Throughput Stats

What are throughput stats?

Throughput stats describe how a strategy uses time and capital rather than how profitable it is: trade frequency, exposure or time-in-market, turnover, average holding time, and capacity. Two systems with identical expectancy can be radically different businesses if one produces five trades a year holding for months while the other produces five a day holding for minutes. Throughput is the dimension along which those differences live.

Traders care because throughput multiplies everything else. Earning power is roughly expectancy times frequency, so a modest edge taken often can out-earn a large edge taken rarely, provided costs cooperate. Exposure, the fraction of time capital is actually in the market, determines both how much risk the account carries on an average day and how honestly returns compare to a buy-and-hold benchmark that is exposed all the time. Turnover, the notional traded relative to equity, is the direct driver of commission drag and the natural companion to cost sensitivity analysis.

Capacity is the forward-looking member of the family: the amount of capital a strategy can deploy before its own orders move prices enough to erode the edge. Short holding times, thin instruments, and reliance on precise fills all shrink capacity. A backtest is silent about this by default, which is why throughput stats belong in every validation report rather than only in the performance summary.

How it's calculated

There is no single canonical formula; throughput is a family of standard measurements computed from the trade list and equity history.

Conventions vary: turnover is quoted per year in many contexts, exposure is sometimes computed per instrument in portfolios, and overlapping positions complicate time-in-market accounting.

Capacity has no closed-form expression; it is estimated from liquidity, participation limits, and impact assumptions.

How traders use it

• To convert per-trade edge into expected earning power, since annual return potential scales with frequency; a 0.2R expectancy at 200 trades a year is a different proposition from the same edge at 20.
• To judge whether a backtest sample is even scoreable: low-throughput systems may need a decade of data to accumulate enough trades for significance of returns testing.
• To normalize return comparisons by exposure, because a strategy in the market 15% of the time that matches an always-invested benchmark is doing far more per unit of risk-time than the raw numbers suggest.
• To anticipate live frictions: high turnover forecasts heavy commission drag and slippage load, and short holding times raise the bar on cost model realism.
• To estimate capacity before scaling, checking average traded size against typical volume and book depth in the traded instrument, since an edge that works at small size may not survive its own footprint.

Throughput stats vs related concepts