| name | trend-projection |
| description | Copy-pasteable scikit-learn and numpy code patterns for fitting a linear trend on recent WTI price history, projecting point forecasts to standard horizons, and calibrating 80% prediction interval widths from residual standard errors. Load references/projection-examples.md before writing any trend-projection code. |
Trend projection skill
Run the statistical-analysis skill first to determine the current vol
regime and appropriate trend window before applying these patterns.
Load references/projection-examples.md via
load_skill_resource("trend-projection", "references/projection-examples.md")
before writing any trend-projection code.
The reference file contains:
- A complete working code pattern using
sklearn.linear_model.LinearRegression
to fit the most recent 30 trading days of WTI close prices.
- The standard interval-width formula:
1.28 * residual_std * sqrt(h / 5),
which produces the 80% CI half-width at horizon h business days.
- A guard for the edge case where the trend line overshoots the 52-week range.
- Worked numeric examples showing expected output for typical WTI vol regimes.
Quick-reference steps
- Parse the CSV history from the task payload into a DataFrame.
- Select the most recent 30 rows (trading days).
- Fit
LinearRegression on [0..29] (x) vs close price (y).
- Project to horizons 5, 10, 21 by evaluating the regression at
30 + h - 1.
- Compute
residual_std = std of (y - y_hat) on the 30-day window.
- Set 80% CI half-width =
1.28 * residual_std * sqrt(h / 5).
- Clip projected point forecast to
[0.5 * 52w_low, 1.5 * 52w_high] as a
plausibility guard — extreme trend extrapolation is usually wrong.
No scripts in this skill. Do not call run_skill_script.
