Thursday, March 19, 2009

When Good Indicators Go Bad

What? No charts?. . . . No visual aids today.
Yesterday Woodshedder posted a thought provoking little piece in which he lamented the lagging performance of his much tested and newly deployed trading system. The exact rules of the system are proprietary so I can't indulge my usual fiendish glee in deconstructing the system and applying my own BZB spin. That's not a criticism, just a fact. Each system developer owns his ideas and decides how to share them. I'm perfectly happy revealing the source code for my humble little systems since I suspect the trading volume in the Qs and GE stimulated by my BUY or SELL signals is not going to skew market momentum one smidgen. That's just my approach.
But I digress.
My point is that I too have noticed a recent flagging in the performance of many of the dirty dozen systems that is reflected in the flattening of the system equity curves. I've mentioned this flattening in several previous posts, without providing a coherent explanation of the phenomenon.
Unfortunately, I still don't have a defensible argument for the flattening equity curves but it's a topic that has consumed much of my time recently as I seek to explore probable causes.
What I have noticed recently is that the dirty dozen systems applied to daily bars tend to UNDERPERFORM the same systems on shorter time frames, including 60, 10 and 2 minute bars. Within this context I also have to caution that the various system inputs must be re-optimized when used in conjunction with intraday time intervals. Sometimes the adjustment is minor, sometimes a bit more dramatic.
While I may be accused of curve fitting the backtest data to create a better looking performance report, I'm going out on a limb and argue that the differential volatility of the various intraday time frames versus the daily bars gives credence to this re-balancing approach.
I know from emails I have received from several active trader viewers that my observations about the systems performing better intraday than daily are supported by their own trading activity.
So what to do?
For my own part, I continue to focus on what works (for me). That means concentrating risk exposure on intraday bars using my reliable stable of signals including the pivots, the NYAD, the MA crosses and the parabolics.
You might call that getting back to the basics, and while I've been able to accumulate a nice little equity bounce with this latest rally, I'm concerned about the sustainability of the surge and am therefore pulling back to my daytrading mode for the near term while I begin a new round of system testing and refinement.


Cucca said...

Hmmmmm, the ATR (14) has dropped below 1 on the Q's. I've also noticed I'm not "SEEING" very many setup's on my "visual" based guessing "systems", IE, I'm struggling.We seem to get a "move", and then sit there for a while, I don't know, maybe it's my Biorythmic scale is down.

Woodshedder said...

BZB- I've been thinking for some time that some patterns/cycles are scaling down into shorter and shorter time frames. Might explain what you're seeing.

I've also noticed (hope to get a post up sometime this weekend about it) during periods similar to the past 10 trading days that what works is trend following (duh) i.e. buying above a short MA and selling-short beneath it, on the dailys.

If that can keep working, it will mean things are really changing.