The Economist Weighs in on Technical Analysis (A Must-Read for Fibonacci Fans)

This week's Economist has a provocative article on the use of Fibonacci numbers and other technical indicators by traders and market analysts. Its starting point is a recent study by professors at Cass Business School in the UK. Entitled "Magic numbers in the Dow", it reports the results of a giant backtest on Dow Jones market data using Fibonacci percentages in an attempt to predict market reversals. Turns out they weren't much help at all.

The Economist article then goes on to assert that the whole field of technical analysis is on shaky scientific ground, with very little evidence to support its claims that certain indicators can consistently predict certain trends. Overall, it's a well-written and rightfully skeptical look at some of the underlying assumptions of technical analysis and the "chartists" who use it. Here are some particularly memorable excerpts:

• "The academics looked at the Dow Jones Industrial Average over the period 1914-2002 and found no idication that trends reverse at the 61.8% [Fibonacci] level, or indeed at any predictable milestone."
  
  
• "The recommendations of technical analysis can be so hedged about with qualifications that they can validate almost any market outcome."
  
  
• "Too often, rules are so vague and complex as to make replication impossible"
  
  
• "Some technical predictions may be self-fulfilling; if everyone believes the dollar will rebound at 100 Yen, they will buy as it approaches that level."
  
  
• "Chartists fall prey to their own behavioral flaw, finding 'confirmation' of patterns everywhere, as if they were reading clouds in their coffee futures."
  

I am in complete agreement that those using technical analysis can skew their judgement through selection bias and wishful thinking - in fact I wrote a whole post about it. I also agree that before using any technical indicator, you should be able to formulate its predictions as a testable hypothesis, and should then rigorously backtest those predictions with as much historical data as you can get your hands on.
However, based on my own trading experiences, I question the rather unkind generalization the writer then goes on to make that the entire field of technical analysis is largely bunk. My personal rebuttal to this assertion is that (a) I've created a trading system based entirely on multiple technical indicators (b) I've backtested it thoroughly (c) the results don't look too random to me. The day this system starts making completely useless predictions is the day I'll give some credence to sweeping claims made by skeptics of technical analysis.

I also question the assumption made in both the university study and the article that traders are using the same rigidly-defined technical indicators all the time in all market conditions. While all of my indicators were created using certain basic concepts of support, resistance, continuation, and reversal, I then tweaked and optimized each of them extensively based on actual market data. And I continue to tweak and optimize them all the time, because the market's behavior changes, and therefore the signals need to change too. I think most experienced traders would look at their signals as dynamic, flexible entities that must frequently be retested and recalibrated lest they get rusty and break down - or even discarded, if they've stopped working altogether.

As for the specific conclusions reached about Fibonacci numbers and their predictive abilities (or lack of them), it could very well be that they're of no use in predicting the Dow, at least in certain timeframes. However, they may be of use in predicting activity in other markets. I've found there's a great deal of variability in how effective signals are based on (a) what timeframe they're being applied to and (b) what market they're being used in. I don't personally use Fibonacci numbers because I find them too rigid a signal - instead, I've identified other ratios that seem to be particularly relevant to the EUR/USD market, and I'll continue to adapt these ratios in response to changing market conditions. Again, flexibility and adaptability are the keys here.

If reading these kinds of studies shakes your confidence, the best way to regain it is to go back to the data and see what it tells you. If you find a signal performs consistently and accurately over time, then congratulations, you've found a type of technical analysis that works. If your signal fails miserably, then toss it out and look for another one. Repeat as needed. And whatever you do, don't take an isolated study on Fibonacci numbers and the Dow as the last word on technical analysis!

Further reading:
The Economist: Technical Failure
Magic numbers in the Dow

Related topics:
Designing a Trading System, Tip #1: The Data is Your Friend
Beware of selection bias when designing your forex trading strategy. (Or, how wishful is your thinking?)