A complete window can still require a drawdown review
Compare two synthetic twenty-month return paths using compounding, paired active returns, tracking error and drawdown from the initial wealth point.
The reference fixture compounds to 11.920900% for the model and 9.765627% for the benchmark. Changing four model returns produces a −15.468102% model return and −20.427714% maximum drawdown. Both windows remain complete; the second report requests drawdown review rather than withholding measurable outcomes.
Establish the aligned window and return units
Begin with the reference preset, which has one model and one benchmark return for each of twenty monthly period dates. Each number is a decimal return: 0.012 means 1.2 percent. The dates run from January 31, 2023 through August 31, 2024 and are declared period keys. The fixture supplies no transaction costs, investor cash flows or intramonth path. Its calculations cannot establish actual execution or returns net of unspecified charges.
Starting from wealth one, multiply by one plus each return in chronological order. The first model periods illustrate the convention: 1.012 multiplied by 0.992 equals 1.003904 after February. Continue through all twenty pairs. The resulting model wealth is 1.1192089974 and benchmark wealth is 1.0976562697. Their returns are 11.920900% and 9.765627%; subtracting them gives a 2.155273 percentage-point difference over the declared window.
Keep active return, active risk and compounding distinct
For each period, arithmetic active return is the model return minus the paired benchmark return. Its mean is 0.001, or 0.10 percent per month in this synthetic sample. That mean is not the compounded performance difference. Tracking error takes the sample standard deviation of those twenty active returns and multiplies by the square root of twelve, the declared monthly annualization factor. It is approximately 0.940325% annualized in the reference fixture.
CFA Institute uses tracking error as a measure of active risk, while this implementation declares the specific sample-deviation and annualization conventions. Annualizing twenty observations does not create an independently validated long-run risk estimate. Neither the cumulative-return difference nor the arithmetic active mean is regression alpha. The report also avoids inventing a Sharpe ratio when no risk-free series, frequency convention or defensible inference framework has been supplied.
| Metric | Reference | Drawdown preset |
|---|---|---|
| Model cumulative return | 11.920900% | −15.468102% |
| Benchmark cumulative return | 9.765627% | 9.765627% |
| Difference of cumulative returns | 2.155273 percentage points | −25.233729 percentage points |
| Mean active return per month | 0.100000% | −1.265000% |
| Annualized tracking error | 0.940325% | 9.803292% |
| Model maximum drawdown | −1.200000% | −20.427714% |
Trace the loss from its running peak
Now select the drawdown preset. Four model returns change: May, June and July 2023 become minus eight, minus seven and minus four percent, and March 2024 becomes minus five percent. Benchmark returns and feature values remain unchanged. The model's running peak is 1.0291160451 in April 2023. By March 2024 wealth is 0.8188911677, giving wealth divided by running peak minus one of approximately minus 20.427714 percent.
The final model return is minus 15.468102 percent, which differs from the worst peak-to-trough loss. Later recovery does not erase the earlier drawdown. The calculation includes initial wealth and initial peak of one, so a loss in the first period would also count. The synthetic benchmark's maximum drawdown remains minus 0.9 percent. Monthly observations cannot reveal deeper drawdowns occurring between the supplied period endpoints.
Review a measured loss without hiding it
The drawdown preset has complete aligned returns, so all outcome metrics remain available. Its maximum drawdown crosses the selected ten-percent review boundary and the report status becomes review. The feature distribution is unchanged, with PSI zero; feature stability therefore cannot establish acceptable outcomes. Conversely, the presence of a loss does not make the return arithmetic invalid. Separating calculation availability from the review decision preserves the evidence a reader needs.
Export both reports with their input versions and identify the four changed observations. A hypothetical reviewer would ask whether the benchmark is suitable, the return construction is consistent and the policy reflects the intended use. Removing a difficult month or replacing a null with zero would invalidate the complete-window comparison. When a required pair is genuinely missing, this report withholds the performance block and retains the explicit reason instead.
Keep in the decision record
- Require a complete paired window before reporting comparable outcomes.
- Compounded difference, mean active return and tracking error are distinct.
- A measurable drawdown can trigger review without invalidating the calculation.
Source notes
- CFA Institute: active return and tracking error
- CFA Institute: active equity portfolio construction and drawdown context
- NIST: sample standard deviation convention
Sources inform the stated methods and governance context. All example monitoring observations are synthetic. No regulatory approval or production monitoring certification is implied.
Continue the analysis
Require an aligned return window before reviewing outcomes
Match model and benchmark periods, units and coverage before calculating a complete-window result.
Separate compound returns from arithmetic active returns
Reproduce wealth linking and avoid calling a cumulative return difference alpha.
Read drawdown from the full wealth path
Include the initial wealth peak and distinguish recovery arithmetic from a decision to trade.
Interpret tracking error with its period and dependence limits
Calculate sample active-return volatility and understand the square-root annualization convention.
Version 1 · Published 16 October 2026. Inspect the method · Open the monitoring report