Premium Market Analysis

Market Statistics

The Great Depression Pattern Unmatched

A chart that compares current price action pattern in Dow Jones Industrial Average to a pattern formed in the index from 1928 to 1930 has appeared in financial social media. Below are a few reasons why such comparisons make no sense.

My replication of the chart is shown below. The black line is Dow Jones Industrial Average between 1928 and 1932. The orange line shows the index from 01/02/2019 to 07/07/2020 with hidden scale.

Note that in the chart above I matched the peaks in respective periods whereas in the chart shown in financial social media the current pattern in the index is shifted forward to match the price decline in the past but this does not make any difference as to the significance of the pattern matching.

Briefly, the probability of finding and arbitrary pattern in historical data that loosely matches any other pattern is relatively high and depending on the looseness of the fit it approaches one.  However, the probability that price evolution of the two patterns will match forward is close to zero. Below are some of the reasons why.

  • There is no causality between patterns and market moves.
  • Economic conditions and tools to deal with them are different.
  • Even if patterns match forward, that will be a random outcome.

There are uncountable possible paths prices can follow in the future. The paths will depend among many other things on volatility and drift. If we assume the drift is zero and annualized volatility is 20%, then a Geometric Brownian Motion model applied for about 1000 days after the end of that Great Depression pattern yields the following for 200 possible paths.

The above model says that the index can fall below 10,000 or even rise above 50,000 in the next four years. The mean value is about 26130 and the standard deviation is about 5,728.

If we decrease the volatility to 10%, the mean index value is about 25700 and the standard deviation drops to about 2,800. The extremes are somewhat smoothed with maximum below 50,000 and minimum around 15,000.

How prices will evolve in the future assuming no surprises will depend on volatility and drift and these are random variables. Patterns cannot account for these parameters in non-stationary series. Patterns can be useful only in short-term forecasting and depending on the significance the expectation will vary widely.

Naive pattern matching of the kind shown in the first chart above is a remnant of old classical technical analysis. Finance has progressed a lot and quants nowadays know that pattern matching of this kind is an exercise in futility. Obviously, the index may fall below 10,000 in the next four years but any relation to the pattern shown in the first chart above will be purely accidental.

Charting and backtesting program: Amibroker

Data provider: Norgate Data

If you found this article interesting, you may follow this blog via push notifications, RSS or Email, or in Twitter


Price Action Lab Blog Premium Content