Forcing Optimization
by Vincenzo Sciarretta
Sooner or later, most system-oriented traders backtest a mechanical system and get a positive result -- and then have it perform poorly in real-life situations. What gives?
Sooner or later, most system-oriented traders experience this dismaying conundrum: They formulate a mechanical system, backtest it over a convenient period, and get a very positive result. They watch the equity line and see it rise in an almost straight line with small but acceptable drawdowns. Things look downright rosy. But then they start trading the market and, despite the positive backtesting, proceed to lose money with the system! Why this occurs with dismal and predictable regularity is too expansive a subject to get into here; I will deal with it at another time. But we can discuss how to limit this sad but all-too-common occurrence, with a methodology called forcing optimization.
FIGURE 1: BACKTESTING COFFEE, 1977-81. The equity line on top rises steadily during backtesting, falsely indicating a robust trading system.
As an example of contrary results, I used a mechanical system from Martin Pring's excellent Technical Analysis Explained. The system presented by Pring was devised for the dollar/sterling on a weekly basis, but that's not the way I'll be using it here. This mechanical system combines a moving average and two rates of change, a short one and a long one. Here are the system rules in general terms:
Rule 1 Go long when price is above its x-day moving average, the y-day rate of change is above zero, and the z-day rate of change is above zero.Rule 2 Go short when all conditions in rule 1 are reversed -- that is, when price is below x-day moving average and both rates of change are negative.
Figure 1 shows backtesting of coffee for a five-year period, 1977 to 1981. On top you have the equity line for this set of parameters: moving average, 30 days; short momentum, 15 days; long momentum, 25 days; entry and exit commissions, 0.5%; and margin requirement, 100%. The results look good, but check out Figure 2.
The result is typical: a trader optimizes a mechanical system, gets a wonderful performance (here, $100,000 is transformed into $2 million), has a satisfactory equity line and then begins trading the market -- only to get a completely different result from the one that is expected.
Vincenzo Sciarretta is a member of the Italian Society of Technical Analysis (SIAT) and can be contacted via his Website at https://www.sciarretta.com/.
Excerpted from an article originally published in the February 2000 issue of Technical Analysis of STOCKS & COMMODITIES magazine. All rights reserved. © Copyright 2000, Technical Analysis, Inc.
Return to February 2000 Contents