoring frequency. A strategy with edge 0.5% per trade and 5,000 annual occurrences outperforms one with edge 2% per trade and 50 annual occurrences — but the former requires intentional engineering while the latter feels more like a "real" trade. Frequency is also more engineerable than edge: you can increase occurrence count by applying the same signal to more instruments, shorter time horizons, or multiple simultaneous variants. You cannot easily double edge magnitude.

What most people do
Optimize signal quality and edge magnitude per trade; accept frequency as a byproduct of the strategy design rather than a variable to engineer.
What the best do
Set frequency as an explicit design requirement; engineer occurrence count by expanding universes, shortening time horizons where the edge survives, and running simultaneous signal variants; treat frequency as a first-class optimization dimension alongside edge and win rate.
Why it's an edge: The central limit theorem only materializes with sufficient occurrences — a strategy with 20 trades per year is mostly luck; a strategy with 2,000 trades per year is approaching statistical reliability. Most systematic strategies are built with 20 trades per year because that's how discretionary trading works.
How to exploit: For your best-performing systematic signal, measure its Sharpe at three frequency levels: current implementation, 3x current frequency (expand to more instruments), and 10x current frequency (shorten holding period to next valid signal window). Compare Sharpe improvement per unit of frequency increase. If the signal degrades slowly with frequency (it often does until you hit microstructure friction), the engineering investment in frequency is the highest-ROI improvement.
Cross-domain parallel
In sports betting, a model with +3% EV per bet and 100 bets/year is less valuable than one with +2% EV per bet and 1,000 bets/year. Sharp bettors deliberately expand their betting markets to maximize occurrence count.
David Sun, "Expectancy Hacking," FWM S5E5, 2022-06-27
⚡ Conventional Wisdom Is Wrong

Once Edge Is Established, Undersizing Is Wrong Risk Management — It's Leaving Money on the Table

model-buildingexpectancy-and-position-sizing →

The intuition "be conservative while developing confidence in the strategy" is correct before edge is proven. After edge is clearly established with sufficient live trading history, continuing to undersize is not conservative — it is the wrong risk management choice. The Kelly criterion provides mathematical clarity: the position size that maximizes long-run wealth growth is explicitly positive, and sizing below the optimal fraction reduces the long-run compounding rate. An investor with a clear edge who sizes at 10% of Kelly is generating 90% less expected return per unit of edge. The risk is in the edge decaying while undersizing persists.

What most people do
Size conservatively throughout the strategy lifecycle, never increasing past "comfortable" levels regardless of established edge; treat undersizing as always safer than full sizing.
What the best do
Establish explicit thresholds for increasing sizing: after N trades with documented positive expectancy, increase to half-Kelly; after 2x that, increase toward full-Kelly; treat each scaling milestone as a risk management decision requiring the same rigor as the original strategy evaluation.
Why it's an edge: Edges don't persist indefinitely — market conditions change, competition grows, information gets priced in. An investor who identifies a genuine edge and extracts maximum value during its active period outperforms one who waits until the edge decays before committing full capital.
How to exploit: Define a live trading milestone in advance: "After 500 independent occurrences with documented edge above 0.5% per trade, I will increase sizing to half-Kelly." Use the Kelly formula with your measured (not estimated) edge and win rate. Treat the milestone as a commitment device, not an option to reconsider.
Cross-domain parallel
In algorithmic trading, a strategy with proven edge should be scaled to its capacity constraint (market impact) as fast as the trading infrastructure allows — waiting allows others to find and trade the same edge, compressing the premium.
Euan Sinclair, "Edge is in the Numbers," 2020-09-25 — "if you've got a real edge, you should be trading it as big as you possibly can."
⚡ Conventional Wisdom Is Wrong

High Win Rate With Unlimited Losses Is the Most Common Way to Lose Money With a Smile

model-buildingexpectancy-and-position-sizing →

Traders instinctively optimize for win rate because wins feel good and losses feel bad — prospect theory in action. A strategy that wins 80% of the time feels excellent even when the 20% of losses are large enough to produce negative expected value. The expectancy formula makes this precise: a strategy with 80% win rate and 1.5x average win but 8x average loss has negative expectancy despite the high win rate. Options sellers, trend-fighters, and mean-reversion traders who don't use stops regularly build strategies with this profile without realizing it. The strategy looks great for months or years until the tail event arrives.

What most people do
Judge strategies primarily on win rate; add positions that improve win rate; remove stops that reduce win rate.
What the best do
Evaluate ALL strategies on expectancy first; only examine win rate in the context of its relationship to loss size; hard-code stop rules that prevent losses from growing large enough to make win rate irrelevant.
Why it's an edge: The psychological pressure to optimize win rate is ubiquitous; investors who systematically override it and optimize expectancy build portfolios that survive tail events instead of blowing up at them.
How to exploit: For every systematic strategy you run, compute the full expectancy distribution: (win rate × average win) - (loss rate × average loss). If this number is positive by less than 50% of the average loss size, the strategy's positive expectancy is fragile to tail losses. Add a hard stop that limits maximum loss to 2x the average loss — accept the win rate reduction this causes and verify that positive expectancy survives.
Cross-domain parallel
In sports betting, fade-the-public strategies have high apparent win rates in backtests when tested on favorable scenarios, but the full expectancy including large losses on bad spots turns negative. The win rate metric alone misleads.
David Sun, "Expectancy Hacking," FWM S5E5, 2022-06-27; Kris Abdelmessih, "Risk Management and Edge," 2022-05-07

Sources

  • David Sun, "Expectancy Hacking," Flirting with Models S5E5, 2022-06-27 — expectancy framework; zero-DTE as frequency optimization; law of large numbers in strategy design
  • Euan Sinclair, "Edge is in the Numbers," 2020-09-25 — exploit strengths not weaknesses; press edge at scale while it lasts
  • Kris Abdelmessih, "Risk Management and Edge," 2022-05-07 — risk management as precondition for expressing edge; risk of ruin in sizing