Some time ago I started working with a couple of guys that wanted a strategy to enter a trade when we get MACD divergence off a Bollinger Band bounce. I annotated some charts to help clarify what they were looking for. See attached.
The trick is when to take the 1st peak, and when to start looking for a second but divergent peak in both data series. For example, in the MACD examples, after the first peak in the histogram, we decided to not start looking for a new but higher swing low in the MACD histogram unless the MACD unless we got at least 3 higher readings after the first peak. Some thought we should cross all the way above the zero line and then back below it before waiting for the 2nd swing lower,while others thought that a cross above the zero line should reset the search for a 2nd peak. Also, regarding price, it wasn't clear if we are comparing lower lows or lower closes.
So there is a lot of discretion in measuring divergence and how you read the 2 data series that you are comparing. These examples that I share are some of the challenges one has to consider.
The following user says Thank You to shodson for this post:
Most trading systems based on divergences have a negative expectancy. The left chart is a good example. MACD produced a divergence, not because the trend slowed down, but because of the Baskerville effect: a moving average also barks if something drops out at the beginning of the period.
By pure luck price entered a consolidation and then continued its prior trend. Every short term consolidation of a trend produces a divergence, so you would not want to trade all of them.
The right chart shows that a trade entry based on MACD divergence came way too early. What about adding a filter that you will only take divergences after a trendline break ?
If you have a strong trend, this will be typically near a Bollinger Band. I have often observed that a trend weakens - i.e. reduces its slope - several times before it reverses. See daily chart of ES below.
Thank you for the comments. The chart also shows that markets are assymetric, fear and greed do not create the same feedback patterns. In the stock market topping patterns take longer time than bottoming patterns. It is not unusual to see double or triple tops, but rarely you will notice a double bottom with low volatility.
I agree that MACD signals work better with a zero line cross and when volatility is high, so it might give you an edge for stock index bottoms or commodity tops. These are U-class peaks and troughs and not similar to the moderate S-class peaks and troughs.
So to apply a MACD divergence indicator, you might first need to qualify the type of market. Apply it to a high volume climax move and it might work, apply it to a wedge, and you will get a quadruple divergence, which is a failed triple and failed double divergence.
The problem of the divergence is that it only measures the peak (trough) behaviour, but not the trendline side of the price action. If you have an ascending triangle, on the peakside it creates a divergence, on the trendline side it is a continuation pattern. How would you qualify this?
if you transfer Richard's post as well, I could answer it here....
Below is a chart of ES during the last four hours. The MACD produced numerous divergences, as ES made four trendline breaks, while gradually moving higher. A divergence is just a slow down of price. If you want to know whether it is a valid pattern to initiate a countertrade, the better way is to scrap the indicator and just watch price and volume.
The chart also shows an indicator that serves as a trend filter. Green means up, so you would be very careful to enter short trades.
The following user says Thank You to Fat Tails for this post: