Another generic question:
I am interested in what your opinions are on 'tuning' or optimizing your tests.
Basically I crafted nifty binary wave combo that uses 6 different optimized macd tests, and of those, a binary wave is created from the 3 most successful to yield a final result. It has worked surprisingly well on a variety of stock issues that have relatively well behaved personalities in both up and down trends. (when I say well-behaved, I mean that the price dosen't spike all over the place randomly...). It is constantly beating buy and hold, and usually yields between 90% and 350% returns over shorter (4-6 month) timeframes of any of the stocks I have tested it on so far. It does not perform as well over longer term timeframes.
The reason that the numbers work well within the shorter timeframes, seems to be because they are optimized to suit the pattern of the stock over the input time-period. Obviously if I take those numbers and work them over a different period, I get less than stellar results. My thought is that I may want to consider using this system, but after a certain time period, say a month or so, It seems like it will be necessary to re-optimize to adjust for any new price patterns. MS dosen't seem to provide an easy way to test this particular situation out, as you'd basically have to add up by hand what your current position was in each timeframe you are backtesting, and when you re-tune, you'd have to see weather the new variable numbers keep you on the same side of your current trade, etc.
I would just like to know what some of your thoughts are on this, weather trading this type of system would be frowned upon for any reason, or weather perhaps re-tuning optimization is commonplace with systems that perform well short term, and what criteria should be used.
Thanks again,
Robert |
