Markout profit measures the economic result associated with a trade by comparing its traded price with the market price after a fixed markout interval, such as one minute. A positive result indicates that the subsequent market move favored the trade; trading cost can be subtracted to make the metric a business measure of strategy performance. section 4.1.1
For a signal-driven proprietary strategy, markout profit links a decision rule to an optimizable outcome. The strategy buys above a positive signal threshold, sells below its negative counterpart, and otherwise does not trade. In a simulator, expected profit rises with signal magnitude, while trading cost and Gaussian market noise reduce or vary the realized result. section 4.1.1
Figure 4.2 A proprietary trading strategy. The strategy computes a signal from public market data. If signal > threshold, the strategy sends a BUY order to the exchange. Similarly, if signal < -threshold, the strategy sends a SELL order; threshold is a parameter.
The trading flow makes the metric actionable: a threshold determines when the signal produces a BUY or SELL order, creating the trades whose markouts are measured. section 4.1.1
Single-trade markout
An illustrative buy of 100 shares at $5.00 compares that execution with a $6.50 market price one minute later. The stated example records $50 of markout profit, then $49 after a $0.01-per-share trading cost. The calculation is applied per trade; a simulation returns zero when the signal does not cross the threshold and no trade occurs. section 4.1.1 section 4.1.1
Individual outcomes are noisy, so compare parameter settings using aggregate mean markout profit and standard errors. Randomly assign threshold values at trading decision times and collect many individual measurements per setting; this reduces uncertainty in each aggregate measurement while limiting the number of settings experimentally tested. section 4.1.2
Response surface methodology (RSM) can estimate markout profit at unmeasured settings. For one threshold, fit a quadratic surrogate with linear regression to three aggregate measurements, interpolate across the measured range, and select the threshold with the largest estimated profit. The selected value is an educated guess rather than a direct production measurement, so it requires experimental validation. section 4.1.3 section 4.1.4
Use markout profit in an optimization loop:
- Choose feasible threshold settings; negative thresholds create overlapping BUY and SELL regions. section 4.1.1
- Measure many individual markouts at each setting and retain aggregate means with standard errors. section 4.1.2
- Fit and optimize a surrogate only within the measured range; validate the resulting parameter experimentally. section 4.1.3 section 4.1.4
- For threshold and order size together, use a two-dimensional model with linear, quadratic, and cross terms; redesign if an outlying measurement violates the near-optimum modeling assumption. section 4.2.2
In two dimensions, the metric remains the response variable while threshold and order size become the controllable inputs. Nine measured parameter tuples can support a two-dimensional surrogate that yields many estimated profit values, but visual ordering of the tuples alone cannot reveal an unmeasured optimum. A final direct measurement accepts the estimate only when it falls within the measurement’s 95% confidence bounds. section 4.2.2
