Outliers Detector with N-Sigma Confidence Intervals (TG fork)

30 nov. 2022

Display outliers in either value change, volume or volume change that significantly deviate from the past.

This uses the standard deviation calculation and the n-sigmas statistical rule of significance, with 2-sigma (a value of 2) signifying that the observed value is stronger than 95% of past values, and 3-sigma 98.5% of past values, and so on for higher sigma values.

Outliers in price action or in volume can indicate a strong support for the move, and hence potentially more moves in the same direction in the future. Inversely, an insignificant move is less likely to be supported. And of course the stronger, the more support.

This indicator also doubles as a standard volume indicator if volume is selected as the source, but with the option of highlighting outliers.

Bars below significance can be uncolored (gray) to unclutter the visuals.

Differently to almost all other similar indicators, the background highlighting is dynamical, so that all values will be highlighted differently, not just 2-sigma or 3-sigma, but also 4-sigma, 5-sigma, etc, with a different value of transparency.
The dynamical transparency value can be calculated in two ways: either statically proportionally to the n-sigma but capped at 10-sigma, or either as a ratio relative to the highest observed sigma value over the defined lookback period (default: 300).

If you like this indicator, which is an extension of previously published indicators, please give some love to the original authors:

* tvjvzl :

* vnhilton :

This extension, authored by Tartigradia, extends tvjvzl's indi, implements vnhilton's idea of highlighting the background, and go further by adding dynamical background highlighting for any value of sigma, add support for volume and volume change (VolumeDiff) as inputs, add option to uncolor insignificant bars, allow plotting in both directions and more.