Page 1 of 1

Filter stock trades based on the market direction

Posted: Tue Mar 04, 2014 7:27 pm
by AdrianJReid
I have been trading long / short equities using mechanical trend following systems for some time. This has been quite profitable for the last 8 years so I am pretty comfortable with my system performance. I have recently been investigating the impact of turning the long side signals off when the market index is going down and turning the short side signals off when the market index is going up.

My initial hypothesis was that a trigger like this should reduce the drawdown on the long side and the short side, so when these two sides of the system were combined the equity curve should be that much smoother.

I thought this was a relatively obvious hypothesis - although often (as we know) what is apparently obvious or logical may not work in reality...

With various types of filters used to determine the direction of the index all I have found is that the equity curve degrades relative to leaving both sides of the system 'switched on' at all times.

Possible theories may potentially explain this:
1) The reduced opportunity due to the filter more than offsets any other benefit
2) When trading individual stocks with trend following systems, the stocks that make you the money are leading the market, so the lagging index filter keeps you out of the best trades
3) Combination of BOTH 1 and 2
4) Some other theory I have yet to come up with

I suspect that the main impact is from point 1 above, because when I invert the index filter (trade stocks long when the index is declining and short when the index is rising) the system still makes money (although the equity curve is a little ugly).

Conventional wisdom (not always right or useful in trading) I thought said that you should trade in the direction of the broader market, but these results seem to say trade both directions all the time to maximise your opportunity.

Has anyone had similar experiences to this? I would be interested to hear if anyone has successfully used the direction / state of a stock index to filter medium to long term stock trend trades to improve their equity curve.

Posted: Wed Mar 05, 2014 8:29 am
by Jake Carriker
You could check the correlation between your long only and short only equity curves. If the correlation between two return streams is low, the benefit of diversification is high. That might explain why your testing shows it best to stay invested in both.

Posted: Wed Mar 05, 2014 9:10 am
by sluggo
Maybe it would be worthwhile to consider changing the way you think about a "filter". So far you've considered the "filter" to activate a simple on/off switch. A digital signal, either One or Zero.

Another possibility is to have the filter control a continuously adjustable dial; a knob. Let's assume this knob can be set to any decimal number between 0.00 and 1.00. (zero percent and one hundred percent). Then maybe

Code: Select all

if(filter says UPTREND) AND (new long entry signal is received)
        positionsize for new long is normal

if(filter says UPTREND) AND (new short entry signal is received)
        positionsize for new short is (KNOB * normal)

if(filter says DOWNTREND) AND (new long entry signal is received)
        positionsize for new long is (KNOB * normal)

if(filter says DOWNTREND) AND (new short entry signal is received)
        positionsize for new short is normal
Setting KNOB=0.00 gives the same behavior you have now: in an uptrend, short entries are taken at zero positionsize (i.e. ignored), but in an uptrend, long entries are taken at the normal positionsize.

Setting KNOB=0.50 means you take short entries during an uptrend, but at half size.

Does this add a parameter to your trading system? Does it introduce another degree of freedom / opportunity for evil curve-fitting? Yes.

Posted: Mon Mar 17, 2014 10:09 pm
by AdrianJReid
Thank you for the great responses. I like both ideas and have done some further work on the correlation of the equity curves as well as the quality of trades from each system to determine whether it may make sense to adjust the position size of each system as Sluggo suggested.

I looked at the expectancy of the trades for the long system and the short system when the broader market was going up/down to determine whether the difference was material enough to bother adding an additional parameter.

When I did this work the difference between performance when the broader market was up vs down for both systems was not that great. However, using the filter dramatically reduced the opportunity to trade each system.

Therefore given:
a) both the long and short systems performed acceptably regardless of whether the broader stock market was going up or down; and
b) The returns from the two systems provide diversification; and
c) I am much more comfortable with a consistent level of long AND short equities exposure for safety in an extreme situation; and
d) Optimisation of the market filter is prone to curve fitting

My conclusion is that I should just trade both systems all the time.

So until I discover some other useful evidence, I continue to focus on individual stock price movements and ignore the broader market. In bull markets I get more long exposure and in bear markets I get more short exposure, but generally there is a bit of both long and short at all times.

Posted: Mon Mar 17, 2014 11:28 pm
by sluggo
"The purpose of simulation is to correct faulty human intuition." --- my boss.

Posted: Wed Mar 19, 2014 8:52 pm
by AdrianJReid
Thanks Sluggo - Sounds like you have an insightful boss!