Normalized multi-horizon historical volatility construction

Construct relative quietness first

Editorial: a later trade idea is clearer when relative quietness is built as its own stack, then a directional overlay is attached. The finished object is a regime filter plus a baseline, not a single raw volatility number.

The historical workflow starts by converting closes to log returns. It then scales several short realized-volatility windows by one long baseline, takes their average, and exponentially smooths that average. A short least-squares line of the close and a price exponential average are added only after that quietness stack exists.

Share one log-return series

Every volatility window is computed from the same series: the natural log of the current close divided by the previous close. That log-return is the shared input to every standard-deviation window.

Form a volatility ratio against one long baseline

A single-horizon historical-volatility reading is the short-window standard deviation of those log close-to-close returns, divided by the same statistic over a longer window. The longer window defaults to 100 periods.

The short lookback in the numerator is a parameter. It defaults to 4 periods, so the same ratio definition can be recomputed at other short lengths against the unchanged 100-period denominator.

Editorial: that volatility-ratio puts different short lookbacks on a comparable scale. It is a relative-quietness reading, not a second price series.

Average three short windows and smooth the mean

A composite volatility reading is the arithmetic mean of three such ratios. The short lookbacks are 4, 6, and 10 periods, and all three use the same longer-window denominator.

That three-window average is then smoothed with a 12-period exponential moving average.

Editorial: the equal-weight mean of the three ratios and the 12-period exponential smoother are the moving-average layer of this construction. They refine the quietness stack. They do not replace the ratio definition.

Add the short line, the true range, and the price average

The same construction set also includes a 10-period least-squares or linear-regression line of the close and a one-day true range. That true-range is treated as equivalent to a one-day average true range.

A companion 20-period exponential average of the close defines a two-bar confirmation. After two consecutive lows sit above that average, a buy stop is placed 10 ticks above the two-bar high. The alert or long state is canceled if a low trades back through the average.

Editorial: the least-squares line is a short directional overlay plotted beside the volatility stack. The 20-period exponential average is a breakout baseline. Neither object is a substitute for the relative-quietness reading.

Read the finished stack as filter plus baseline

Editorial: once the ratios are scaled, averaged, and smoothed, the stack answers how quiet the recent windows are relative to the long baseline. The price average then locates a baseline for a later idea. The construction stays intact when that order is kept, rather than collapsing everything into one unsmoothed short-window standard deviation.