Position Sizing for the CMT Exam: Match Units to Risk

Position size is a portfolio decision

An entry signal says when an idea is attractive; position size says how much loss the portfolio can absorb if the idea fails. The 2026 CMT Program Guide addresses risk principles at Levels I and II and explicitly includes position sizing in Level III systematic countertrend and trade-design work. This guide is a practical extension of those objectives. The official curriculum, not a single sizing formula, controls exam scope.

Three common approaches answer different questions:

A robust order uses the minimum size permitted by all relevant constraints. A volatility estimate can inform the stop distance, but it should not replace an invalidation drawn from the actual setup. See risk management for VaR and drawdowns.

Worked order ticket

A portfolio is 250,000. The trader permits a 0.50% planned loss on this idea, or 1,250. The proposed entry is 80, chart invalidation is 75, and an allowance for costs and slippage is 0.25 per share. Per-share planned risk is 80 - 75 + 0.25 = 5.25. The risk rule permits floor(1,250/5.25) = 238 shares.

The mandate caps one name at 7% of equity, or 17,500. At an 80 entry, that allows floor(17,500/80) = 218 shares. Average daily trading volume is adequate, but a separate participation rule limits the order to 200 shares during the intended execution window. The ticket is therefore 200 shares, 16,000 gross exposure and 200 × 5.25 = 1,050 planned loss. The most restrictive constraint is execution participation, not the 1,250 loss budget.

If the market opens at 71 after news, the loss from an 80 entry to a 71 fill is 9 × 200 = 1,800 before costs. Stop-based sizing did not cap the actual loss. A committee should stress several correlated positions against the same news or macro event. If that combined stress breaches the portfolio limit, reduce the order further.

Volatility and equal-risk ideas

Suppose two assets each receive a 10,000 nominal allocation, but Asset A has a 2% typical daily movement and B has 5%. Their approximate daily dollar movement scales are 200 and 500, respectively. Equal dollars are not equal risk. A simple inverse-volatility rule would assign weights in inverse proportion to those estimates; before normalization, A:B weights are 1/0.02 : 1/0.05 = 50:20, or 5:2. This is a starting point, not a guarantee of equal realized portfolio risk, because correlations, tails, and estimation errors matter.

Volatility estimates can jump after a loss. If the strategy mechanically cuts exposure at every spike and restores it after quiet markets, turnover and poor timing may result. Define rebalance frequency, lookback, maximum change per review, and portfolio concentration limits before backtesting.

Decision and invalidation

For the worked ticket, place the order only while price and volume conditions support the entry, the 7% cap holds, and participation remains feasible. If 75 is breached under the chosen close or intraday convention, exit or reduce according to the prewritten plan. If volatility or equity changes materially, recalculate allowed size for new orders. Do not widen the invalidation level merely to keep the same share count.

Size is not a confidence score. A persuasive story cannot override a binding limit. When several positions share an underlying factor, a per-name risk budget may understate aggregate exposure. A portfolio construction check belongs on every substantial order.

Recalculate after the portfolio changes

A fixed fraction of equity changes with account value. If a portfolio falls from 250,000 to 225,000, the same 0.5% planned-loss rule permits 1,125, down from 1,250. If the entry and per-share risk stay at 80 and 5.25, the risk-only maximum becomes floor(1,125/5.25) = 214 shares. The 7% name cap becomes 15,750, or floor(15,750/80) = 196 shares. The participation cap of 200 shares would no longer bind; the name cap would limit the order to 196.

This shows why sizing should be recomputed at the decision time. Carrying forward a share count from a prior portfolio value can quietly increase the fraction of equity at risk. It also shows that the binding constraint can change even when the chart does not.

A losing streak does not require an improvised "win it back" size increase. If a strategy uses a drawdown-based reduction, define the rule before the drawdown. If a strategy uses volatility targeting, use a stated estimation window and maximum adjustment. In both cases, maintain enough liquidity to exit and consider whether several positions depend on the same market factor. The final share count is the minimum permitted by the whole set of constraints.

The order ticket should also state whether the loss budget is measured against current equity or the value at the start of a review period. Those choices produce different sizes after a drawdown. A fixed-dollar budget can become a larger percentage of a shrinking portfolio, while a fixed fraction automatically scales down. Neither choice eliminates gap risk, and both need a rule for simultaneous positions that share one shock.

Common traps and self-check

Avoid rounding up, omitting fees and slippage, treating a stop as a guaranteed exit, and applying inverse-volatility weights without a concentration cap. An ATR multiple chosen after observing a losing trade is not a stable rule.

Self-check: A 100,000 portfolio permits a 0.6% planned loss. Entry is 40, invalidation 37, costs 0.20 per share, and the name cap is 5%. What is the maximum share count? Answer: Risk budget 600; risk-based size floor(600/3.20) = 187. The name cap is 5,000/40 = 125 shares, so 125 is the binding maximum before liquidity checks.

Next, complete a one-page ticket with all three limits, a gap stress, and the trigger that would void the order. Compare it with the current official objectives.