Module 13 of 14 · Running it for real

Small accounts and whole contracts

Module 9 said: trade as many markets as possible. Module 2 said: you need enough capital to hold several whole contracts in each. With a retail-sized account those two pieces of advice collide. Spread $250,000 across thirty markets and most ideal positions are a fraction of one contract, which rounds to zero — or to one, at several times the intended risk.

The obvious workaround is to pick a handful of markets you can afford and trade only those. That throws away most of the diversification you were trying to buy, and the choice of which handful becomes a big, unintended bet.

A better approach treats rounding as an optimisation problem. Each day:

  1. Compute the ideal, fractional position in every market as if capital were unlimited.
  2. Search for the whole-contract portfolio whose risk most closely matches the ideal one — minimising tracking error, the volatility of the difference, measured with a covariance matrix.
  3. A simple greedy search works well: start flat, repeatedly add the single contract (in the direction of its forecast) that most reduces tracking error, stop when nothing helps.
  4. Add a cost penalty and a tracking-error buffer so the optimiser only trades when the improvement is worth paying for.

🎛 Whole-contract optimiser

250000
20%
2

Do nothing

12.3%

tracking error

Round each position

5.1%

7 markets held

Greedy optimiser

3.6%

7 markets held

MarketForecastIdeal contractsRoundedOptimised
MES Micro S&P 500+81.5421
MNQ Micro Nasdaq+60.5911
ZN 10y note-9-1.34-1-1
ZF 5y note-7-1.62-2-2
MGC Micro gold+141.9423
SI Silver+110.1300
MCL Micro WTI-5-2.23-2-1
NG Natural gas+40.2200
ZC Corn-12-2.42-2-2
6E Euro FX+30.2600

Ten illustrative markets with fixed forecasts. Ideal weights come from the usual volatility-targeted sizing with equal instrument weights and a diversification multiplier. Tracking error is the annualised volatility of (held portfolio − ideal portfolio), using a simple correlation model (0.75 within an asset class, 0.15 across). The greedy optimiser starts flat and keeps making the single one-contract change that most reduces tracking error — adding in the direction of a market's forecast, or taking back a contract it already added — until no change helps.

Correlated markets stand in for each other

The optimiser exploits correlation. If you can't afford a full silver contract, a few micro gold contracts capture much of the same metals exposure. If the 5-year and 10-year notes both want to be short, one 10-year contract can carry both views. The result is a portfolio of fewer markets that behaves like the many-market ideal far more closely than naive rounding does.

Two cautions. The answer is only as good as your correlation estimates, so cap concentration in any single market with position limits (Module 14). And the method suits daily strategies; fast intraday strategies need their own whole-contract capital in every market they trade.

Things to try

  • • Drop capital to $50,000: rounding puts most markets at zero, while the optimiser still finds a few contracts that cover several views at once.
  • • Raise capital to $3m: rounding error becomes small and the two methods converge — optimisation matters most for small accounts.
  • • Watch silver (SI): at small capital it is usually zero in both columns, and micro gold carries the metals view instead.

Original interactive lesson. Labs run on a seeded synthetic market in your browser, built to illustrate mechanics, not to predict any real market. Contract figures are illustrative; check exchange specifications. Educational analysis, not investment advice.

How are we doing?