Evolution of a Timing Model by Nelson Freeburg
How does a trading model evolve? Formula Research newsletter publisher Nelson Freeburg discusses the evolution of a trading model and explains such concepts as out-of-sample testing, parameter sensitivity testing and the inclusion of nonprice indicators for timing trades.
Many come to trading and investment with high expectations. In my case, the markets soon made
clear the limitations of my intuitive trading skills. There are talented traders who can sense an impending
market move, adjust for risk and enter with a properly scaled commitment, all with instinctive finesse.
For me, hard experience in the stock, options and commodity markets told me I was not so blessed.
For those who lack fine trading instincts, mechanical timing systems are a ready alternative. Here, the
buy and sell signals are clear and precise. A structured algorithm dictates market action, not a fleeting
insight that is subject to emotion. Trading rules can be tested over a broad sample of the data, noting how
performance varies through different market cycles. Not only that, the track record is objective, not
anecdotal. Everyone who tests the system will get the same results, a big plus.
The chief problem with mechanical trading systems is uncertain future reliability. Timing methods that
test well in theory may break down in practice. Trading systems can be so elaborately patterned on the
past that they no longer work in real time. The fact is, a determined signal jockey can methodically tweak
a system to generate almost any desired result. Such curve-fitting is self-defeating. By overoptimizing to
a narrow range of market conditions, poor results in actual trading are all but assured.