Sunday, 17 April 2011


It sounds like classic Hollywood B-movie shlock.  

But as world stock markets storm back towards their highs in the wake of the March 2011 'nuclear correction', history may be about to inflict its equivalent of a mushroom cloud on this herd of raging bulls; because while they are busy making projections to sunnier uplands on the Dow, a number of important markets now appear to be forming a major top, in the image of a pattern known to technicians as 'Three Peaks and a Domed House'.


In 1970, technical analyst George Lindsay published a paper describing a pattern which, he claimed, was found in the lead up to around half of all bull market tops in the Dow Jones Industrials since the turn of the century. Its peculiar name comes from a series of precise peaks and troughs, which he numbered from 1 to 28, appearing in a sequence taking on average eighteen months to complete.  

The clarity of the structure means that, if correctly analyzed, it should be possible to identify the peak of a bull market with some precision - and to exit before the start of what is always a major decline.  

What is fascinating is that a number of emerging markets, as well as Australia, now appear to be in the final stages of forming this pattern.  Given that emerging markets have over recent years become a reliable leading indicator of our own, it is more than likely that major weakness there would soon be reflected in US and European stocks.  

By implication therefore, we are talking about the end of the world's two-year stock market rally and, conceivably, the beginning of a new bear market.

Here is a schematic diagram, based on Lindsay's findings*, which shows the idealized progression of the sequence.

*In the late 1990s, for his work on this pattern and other writings, Lindsay was posthumously awarded a lifetime achievement award by the Market Technicians Association.  

Source: Claasen Research, LLC

Following a fascinating note on the subject from top technician Matthew Claasen, I decided to undertake a thorough check through my database of the Dow Jones Industrials Average to see if this pattern is all it's cracked up to be and what, if anything, history might be trying to teach us.

Lindsay's analysis stretches back to the 1890s, but in the absence of detailed charts from that era I was only able to check back to 1922.  Still, you would hope that if there was anything worth gleaning 90 years would be enough.


Mark Twain lectures us that history sometimes rhymes but rarely repeats.  Nowhere is this embarrassingly clearer than in technical analysis where, within the utterly predictable cycles of fear and greed, price movements constantly befuddle chart watchers with their endless subtle variations.  

Sometimes the mind can play tricks and find patterns in market moves which are entirely random.  But the precision of this particular sequence and the fact that it has proved to be predictive on numerous occasions since publication in 1970 - not to mention the fact that it has always preceded a significant decline (minimum 13%, maximum 90%), makes this one at least worth taking seriously. 

According to my analysis, from 1922 to date just sixteen examples exist in the Dow Jones Industrials Average: this includes those cited by Lindsay plus instances generated since his paper was published.  The paucity of data is due to the fact that not all bull markets end in this way and, by definition, the peaks only come around once every several years; plus the average pattern takes a year to eighteen months to complete (a couple of examples took just four months while a few stretched to two years). 

A statistician would caution against making any iron-clad predictions based on such a limited sample - but then, no sensible investor even has the phrase 'iron-clad' in his vocabulary.  


Lindsay's own notes describe the point-by-point 

progression of the sequence

One can (and later in these pages I will) prescribe a couple of technical rules which make it difficult, though by no means impossible, to create a 'false positive' and see a pattern where there is none.  But looked at from a market timing perspective, those wishing to capitalize on any forthcoming market decline can still create an extremely favourable risk/reward set-up with the use of careful entry and stop-loss strategies, as we shall see.  


1923 - 2007

I'll illustrate each instance with a chart, then we'll take a look at the current situation and potential timing and trading options.  Rather than blind you with a blizzard of numbers I'll keep the charts simple in this section so you can concentrate on the bigger picture.  Also included for your viewing pleasure with each is a MACD momentum indicator.  Declining yellow trendlines show whether there are any divergances between indicator levels (lower highs) and price action (higher highs) - again, we'll discuss this later.


Decline: -19% 

Dow Jones Industrials Average, daily.
Red line:  200-day Simple Moving Average
Indicator:  MACD 12,26,9.

The initial Three Peaks are circled in red.   Note how, in this and all the other examples, the final trough of the three (point 10 in the progression) is the steepest and drops below the previous two, separating the Three Peaks from the remainder of the pattern.  

The market then begins construction of the Domed House.  Prices rise in three stages: a base comprised of two more highs & lows, a sharp move up to a plateau or first storey which carves out three further highs & lows, followed by a final ascent to new market highs at the dome.  A decline then begins which results in price collapsing to, at the very least, the lowest trough of the entire pattern - although in many cases it falls well below that point.  

Here are the other fifteen examples.  Click on each to view in closer detail.


Decline: -90%

Although it doesn't look like it at first glance, the greatest momentum reading for this bull market was in December 1928, almost a year before the top.  All 16 instances feature a divergance between peak price and peak momentum.
















Decline: -45%

A strict interpretation might fairly quibble with the inclusion of this example, since the three peaks are more like two-and-a-quarter.










Decline: -19%

Again, the form police will argue that the third peak in this example, which fails to fall to a trough below the previous two, disqualifies it from consideration.  One might reasonably respond that the unusual violence of the decline from peak two (12%) makes a greater decline from the third peak unnecessary.  Additionally, there is a pattern-within-a-pattern - with three peaks beginning in April 1998 leading to the dome in July.


Decline: -54%

Total instances: 16
Avg decline: -32.5%
Time from first peak to dome: 4 to 25 months

Skeptics' Corner 
intelligent question of the day:

"Why would such a pattern conceivably exist?"

As technical analysts, we can so often fall into the trap of imbuing chart patterns with meaning where in truth little or none exists.  Indeed without a root in simple logic and economic or market fundamentals, this kind of navel-gazing can, if we're not careful, lead us into making big mistakes.

While it's rarely possible to prove any whys and wherefores conclusively, it's vital that we at least make an intelligent attempt.  So here's my stab at explaining why this pattern has proved so successful and persistent over time: 


As the bull market ranges into new sunlit uplands, buying pressure gradually wanes and sellers begin to sniff a top in the market.  During the period of the Three Peaks, short sellers build up large positions and when the third peak collapses and breaks below previous troughs, investors pounce on what they see in their charts - which is at that stage either a 'Head-and-Shoulders' pattern or a 'Triple Top', both of which are seen at major turning points. This encourages them to press their short positions strongly into the trough at point 10. 

What investors see in early 1968 is a triple top or head-and-shoulders pattern, inviting them to sell or pile in short.  When this fails, it sets the market up for a final surge of short-covering and speculative buying - a kind of mini-bubble - which finally bursts when the last fool has bought.  With no fundamental support, the market then succumbs to a spectacular decline.

At that point, however, something occurs to up-end the supply/demand balance.  It may simply be that a new group of investors step in who are looking for an opportunity to buy.  It may be that those negative events or economic fears which induced prices to break down in the first place have suddenly been alleviated.  Whatever, buyers quickly recover their confidence and sellers lose their nerve.  

The trend reverses sharply higher and shorts scramble to get out.  To do so, of course, they must buy and as prices rise, those who have missed out on the early part of the bull market finally seize their chance.  Panicking at the prospect of being left behind yet again, they rush to get in on the action.  

A self-reinforcing spiral of buying then begins, based not on fundamentals but mainly on late-stage, 'dumb-money' speculation and short-covering.  

Stock valuations have usually by this point in the cycle become over-stretched and economic expectations are unable to be met. Reality gradually begins to dawn. Selling by the 'smart money' has begun in earnest around the first storey (probably why it appears as an extended plateau) but late-to-the-party buyers step in once more, lifting the index to the dome.  

Eventually, once the last fool has bought, values begin their inevitable reversion to the mean.  With almost no buyers left at these prices, the market falls rapidly and speculative money rushes to sell, creating a decline which cascades all the way back to the start of the frenzy, and often beyond.


APRIL 2011: 

Now that we're thoroughly familiar with the shape and character of the pattern, let's take a look at those stock markets which are the cause of current concern. Armed with Lindsay's original schematic diagram - for your convenience here it is again - I invite you to compare, contrast and judge for yourself. 


as represented by ETF 
ishares MSCI Emerging Markets

As things stand at time of writing, point 20 seems to have been established in mid-March; however we may see one more sharp downward move which would force us to re-label and bring forward points 18 - 20.  This has implications for timing an exit - see section below.  Peak MACD momentum for the bull market occured way back in June 2010.


as represented by ETF
ishares MSCI Australia 


as represented by ETF
ishares MSCI Brazil 


as represented by ETF
ishares MSCI Hong Kong

If this analysis is correct, we have the ishares Emerging Markets ETF - the world's most popular EM investment vehicle, a weighted average of 25 countries' principal stock markets and key holding for financial institutions and millions of small investors - closing in on what appears to be a major peak. 

These countries represent pretty much the only sources of powerful, 'organic' growth left in the world's economy. If they start to struggle, one of the slim pillars upon which the developed world's two-year recovery rests will begin to crack.  

Australia also appears to be at an inflection point.  Oz's recent economic resilience has come largely through its pre-eminent role as supplier of raw materials to China. Brazil?  Another commodity-producing behemoth, whose immense mining and materials corporations are hugely dependent upon China's continued rapid expansion.  

So it's interesting that the other stark example of the Three Peaks pattern is China itself, as revealed by the MSCI Hong Kong index representing 85% of all issues on the Hong Kong stock exchange.

One is forced to wonder whether the epicentre of any upcoming economic earthquake may in fact be China itself.  If so, the implications for commodity prices, currencies and many other markets including of course our own, will be far-reaching.

A newly-built Chinese apartment building lies casually by the side of the road,
having toppled over.  Dang...guess they'll just have to put up another one.


So what kind of damage could we be looking at?  Well the Three Peaks formation concludes at point 28 - which is always at or below point 10, the trough of the third peak. This tells us in advance the minimum decline we're likely to see.  

Measuring the distance from current levels down to the lows of June 2010 produces the following minimum percentage declines one can reliably expect in each of the ETFs mentioned above, were they to peak today:

  • ishares MSCI Emerging Markets     -26%
  • ishares MSCI Australia                      -32%
  • ishares MSCI Brazil                            -24%
  • ishares MSCI Hong Kong                  -27%

Fortunately, one surprisingly accurate timing technique outlined below suggests that these markets may well experience a few weeks of further gains before the bull is done, making the potential wreckage even greater - and the opportunity for those looking to sell short even more enticing.


Let's say that you're considering taking profits from your emerging markets investments, or would like to trade this pattern as a speculative short position.  When should you make your move? 

In his paper, George Lindsay describes a method for predicting the timing of the ultimate top which involves measuring the time from the first peak to the third. When that is 6-10 months, he claimed, it takes approximately 7 months from point 14 to reach the dome. With the benefit of having been able to study several more examples of the pattern since 1970, I can't vouch for the accuracy of his method.  

But I did come across another interesting feature of the formation, which fits with standard technical charting practice and which does turn out to be an remarkably useful timing tool. 

I found that: 

  • the length of time it takes for the market to climb from point 14 to point 15 (ie. from the base to the first storey) is very often the same as the length of time it takes to climb from 20 to point 21 or 23 - ie. from the first storey to the dome.  

By simultaneously using a good momentum indicator, an investor then has a good chance of exiting close to the absolute top of the market.  

Here's how it works, in an example from 1987:

The two bouts of speculative excitement lasted almost exactly the same time - three months - and the final push coincided with a momentum divergance.

Once the base of the domed house had formed, prices broke upward in a burst of speculation between January and the end of March (yellow box).  At that point momentum waned and, as selling picked up, the first storey began to form.  

By the time prices broke up out of their range from point 20, the overall pattern would have become clear and an investor could have noted the length of time between points 14 and 15.  By adding three months to the date at point 20 (third week of May), he would have come up with a window around the third week of August in which to look for an exit.

As things turned out, all he would have had to do is sell in the third week of August in order to exit at the absolute peak of the market - a triumph of timing he would probably still be bragging about today.  To pin it down more objectively though, he could have used a common momentum indicator such as the MACD.

You'll have noticed from the charts that, in every single case, the very top of the market always featured a divergance between price and market momentum.  In fact:

  • in not one of the 16 instances did a peak in prices coincide with the bull market's peak momentum.

The 1987 example was the closest shave of them all, since peak momentum occured at point 21 during the second week of August.  A few days later prices pushed up to point 23, but momentum had stalled and begun to fall, creating a divergance and sell signal during the last week of the month.

Typically, divergances develop over a much longer period as the 1961 example shows.

The momentum peak for the bull market actually occured a year before and gradually fell away during the formation of the pattern.  A burst of buying pressure from the base at point 14 propelled the market for almost exactly 4 months, when selling began to pick up and momentum slowed.  

When another buying spree began after point 20, an investor projecting forward 4 months from mid-July would have looked for a signal from the MACD somewhere in the region of mid-November.  It gave him a perfect divergant sell signal on 29th November, within a whisker of the absolute top.

Of course it's not always that easy.  This mutant pattern is from 1976:

It turns out to be one of the trickiest examples to label. The run up from 14 began to slow after two months, when a rather sloppy consolidation from 15 begins.  Point 15 also marks peak momentum for the entire bull market, as measured by MACD.  From then on as prices struggle higher, divergances between price and momentum become obvious.  

A third thrust downwards into 20 marks an end to the consolidation phase, and prices push to a new high at 21. At that moment, our sharp-eyed investor would measure two months forward from the trough at point 20 and look to find an exit around early August.  The MACD would have given him a sell signal in mid-August and then another chance in late September.  

So this timing method is not exact, nor of course can it be guaranteed, but check each chart and you'll find that this simple measure does put us in the ballpark of the final high in virtually every case. 

And if it doesn't next time, you always have one final line of defense: the 200-day simple moving average.  Selling or shorting once prices break below that will still protect your portfolio from most of the ensuing damage and / or afford an excellent shorting opportunity.*

*If you're considering going short any of these markets, here is a link to the most comprehensive listing of emerging market ETFs, including short and ultrashort ETFs, I've been able to find.




Using the chart for the broad ishares Emerging Markets ETF (ticker EEM) as our template, we see a clear surge up from point 14 to point 15 which lasts two months, give or take a week.  The low of the consolidation phase at point 20 appears to have been set in mid-March - although it's possible we could see one more push down towards the 200-day moving average, requiring us to re-label the last couple of points in the pattern.  

But as things stand, projecting forward two months from mid-March, we should be looking for a peak in the Emerging Markets ETF, as well as in Australian, Brazilian and Hong Kong stock indicies, somewhere around mid-May.

Since there is ample opportunity to start with modest short positions and to add to them aggressively only if the market falls and confirms our thesis - and as long as one also uses fairly tight stop-losses - I would find it hard to imagine a more favourable risk/reward opportunity presenting itself in 2011.