JorgeStolfi
Hero Member
*****
Offline Offline

Activity: 910
Merit: 1002



View Profile Personal Message (Offline)

Ignore
March 18, 2014, 08:20:07 PM

So as I mentioned, I'm thinking of dropping a volatility index in with ChartBuddy's posts. It's fairly simple so far, it is the base 10 log off the percentage of high-low against closing price ( log(100*(high-low)/price ). This would only be once per day at 3pm (when the NY market closes). Here is a sample with a few arbitrary currencies (price, high, low, index). Prices are from Bitstamp. Thoughts?

Code:
GBPUSD 1.6573 1.6648 1.6547 -0.215079756574918
EURUSD 1.3915 1.3943 1.3880 -0.344142661216467
JPYUSD 0.0099 0.0099 0.0098 0.00436480540245543
UAHUSD 0.1024 0.1041 0.1024 0.22014896473846
BTCUSD 612.72 612.72 591.13 0.546990585818375

Sounds good, but why put the closing price in the formula? You could use  the  geometric mean of L and H
instead, namely m = sqrt(H*L).  Then your indicator would be
   d = log10(H/m) = log10(m/L) = (log10(H)-log10(L))/2 = log10(sqrt(H/L))
That is simply the half-width of the [L,H] interval in log scale. 

To make the number easier to understand you could convert it back from log scale to linear,
namely use D = 100(10^d - 1) = 100(exp(ln(10)*d) - 1).
That is simply he percentual increase of H relative to m (or of m relative to L).

In any case, it would be very interesting to plot your indicator as a function of the time span considered to compute H and L (hour, day, month, year).  The shape of that plot should be related to the fractal dimension of the price plot.