glancing at multiple commodity charts, one can see phases of linearity. The linearity is time variable. Meaning, for a certain amount of time over a historical record, there will be periods of lines with low variance around the regression.
Is there a statistical measure for segmentation of a number series, and calculating the average duration of linearity?
Once the average duration of linearity is obtained. Then price series can be screened for duration thresholds.
If duration thresholds are being exceeded, in a sense its breaking out from past behavior.