George Lindsay's Famous Topping Pattern Returns

We must point out from the outset that in spite of the fact that self-similar patterns are fairly frequently observed in the stock market, one must heavily discount their predictive power. The so-called 'Three Peaks and a Domed House'  pattern first described by George Lindsay has become quite well known because variations of it appeared prior to the crashes of 1929 and 1987, and in less obvious form on a number of other more or less memorable occasions (which is why we call it a 'doomed' rather than a 'domed' house). When Lindsay published his paper on the pattern in 1971, he listed the following examples (including partial pattern formations):

July 26, 1893 (point 1) to Sep 4, 1895 (point 23) in the Dow Jones 20 stock average.

July 26, 1910 (point 1) to Sep 30, 1912 (point 23) in the Dow Industrials.

October 1946 to November 1948

September 1964 to May 1966

Point 3 to point 23 in the Dow Industrials:

Oct 22, 1915 to Nov 21, 1916

June 5, 1919 to Nov 3, 1919

Sept 11, 1922 to Mar 20, 1923

Feb 5, 1929 to Sep 3, 1929

Nov 17, 1945 to May 29, 1946

Sep 13, 1951 to Jan 5, 1953

Apr 6, 1956 to Jul 12, 1957

Aug 3, 1959 to Dec 13, 1961

Lindsay also noted that several examples of an inverted form of the pattern could be found (a 'domed house' followed by three peaks). The schematic of the pattern looks like this:

Lindsay's Three Peaks and a Domed House (3P+DH) pattern, schematic – click to enlarge.

Due to the fact that the pattern is quite complicated, it is almost never seen in its 'pure' or in entirely completed form. But if a pattern that looks very similar pops up, then one should perhaps consider whether other data possibly indicate that its message should be heeded. Here is what the DJIA looks like over the past year:

Not a 'perfect' 3P+DH pattern, but very close – click to enlarge.

Pattern Similarity with 1929 (Spoonful of Salt Recommended)

The reason why we are taking another look at the pattern is that Tom DeMark pointed out back in October that there was a developing self-similarity between the current DJIA pattern and the pattern built prior to the 1929 market top  (we wrote about this at the time in “Which Way Will the Cookie Crumble”). DeMark showed this chart in October. Since the 1929 pattern was what originally inspired Lindsay's 'Three Peaks and a Domed House' pattern and his search for it elsewhere, it followed that this had to be the pattern that was going to appear if the DJIA were to continue to follow a path analogous to that of the 1929 Dow.

Tom McClellan heard about DeMark's comparison chart and since he is a big fan of market analogs, started to follow the pattern similarity as well. Here is a chart McClellan posted in late November:

Tom McClellan's November update on the tracking of the 1929 pattern by today's DJIA (the 1929 pattern was the original '3P+DH' pattern) – click to enlarge.

What makes this whole exercise rather eerie is that ever since Tom DeMark first pointed out the similarity – recall that at the time only the 'three peaks' had formed – the pattern analog has continued in a near perfect manner. Here is an update of the chart which McClellan posted in December:

In December, the DJIA had taken another step toward looking like a near perfect 1929 analog … - click to enlarge.

In the article McClellan posted along with this chart update, he made a number of interesting points addressing criticisms of the exercise. One important point is that even if the patterns continue to follow each other (this is to say, a big price decline occurs once the pattern is completed), it e.g. tells us nothing about the extent of the decline. It could be a far smaller, less remarkable correction, but still  a correction one would either want to sidestep or exploit.

One obvious point of criticism is that today's economic fundamentals are completely different. This is of course true: no period of history is exactly similar. This is also why economic history is essentially useless as a 'test' of the validity of economic theory. Theory can only be disproved by ratiocination and logic, as it is not possible to conduct repeatable experiments in the social sciences.

However, that does not mean it is completely futile to consider pattern similarities in the stock market or other financial markets. After all, there is evidence that many patterns do tend to repeat, regardless of the precise fundamental backdrop. Among the possible reasons for this could be: a few especially important individual factors are perhaps similar.  For instance, the bubble of the 1920s was a result of heavy monetary pumping, which the Fed thought would be harmless because there was no 'inflation' (i.e., CPI failed to increase much throughout the boom period). That is certainly a major similarity to today, even though the situation differs in a great many details.

Another point worth considering is that financial markets reflect psychological factors to some extent. Especially so-called 'herding' effects can definitely be observed. These are quite prevalent among fund managers, not least because  they are under great pressure to deliver performance close to their 'benchmarks' over rather short time periods (this is to say, between one quarterly report and the next). We can see this in various positioning and survey data, such as those we have recently discussed. It is probably not too big a stretch to conclude that these mass psychological effects can be reflected in certain price patterns (see also Sornette's work on bubbles and 'anti-bubbles'. For instance, Sornette discovered that in their late stages, stock market bubbles tend to exhibit declining short term volatility).

Anyway, this is probably a good moment to stress again that we have grave doubts about the predictive value of such patterns, as we cannot foresee with certainty what future states of knowledge traders will act upon. Specifically it should be noted in the context of the current situation that this is not the first time that the 3P+DH pattern has appeared since the 2009 low in the stock market. We know because we first discussed it already more than a year ago, and so did several other observers at the time (including, as it happens, Tom DeMark!). If you follow the link, you will find further links in the article to numerous other articles about the pattern's appearance at the time. Today we know that it was a false alarm, as the power of the printing press has evidently been underestimated by many observers.

This incidentally remains a major caveat at this moment in time as well: last we looked, the year-on-year growth of the broad US money supply TMS-2 was at just above 8%, historically a magnitude that was sufficient to support the stock market for a good while yet (keep in mind that changes in money supply growth only affect prices with a lag). Of course even with regard to this there is a lot we don't know with any degree of certainty. For instance, we don't know how big a decline in money supply growth will suffice to upset the apple cart this time around. This particular threshold is likely to be higher than in the last boom, but that is just a guess (we have pondered this topic inter alia on this occasion).

Anyway, the 'eeriness factor' has in the meantime increased further, as this very recent chart update of the pattern analog published by Zerohedge shows:

A recent update of the analog by Zerohedge – click to enlarge.

We would also note, the mere fact that this chart analog has received so much exposure lately makes it automatically less reliable with regard to its potential predictive value. We know this only from experience and the assertion is solely based on anecdotal evidence, but recall e.g. the frequent mentions of the 'Hindenburg Omen' in the financial press in recent years (a technical signal that measures the dispersion of new highs and new lows and therefore tells us something about he breadth of an advance. If the market is close to price highs, then the appearance of the 'omen' may indicate distribution). The rule of thumb with that particular indicator was clearly that the more articles about it were published, the less meaningful it turned out to be.

That could well be the fate of the above chart analog as well, but it does coincide with a number of technical and sentiment extremes, not to mention a new high in margin debt, so one should probably not dismiss it out of hand either.

In closing, we want to remark on another interesting factoid: so far, the market's price high was on the penultimate trading day of last year (on the final trading day, the market closed early). This is a similarity to the Nikkei's 1989 high, which also occurred right at year end. So there is certainly no lack of strange coincidences (enter fear strings).

The Nikkei during the last months of 1989 and the first half of 1990. On the final trading day in 1989, it reached what to this day is THE high of almost 39,000 points – a price it never again came close to – click to enlarge.

Charts by  Carl Futia (schematic), StockCharts, Tom McClellan, Zerohedge, BigCharts