|
PureBytes Links
Trading Reference Links
|
On Feb 13, 5:52am, Paul Cote wrote:
> Subject: Trading System Design-choppiness index
> Realtraders:
>
> I would like to know if anyone has a method for determining or
> evaluating the choppiness of the market in order to determine a
> choppiness index. This will be to switch the trading system from say a
> volatility breakout to a trend following system when a strong trend
> comes underway. I would like the system to be predominantly a
> volatility breakout of say some fraction of the 3 day average true range
> from the prior days high and low.
Since a volatility breakout system is also a trend-following
system, I'm not sure what you see as the difference? Perhaps you're
looking for a filter that will tell you when to ignore certain volatility
breakouts that would cause you to reverse against a strong trend?
A while ago, I did a study of channel breakout systems (not all
that different from volitlity breakout) where I recorded the value
of several parameters at the point that a new signal was taken.
I looked at (1) statistical volatility over the past 20 days
(essentially the standard deviation of the daily change, described
by McMillan and Natenberg in their books on options), and (2)
Wilder's ADX taken over a 20 day period. I then plotted the
profitability of that signal versus the other two filter parameters.
I looked at a few different markets, but focused mainly on the S&P
and long bond (which are probably not the best candidates for trend
following systems).
I found that the breakout system performed poorly at lower
levels of volatility - note: not the _relative_ volatiliy used in
volatility breakout but _absolute_ volatility (for example, if the
annualized statistical volatiliy over the past 20 days is less than
10%, don't take the breakout). Performance also dropped off at
very high levels of volatility (eg, if the statistical volatility
is over 25%). A certain level of volatility seemed necessary
to make the breakout system work (I was looking at intermediate
term systems such as channel breakouts with 10 to 20 day lookbacks,
so my results won't be equivalent to the 3-day volatility breakout
you're using).
Using a trend indicator like ADX didn't have much effect, and
looking at the plot of profit versus ADX, there was no particular
relationship (higher ADX readings occur when a strong trend has
been in place) to overall profit on the trade. So, even though
it seemed like a good idea to only take breakouts in the direction
of a strong trend already in place, this idea didn't seem to pan out.
Perhaps that's because the S&P and bond have tendency to retrace.
Whatever method you use, I'd suggest that you look at the effect of
your input filter paramaters on the profitability of the trade.
--
--
| Gary Funck, Intrepid Technology, gary@xxxxxxxxxxxx, (650) 964-8135
|