Traders use momentum indicators like moving average convergence/divergence (MACD) to confirm price trends. If both prices and momentum are rising, then they expect momentum to keep pushing prices higher. When momentum slows, they watch for a reversal in the price trend.

A bearish divergence is seen on a chart when prices reach a new short-term high but a momentum indicator fails to reach a new high. Traders tend to believe that momentum will change directions before price, and if momentum weakens, then price weakness should soon follow. An example is shown in the chart below.

In February 2012, the price of SPDR S&P 500 (NYSE: SPY) continued higher after the MACD indicator, shown at the bottom of the chart, peaked. In April 2012, prices peaked and subsequently fell more than 10%. Traders who spotted the divergence and sold would have avoided the loss.#-ad_banner-#

That is usually how a discussion on divergences is presented. The idea is explained and a single well-selected example is offered as proof. In my opinion, traders should prefer to see data rather than single examples before they accept any idea, even one that is as widely accepted as the importance of divergences.

In reality, programming divergences can be a challenge, and in my experience, the more challenging programming results are done by large firms and rarely shared with the public. I’m trying to change that and show readers what really works in the market.

We can test divergences using the MACD indicator and S&P 500 index data back to 1928. The advantage of using MACD is that it is almost always calculated with standard parameters so most traders will accept the results.

There will be only one variable in the test period — the amount of time for the divergence to form. The divergence in February 2012 formed over six weeks. For testing, we will look at divergences that formed over time periods as short as one week and as long as 26 weeks.

The first test is summarized in the next chart. There were more than 700 divergences lasting at least one week in the test period of 85 years, an average of 8.5 a year. Buying each divergence and selling after a prescribed time would have been slightly profitable if the positions were held for less than 13 weeks. None of the profits were significant when risk is considered, but the conclusion of the test is clear: The divergence lost significance as time passed.

While profitable over short time frames, less than half the trades were winners in any time frame. And very short divergences do not offer a significant edge to traders if we define an edge as the ability to make consistent profits.

Fans of divergence analysis will point out that the divergence should build up over several weeks in order to be significant. The next chart shows the impact of varying the length of time the divergence takes to form, from one week to 26 weeks. There is some profitability for divergences formed in one week and then another spike of profitability between 21 and 24 weeks.

This chart shows that divergences don’t offer any more information than a random coin toss would.

Similar results are obtained with daily or monthly data. Bearish divergences between MACD and price do not offer tradable insights. Thousands of tests with other indicators and different parameters also showed that divergences are not tradable.

Action to Take –> Despite the test results, some analysts will still point to divergences as a warning of a decline. I’ll continue testing and will let you know if I ever find a tradable divergence, but for now, the evidence says we should ignore divergences as a warning of market declines.

This article originally appeared on ProfitableTrading.com:
Bearish Divergence: This Widely Accepted ‘Sell’ Signal is a Fallacy