Better Bollinger Bands (now open source)
General purpose Bollinger band indicator with a number of configuration options and some additional color-coded information. The main advantages of it over standard Bollinger bands are:
1) Better statistics:
* Uses volume weighted moving averages, variance, and standard deviation by default. The volume dependence can be disabled with a checkbox option, but generally makes it more responsive improves its ability to distinguish true outlier events from random variation.
* Lets you pick between different time windows (simple, sawtooth (WMA), exponential) in addition to the volume weighting, with appropriate Bessel corrections to make the estimators unbiased and to get consistent result for different weights.
* Has a checkbox option to use a linear regression in the band calculation if you don't want average momentum to be counted in the volatility. This turns the centerline into a last squares moving average, and the band width at each time step is given by the variance away from the regression line instead of from a moving average. Weights in the least squares regression are changed according to the other options. For tickers with a strong long-term trend this makes the bands track the price action more closely.
2) Geometric
* This does all calculations on log(price) instead of the prices themselves.
* Makes almost no difference in most cases, but gives better results on charts with strongly exponential behaviour that range between several orders of magnitude.
* Properly centered around price action on log plots.
* Will never annoy you by rescaling a log plot due to a negative lower band. The lower band is always positive for positive prices.
3) Some built in oscillators.
* This aims to reduce clutter by building in some other indicators into the band color scheme. You can pick between various momentum & RSI operators to color the center line and the bands, or leave the bands plain.
I've been using these bands myself for a few months & have been gradually adding functionality & polish. Feel free to comment, or to refer to me if you borrow any ideas.
1) Better statistics:
* Uses volume weighted moving averages, variance, and standard deviation by default. The volume dependence can be disabled with a checkbox option, but generally makes it more responsive improves its ability to distinguish true outlier events from random variation.
* Lets you pick between different time windows (simple, sawtooth (WMA), exponential) in addition to the volume weighting, with appropriate Bessel corrections to make the estimators unbiased and to get consistent result for different weights.
* Has a checkbox option to use a linear regression in the band calculation if you don't want average momentum to be counted in the volatility. This turns the centerline into a last squares moving average, and the band width at each time step is given by the variance away from the regression line instead of from a moving average. Weights in the least squares regression are changed according to the other options. For tickers with a strong long-term trend this makes the bands track the price action more closely.
2) Geometric
* This does all calculations on log(price) instead of the prices themselves.
* Makes almost no difference in most cases, but gives better results on charts with strongly exponential behaviour that range between several orders of magnitude.
* Properly centered around price action on log plots.
* Will never annoy you by rescaling a log plot due to a negative lower band. The lower band is always positive for positive prices.
3) Some built in oscillators.
* This aims to reduce clutter by building in some other indicators into the band color scheme. You can pick between various momentum & RSI operators to color the center line and the bands, or leave the bands plain.
I've been using these bands myself for a few months & have been gradually adding functionality & polish. Feel free to comment, or to refer to me if you borrow any ideas.