Decisions, Decisions

I have been bothered for some time regarding two different formulae that I could use for measuring industry group strength over n days.  I've finally decided to switch from the fmedian(...) formula to a variant.  It provides a higher level of profit but has a disturbing drawdown at the end of the equity curve.  I ultimately decided that the benefits of the new formula outweigh the additional drawdown.

Up until now I have been using the following formula to determine industry strength:

FMedian("Close(0)/Close(n)",#Industry) > X

where n is the lookback period and X is the threshold. 

The idea is to determine whether the median value of Close(0)/Close(n) within the industry group is on the rise.  "On the rise" is determined by comparing Close(0)/Close(n) to a threshold value.  If above this value then the industry group as a whole is deemed to be on the rise.  But I have been playing with a variation of this formula as follows:

FMedian("Close(0)",#Industry) / FMedian("Close(n)",#Industry) > X

I would have thought that this variation would give same or very similar results to the original formula.  But as can be seen from the following graph the results are somewhat different particularly for the last half of the equity curve.  n = 160 and X = 1.2 for the following equity graph.

I am not sure which formula is the "correct" version or even if there is such a thing as a "correct" version.  The new variation gives higher profit over time but also incurs a hefty drawdown at the end.  After some sole searching I decided to opt for the new variant.  I like the more continuous progression in the second half of the equity curve.   The drawdown is caused by one transportation industry group.  With good money management i.e. limiting percentage trading capital for any sector and industry and stops should make this drawdown much smaller.  And don't forget the market conditions are very unusual.  We may never see these conditions again in our lifetimes.  So I've made my decision.  From now on I will be using  FMedian("Close(0)",#Industry) / FMedian("Close(n)",#Industry) > X.

Something I need to clarify before going any further is that the screener does not constitute a "system".  There are a variable number of stocks called up each week from zero right into the hundreds.  So the next step will be to attempt to develop a ranking system that is compatible with this industry breakout strategy.  This is usually easier said than done...  But I love the challenge.

All for now.