Re: Standardizing a Volatility Filter?

[heypeter@xxxxxxxxxxxxxx]

Hey BJ

1/ Try with something simple (maybe elegant?) first as yur definition of
volatility eg range or true range. 
2/ I would suggest a slightly different and SIMPLER yet standardized
approach . . .
- choose  a period depending on yur time frame  eg 50 days cf 100 days
- calc the average and stddev
- add/subtract some percentage of the stddev from the average to get
your bands, and hence areas of hi or lo vol.
something like this . . .

Inputs: Period(50), Factor(2);	{eg use a 50 period moving window and 2
std dev from mean}
Vars: AverageVol(0), DispersionVol(0), LoVol(0), HiVol(0);

AverageVol = AvgTrueRange(Period);
DispersionVol = Factor * StdDev(TrueRange,Period);
LoVol = AverageVol - DispersionVol;
HiVol = AverageVol + DispersionVol;

Condition1 = TrueRange > LoVol and TrueRange < HiVol;
etc . . .

- try playing with Period and Factor
- off the top of my head (educated guess) you may want to use different
Factors when calcu;ating hi and lo vol, as price changes (volatility?)
are not normally distributed, as defined above you'll get more
occurrences of hi vol than lo vol.  So amend the code  and use say a
factor of 2 for hi vol and 1 for lo vol if ya want. . .

NB no hard and fast rules here as we are dealing with a moving target
(non stationary distribution) - that's what makes this game so damn
interesting . . .

Good luck

E. Peter K. Chan
Independent Private Trader


Peter Iovanella wrote:
> 
> I'm trying to add a volatility filter to a system I'm writing, but have come
> across two problems:
> 
> 1) Which formula to use to figure out volatility, specifically in futures
> markets, and whether such a filter is appropriate to use in futures markets
> (any ideas?)
> 
> 2) How to keep the system from producing a signal if volatility is within,
> say, one standard deviation of the maximum and minimum historical values for
> volatility in a given market.  This problem would be a bit easier to solve
> if volatility had equal ranges in different markets, but I have found that
> it varies significantly.
> 
> So, as far as I can tell, it looks like I need a formula that will
> 
> a. Determine volatility in a given market;
> b. Determine the high and low values for volatility in that market;
> c. Determine standard deviations based on the volatility range;
> d. Add one standard deviation to the low value and subtract one standard
> deviation from the high value;
> e. Output this result into the following filter formula:
> 
> Condition1 = Volatility(10) > {insert (low volatility+one standard
> deviation) here}
> AND Volatility(10) < {insert (high volatility-one standard deviation) here};
> IF Condition1 THEN Begin;
> 
> In short, what I'd like to do is come up with a formula that will allow me
> to standardize this filter.  Any takers?  Thanks!
> 
> BJ