A lot of talk is about how you adjust your position size with respect to a given market's volatility, but I'm wondering if anybody had any ideas on how one might normalize a markets velocity and/or volatility over a given period of time.

My first idea was to take average true range over a given period and divide by the average price during the same period. Commodity X roughly has a range of Y percent each day for the last Z days. Seemed straight forward enough.

The problem is that my historical data (from Pinnacle) is adjusted in a fashion that allows price to go negative. Not only do the negative numbers foul up the calculation, but also very small closing prices result in a very high percentage which isn't "true".

So does anybody have any ideas on a simpler, possibly more robust, way to measure price velocity or volatility and normalize those figures across markets?

## Normalizing price velocity/volatility with respect to what?

- (Linear Regression Slope) / (Average True Range)

If you want to get fancy you can modify the numerator. You can choose to take the linear regression slope of a smoothed version of price, rather than price itself. For example you could take the linear regression slope of a 3-bar EMA of ((H+L)/2). Fans of the Jurik Moving Average will have additional devious ideas.