The saved answer
A historical gain, with important limits
After the assumed costs, the rule returned 22.28%, compared with -21.72% for holding NZD/USD over the same study period.
- Net return
- 22.28%
- Maximum drawdown
- -8.50%
- Closed trades
- 41
- Trade win rate
- 53.66%
- Daily Sharpe
- 0.82
- Total costs (USD)
- 462.92
The saved statistical test cannot rule out luck. This result is an exploratory lead, not evidence of a reliable trading edge.
95% block-bootstrap Sharpe interval: -0.066 to 1.515. 200 resamples; block length 5.
Equity and drawdown
Hypothetical account equity in USD. Each chart includes NZD/USD daily reference rates on the right axis. Full saved span: 2021-11-01 to 2026-10-30.
Data: the saved FXMacroData study. Daily reference observations are not executable quotes. Use the chart controls to inspect dates, compare series or change the display window.
What was tested
Sell the US dollar when the monthly change in US core CPI rises from the previous recorded figure; buy it when it falls.
- Signal
- A new US release compared with the previous recorded figure, not a market consensus. Unchanged figures do not trigger a trade.
- Entry and holding period
- Next eligible reference observation after the signal; no same-date fill. Hold for 10 days unless the stop or a new signal closes the position.
- Account
- Unlevered fixed-unit reference positions in USD; cash and fees reconciled. Starting capital: USD 10000.
- Costs and stop
- 5.0 basis points per side; 1.0% stop. A stop closes the trade at the first daily reference price beyond it, gaps included; intraday prices are not used. After a stop the position stays flat until the rule gives a new signal.
- Risk calculation
- Sample standard deviation; calendar-day account returns including cash days; sqrt(365.25); zero risk-free rate. Dependence may invalidate naive annualization.
- Evidence
- Reference-price simulation. No untouched holdout. This is a frozen historical example, not a current trade recommendation.
- Reference-price simulation; no intraday fill or margin evidence.
- Financing and broker rollover are not modelled.
- Costs are declared per-side assumptions.
- Bootstrap uncertainty does not adjust for strategy selection.
- A release rule trades the days after a report at daily reference rates; the first minutes after a release are not tested.
Calendar-year results
| Year | Return | Coverage |
|---|---|---|
| 2021 (partial) | -1.73% | 2021-11-01 to 2021-12-31 |
| 2022 | 9.07% | 2022-01-01 to 2022-12-31 |
| 2023 | 5.99% | 2023-01-01 to 2023-12-31 |
| 2024 | 11.94% | 2024-01-01 to 2024-12-31 |
| 2025 | -6.00% | 2025-01-01 to 2025-12-31 |
| 2026 (partial) | 2.29% | 2026-01-01 to 2026-10-30 |
Costs and entry delay
| Scenario | Net return | Maximum drawdown |
|---|---|---|
| Selected costs | 22.28% | -8.50% |
| Double costs | 17.38% | -9.50% |
| One additional observation delay | 21.18% | -8.34% |
Study identity and metric definitions
Strategy: 35818cbfec066ba44e92027037b541abc488012b33be64c746f7069ab3897beb
Engine: reference-account-v2. Metrics: calendar-account-1.
Net return measures the account change after assumed costs. Drawdown measures the fall from an earlier account peak. Win rate counts profitable closed trades. Sharpe uses calendar-day account returns, including days in cash.