A trading strategy is a specified policy for taking and changing market exposure: what can be traded, what information may be used, when to enter, how much to hold, when to exit, and how risk, costs and execution are handled. Its practical purpose is to turn an investment idea into decisions whose financial consequences can be tested and monitored.
“Buy when the forecast is positive” is a signal or an incomplete trading instruction. It leaves open whether to buy a share or an entire portfolio, what price is feasible, how long to hold, and what to do when the trade cannot be completed. Those choices can determine whether the same forecast produces a profit or a loss.
What the specification has to settle
A strategy need not contain complicated models. It does need decisions that are sufficiently clear for its intended use.
| Part of the policy | Questions to answer before evaluating it |
|---|---|
| Universe and information | Which instruments are eligible at each decision time? Which data are actually available then, including revisions and corporate actions? |
| Entry | What observable condition authorises a new position, and when can an order first be sent? |
| Size and portfolio | How does the signal become shares or weights? What happens when positions compete for cash or exposure limits? |
| Exit and maintenance | What price, time, signal reversal or risk event changes or closes the position? When are targets reviewed? |
| Risk | What limits apply to concentration, leverage and loss exposure? When is new trading suspended, and who can override it? |
| Execution and exceptions | Which orders are allowed? How are partial fills, rejections, outstanding cancellations and unavailable liquidity handled? |
| Economics | Which fees, spread, impact, funding and borrow assumptions enter the evaluation? Which reference prices are used? |
A systematic trading strategy makes these decisions through explicit rules. A discretionary strategy can reserve judgement to a trader, but that discretion still affects exposure and needs a defined mandate. Algorithmic trading concerns the automation of trading decisions or orders; automation alone does not establish that a strategy has an economic edge.
For research, the specification also needs a reason the idea might work and evidence that could disprove it. Economic Mechanisms and Falsifiable Hypotheses develops that research contract, including information timing, a feasible baseline and rejection criteria. The full lesson requires membership; the linked page provides a public preview.
A profitable trade is an outcome, not a specification
Consider an original fictional round trip in a stock quoted in USD per share. A strategy requests a long position of 10 shares and later closes it. Stipulate complete purchase fills at USD 100 and complete sale fills at USD 105, with explicit fees of USD 2 on each side. There are no dividends, financing costs, taxes or currency conversions in this illustration.
| Event | Change in shares | Trading cash flow | Explicit fee | Net cash flow |
|---|---|---|---|---|
| Purchase | +10 | −USD 1,000 | −USD 2 | −USD 1,002 |
| Sale | −10 | +USD 1,050 | −USD 2 | +USD 1,048 |
| Completed round trip | 0 | +USD 50 | −USD 4 | +USD 46 |
The reconciliation is
Original schematic and fictional USD ledger. Both fill prices and full execution are stipulated. The USD 46 is the result of this invented round trip, not historical evidence or a forecast.
The fill prices already determine the trading cash flows. Relative to a chosen reference price, they could include spread and slippage; charging the same price difference again as an additional cost would double-count it. Other real-world costs would need to be added when applicable.
The ledger does not tell us why the position was entered, whether the information was available before the order, or whether the sizing and exit rules were sensible. Nor does it estimate the strategy's trade expectancy. A single stipulated winner supplies no distribution of future outcomes. Without a declared capital base and timing, USD 46 is a cash profit, not a portfolio return or an annualised performance measure.
Positions, Trades, Fills and P&L develops the holdings-and-cash accounting, including partial fills and the difference between a desired position and actual exposure. Its full lesson also requires membership, with a public preview.
Two kinds of uncertainty remain
Financial uncertainty remains even when every rule is followed. Forecasts can fail; relationships can change; liquidity and trading costs can deteriorate. A backtest estimates how a specified policy behaves under historical data and modelling assumptions. Testing realistic information timing, feasible execution, costs and sensitivity to assumptions matters more than displaying an attractive gross curve. Changing the rules after seeing the test result makes that result part of the research process for the revised policy.
Operational uncertainty concerns what the system has actually done. An order acknowledgement can be missing, a cancellation unresolved, or a price feed stale. The strategy needs an exception policy: for example, suspend new submissions when order state cannot be reconciled, verify outstanding orders and fills, and resume only through a defined recovery process. Suspension alone neither cancels every working order nor closes an existing position. A stop-loss rule likewise cannot guarantee its trigger price as a fill price.
For U.S. FINRA member firms using algorithmic strategies, Regulatory Notice 15-09 discusses testing, deployment, monitoring and control practices. Its regulatory context is specific; the underlying distinction is broadly useful: a trading idea, an implemented policy and a reconciled financial result are separate things to examine.
Sources
- Rishi K. Narang, Inside the Black Box: A Simple Guide to Quantitative and High-Frequency Trading, 2nd edition (Wiley, 2013), Chapter 2, pp. 17–20 — the roles of alpha, risk, transaction costs, portfolio construction, execution, data and research. The schematic, decision table and numerical ledger here are original teaching constructions.
- FINRA: Regulatory Notice 15-09, Guidance on Effective Supervision and Control Practices for Firms Engaging in Algorithmic Trading Strategies, 26 March 2015 — development, testing, deployment and monitoring practices for FINRA member firms; checked 26 October 2026.