Divergences are everywhere if you listen to the analysts on CNBC. There’s almost always a chart that shows two lines moving in different directions and viewers are told this is important. Because it’s CNBC, it seems that most of these divergences are bearish, but that might be because it seems most of the analysts on CNBC are bearish. Divergences are an important topic in technical analysis but the question that hasn’t ever really been answered, until now, is whether or not divergences are useful. Before addressing whether or not divergences are useful, we should be clear about what a divergence is.
Divergence analysis can be applied to any indicator, including RSI, stochastics or breadth indicators like the advance-decline line. For the sake of clarity, I’ll use MACD for my analysis in this article so that we have a consistent frame of reference.
MACD, the moving average convergence/divergence indicator, is intended to measure the momentum of the price action. If prices are rising, we would expect momentum to be moving up and we would also expect momentum to have a high value. To understand this idea, let’s start by thinking of the momentum of a car.
The car’s speed is the velocity. The car’s momentum, in physics, is equal to its weight times the velocity. There are two important points related to this calculation:
- This would mean if we have two cars moving at the same velocity, the heavier one will have more momentum. The one with more momentum will be harder to stop than the other one.
- As the car slows, its velocity will be reduced and that means momentum will slow, unless there is a sudden crash. This is because momentum equals mass times velocity and the car’s mass or weight is fixed. The only way momentum can change is if the speed changes in this case.
Back to MACD which is a measure of momentum, higher values of the indicator should indicate the trend is likely to continue. If we have two stocks where the trend is up, the trend is stronger, by the definition of MACD, for the one with the higher MACD. This indicates buying when momentum is high should be a winning strategy and numerous studies have shown high momentum stocks tend to outperform low momentum stocks over the next three to twelve months.
While momentum can help us spot trends, it’s also widely used to forecast the end of a trend. Technical analysts like to say “momentum precedes price” because they believe a slowdown in momentum usually occurs before the uptrend ends.
This is based on the second point above. The stock’s “mass” obviously doesn’t exist but if it did it would have to be the same no matter how fast the stock was moving. Because there is no way for mass to change, momentum can only change when the velocity of the stock price changes. Therefore, a change in momentum indicates the speed of the price changes is slowing, in other words the trend is weakening as momentum slows.
This all shows up as a divergence on the chart. In this case – price is rising so the trend is up but momentum is slowing so we see MACD falling. That’s a divergence.
A review of some charts might help. The chart below shows MACD and price. I added MACD to the price pane to make it all more clear. This is normal price action with MACD confirming the price action. Notice how MACD is rising in an uptrend and falling in a downtrend, that is normal and means prices and MACD are both moving in the same direction.
The next chart shows a divergence. We can see a large amount of normal price action with MACD and prices both rising and falling at the same time. As the uptrend ends, we see that price set a new high, by a few cents, but MACD did not reach a new high. This is the definition of a bearish divergence. The divergence is bearish because momentum leads price and in this case the momentum indicator, MACD, is bearish. We expect price to follow and turn bearish and in this case it did.
The next chart moves the MACD under the price action. This is the way you would normally see the indicator.
Momentum leads price so when a divergence develops between momentum and price, we expect the price trend to reverse as it did in this example.
A bullish divergence occurs when the indicator turns up while prices are moving down. That’s in the next chart. Notice price in early February came down to the same level as the Jan low while MACD was moving higher. Momentum leads price and now momentum is up while prices are down, forming a bullish divergence which should be followed by higher prices.
Here’s the traditional view of a bullish divergence.
Highlighting the divergences on these charts allows for what statisticians call the “well-selected example,” a technique that allows the presenter to prove their point with a few examples. Of course, those examples are selected solely because they prove the desired point, they are well-selected. They might not be, and probably aren’t, representative of what typically happens. To truly discover whether or not divergences are useful, we need to run a backtest.
Backtests use preprogrammed rules to take all trading signals rather than just the ones a trader notices on charts. Rules mean that signals are objectively identified and the test results tell us whether or not the signals beat the market. In the case of divergences, the results of a backtest are disappointing.
For a comprehensive test, we can use the SPDR S&P 500 ETF (NYSE: SPY) and record the results of all divergences over the ten-year period ending in April 2016. To obtain objective results, we will simply exit one-month after the divergence develops. Using this strategy, we can test for divergences that develop over various lengths of time. For example, we can look at how useful a divergence lasting one day is and step through the length of time required for the divergence to form. This test looks at all divergences lasting from one day to one month.
Looking back over the past ten years, we know SPY has shown a general upward bias. In fact, if you bought SPY on any random day and sold one month later, you have recorded a gain about 63.3% of the time. That figure provides a baseline we can compare divergences to. If divergences increase the winning percentage, they would be useful. If we win less often after spotting divergences, then divergences should be considered useless or even harmful. If we see about the same number of winning trades, divergences can be thought of as random noise in the markets.
If you bought after a bullish divergence developed and sold one month later, you would see a gain 62.6% of the time. A bearish divergence resulted in a gain 62.8% of the time. These two figures are not really different from our baseline and lead to the conclusion that divergences are simply market noise that can distract traders from their primary goal of making money.
You might be thinking the test results are wrong because you have seen a number of charts showing divergences that developed before market crashes or close to important bottoms. We have all seen those charts and they illustrate the problem of the “well-selected example.” Seeing a particular divergence worked once could provide an incentive to hunt for that divergence again. In doing so, you could be ignoring more important chart patterns.
Even though testing shows divergences will not improve your trading results, divergences do serve an important purpose. Divergences allow CNBC and other business news networks to fill hours a day with colorful guests, claiming they know what the future holds based on the latest chart they saw. Of course you know, that you should be watching CNBC for entertainment purposes only, ignoring divergences unless the speaker provides backtested results, showing their observation improves your odds of beating the market.