Dynamic Regression Bandings (Base10)
Dynamic Regression Bandings (Base10) is designed to provide a statistical range of outlier pricing within an established trend. Instead of calculations being performed on a linear scale, spot price is adjusted logarithmically, allowing for regression to be performed over longer periods without compound movement creating abnormal behaviour.
The range is set through user input of a minimum and maximum values; from which the script identifies the backward length (candle count) with the greatest correlation to price. This process is performed for each candle, so the regression length may change dynamically across time. By doing this, we are able to look at the current candle for its probability of being an outlier compared to the mean of the regression. If the spot price is outside the range of the expected deviation (e.g. +/- 2 standard deviations from the mean); a buy or sell signal is triggered.
IMPORTANT: This does not aim to validate the volatility of a trend, so the user must identify the historical fit. It is recommended to use the replay functionality to make these adjustments with historical data in order to avoid over fitting the model to the data; which will create long term issues with performance.
When a trend is found in the specified range; it is assumed that the white noise (movement +/- to the trend) happens in a normal & unbiased way. In a fair market; the buyers and sells should balance themselves out in such a way that there is no inherent bias outside of the trend. As such, we can assume that almost all movement within the trend will be within +/- 3 standard deviations. So if the selected deviation range is greater than that; it is likely that the model is being over fit to account for extreme volatility.
Below are examples of the indicator on different charts:
USDAUD
BTCUSD
AMZN
A2M
The range is set through user input of a minimum and maximum values; from which the script identifies the backward length (candle count) with the greatest correlation to price. This process is performed for each candle, so the regression length may change dynamically across time. By doing this, we are able to look at the current candle for its probability of being an outlier compared to the mean of the regression. If the spot price is outside the range of the expected deviation (e.g. +/- 2 standard deviations from the mean); a buy or sell signal is triggered.
IMPORTANT: This does not aim to validate the volatility of a trend, so the user must identify the historical fit. It is recommended to use the replay functionality to make these adjustments with historical data in order to avoid over fitting the model to the data; which will create long term issues with performance.
When a trend is found in the specified range; it is assumed that the white noise (movement +/- to the trend) happens in a normal & unbiased way. In a fair market; the buyers and sells should balance themselves out in such a way that there is no inherent bias outside of the trend. As such, we can assume that almost all movement within the trend will be within +/- 3 standard deviations. So if the selected deviation range is greater than that; it is likely that the model is being over fit to account for extreme volatility.
Below are examples of the indicator on different charts:
USDAUD
BTCUSD
AMZN
A2M