Constructing a volume-weighted moving average as a forecast baseline

How the average is assembled

A volume-weighted moving average is a moving average that weights each price observation by the volume observed on the same bar. It is constructed by summing each bar's price multiplied by that bar's volume over a chosen lookback, then dividing by the total volume in the same lookback.

The construction takes two explicit inputs: the price series being averaged and the number of bars in the averaging window. Volume is the weight applied to each price inside that window.

Why the volume weight differs from a simple moving average

The stated purpose of the volume weight is to make the average more responsive during periods of higher volume than a simple unweighted moving average on the same lookback.

Editorial note: hold that lookback fixed and treat the volume-weighted moving average as an explicit forecast baseline beside the simple moving average. The contrast is in how the average is assembled, not in a result.

A separate triangle pattern rank

Volume-price analysis can use a different mechanical procedure. A triangular-formation routine identifies a zigzag-defined window, fits a linear regression to closes in that window, then fits secondary regressions to positive and negative residuals around that line.

That procedure also applies a crossover count and a declining-volume test, then sums the scored factors into a triangle pattern rank. A triangle pattern rank is a summed score of several mechanical tests used to rank a candidate triangular formation.

In one supplied implementation, the volume-slope factor adds to the rank only when the linear-regression slope of volume over the zigzag window is negative. The same implementation scores the window length, residual standard error, and retracement size as additional additive rank components before any trade-window logic is applied.

Editorial note: that rank is not the volume-weighted moving average. Keep the pattern score apart from the moving-average baseline comparison on a shared lookback.