Divergence Theory

The Concept of Divergence
How marvelous is a tool that lifts the skirt of the life blood of the markets, greed. Greed propels the markets to extremes. It is the idea that the folks that reaped the reward of driving the market in one direction, can’t resist going to the well one more time. Ultimately, they realize that the game they have so cleverly played is about to end so they bail out and head in the other direction. The final act of pushing the market to extremes frequently provides telltale signs of weakness, sensed so easily by the floor traders. The final move is often a volatile event without the strength and conviction of previous surges. Tools can be written to evaluate the nature of these surges and objectively quantify the strength of the price moves. This is frequently done by using an oscillator like RSI or Stochastics in conjunction with the price. However, price of one commodity can also be compared with another price if they are expected to be correlated.

The tool used to discover market weakness is called divergence. Divergence usually requires price to be moving into a strongly overbought or oversold region. We then look to see when the price previously displayed similar strong movement. Ordinarily we require the price to move to an even more extreme level than the previous peak. Then we look at the oscillator peaks. If the oscillator fails to follow prices to new extremes we say the price diverges from the oscillator.

Nothing works every time and the same can be said about divergence. Divergence provides no certainty that prices will reverse direction. Divergence can be followed by even more extreme prices. Likewise, not every "obvious divergence" can be detected by the divergence tool. It detects that which it is programmed to detect. Programming to avoid false divergence signals will inevitably lead to screening out otherwise valid divergence signals. This is all part of the game.

A Few Words About Divergence
Divergence seems to be one of those things that everyone seems to intuitively understand but are unable to express objectively. How often have you heard something like, "…and then, when you see an obvious divergence…?" I am frequently approached by well meaning clients who ask me to program something they learned at a seminar or from a book where divergence is one of the required ingredients in a sure fire trading method. All I need to do is to write a piece of code to detect the "obvious divergence" and the methodology is ready to be fully baked.

Unfortunately, spotting divergence by eye is easier than designing an objective tool to detect it. That is because of all the exceptional cases the viewer should recognize as divergence but for one reason or other the eye fails to identify. If you want to systematize your trading then you need objective rules. When you begin to be objective, many of the subtleties pose difficult programming problems. Let’s look at a few of the issues that must be addressed.

Peak Separation: The concept of divergence begins with the thought that there are two clearly identifiable peaks in price. For instance, movement in the direction of the uptrend appears to be confirmed because the most recent peak in price is higher than the prior peak. The problem is to define what minimum or maximum number of bars between peaks is allowed. For instance if price has risen dramatically for 10 bars and then drops lower for one bar before continuing higher, can we say that there are two peaks in price? Most traders would simply say that the lower bar was simply a momentary pause in the price movement that comprises the same peak. They would not split the move into two peaks. That begs the question, what is the minimum number of bars between peaks?

Move Termination: Not only do we need some minimum number of bars between peaks, we need to define what event officially terminates the price move. If price has been moving higher, we usually want to see the price fall back to confirm that the upward thrust has potentially completed. Otherwise, divergence might be detected while prices continue to set new highs, thereby fostering dangerous top picking. So, what defines the end of a price move?

Peak Strength: Some peaks are weak and others are strong. Our eye is quick to connect the most recent peak with some "obvious" previous peak, effortlessly ignoring the minor peaks in between. How do we tell the computer which peaks to ignore and which to count?

Asynchronicity: The eye is also instantly capable of divining divergence from price and an oscillator without hesitating to recognize or provide import to the fact that the peaks in price did not occur on the same bars as the oscillator peaks. The peaks are not synchronized. How do I tell my computer about this little problem? How far back or forward do we search for the oscillator peak if we know the bar on which the price peaked? What event will terminate my search for the oscillator peak?

This is not offered to convince you how difficult the programming effort was. Moreover, we hope to provide you with enough information so that you won’t fall into the trap of ever using the words "obvious divergence" in connection with trading. If you are looking for objectivity, there is no such thing as "obvious divergence." Objectivity requires rules and whatever meets those rules qualifies as a valid divergence, until we change the rules.