Thursday, October 16, 2008

Sell New Highs - Correction

OK, I messed up. Humble thanks to Ramon for catching the error. Yesterday's Sell New Highs system had some poor coding on my part. I think that derives from my use of the HighestFC function to identify new highs. I originally decided to use that function because the intent was to scan a larger universe of ETFs and optimize the lookback period. For the IWM that lookback period ended up being 1 day and as a result TS generated trades both on the next bar and on 9 bars out, hence the 4 day average holding period. I should have seen there was a problem when a sytem generates 50 trades in a year with a 9 day hold. The chart above reflects the error in the coding, which has now been corrected to read as follows:

If Close > Close[1] and High > High[1] and Low > Low[1]
Then Sell This Bar at Close;
If BarsSinceEntry=2 Then ExitShort this bar at Close;

The resultant system has the same frequency of trades as the original faulty system, but the holding period (and risk exposure) is reduced to just 2 days . . a holding period I can live with. This is actually an attractive system to me as with the corrected code, the system has a relatively low intraday drawdown, a good max winners/max losers ratio, and trades on an average of once a week. If nothing else, the system should serve as a warning to momentum trades who still like to buy new highs.
Thanks again Ramon.


Mateo said...

I had a question about what your criteria is for actually entering into the market with this system?

bzbtrader said...

As with all these systems, I use them to guide my daytrading and occasionally a 1 or 2 day hold. When several of the non-correlated systems generate a BUY or SELL, that guides my trading bias, long or short for the day within the confines of the intraday pivots. This is how I use my basket of systems approach and the basis for the KOP studies.

Mateo said...

thanks alot