|
|
 |
|
|
ADX vs. Random Strategy
Originally published November 24, 2007
Page 1
There is something intoxicating about the idea of automated strategies. What could be better than to program a black box to do nothing but sit on your desk and churn out profits day after day, month after month. Those new to the game can easily get carried away with dreams of riches as they develop and backtest strategies. Most programs allow the developer to automatically optimize variables. Often strategies which initially look rather mediocre can be optimized on a particular stock to deliver excellent backtested returns.
It is easy for the uninitiated to imagine the immense profits which could be churned with a strategy with some good profit statistics. It is a simple matter to develop an optimized, backtested strategy with an annualized return of 185%. And it is just as easy to get carried away with ideas of the expensive cars and boats one could purchase with the proceeds especially if the profits were reinvested.
Assuming this interest rate, compounded monthly, in three years an investor beginning with $10,000 would see the investment mushroom to $1.74 million. Not bad considering the entire system could be sitting in an empty corner in the spare bedroom quietly churning profits. No human intervention required.
Of course there is a problem.
In the majority of cases it never works out this way.
The transition from idealized, backtested results to real world profits is a notoriously rocky one which brings to mind the old chestnut ascribed to Captain Edward Murphy: “If anything can go wrong, it will.” Just as machines designed to travel to Mars often crash and burn before they ever perform useful work, or drugs intended to correct human maladies wind up harming unsuspecting patients even after rigorous clinical trials, so it seems automated strategies have a particular knack for failing to generate profits once they are applied to real world markets.
In the May 2006 Issue I showed one interesting example of how misleading results from backtesting can be. I developed and tested a strategy called the “Black Box HR”. By virtually all accounts it was an excellent strategy. In just over two years the strategy managed to turn a $54,000 gross profit from an initial investment of $100,000. Assuming monthly compounding this represented a very respectable 22% annualized return. But much more impressive were all of the trade statistics. For example a full two thirds of the trades were profitable. The maximum consecutive winning trades completely eclipsed the maximum consecutive losing trades at a ratio of 4 to 1. The Profit Factor was outstanding. Dividing Gross Profit by Gross Loss produced an astonishing figure of 5.16, something most strategy developers would be overjoyed to obtain.
Looking at these and many other trade statistics would lead many people to conclude that this was an effective, well designed strategy.
But in fact it was generated completely by chance.
Figure 1. “Black Box HR” Strategy Performance on Daily S&P 500. Jan. 1995 - Feb. 1997 Report Generated by TradeStation |
|
|
|
|
