Let me start by saying that I do attempt to connect peaks and valleys on the oscillator when drawing a trend line for divergence. This is the text book, picture perfect way of doing it.
This is what we want to see…
But the market is often messy and rarely picture perfect, so I am not opposed to drawing the trend line right through the oscillator in order to precisely connect the values that correspond with the peaks and valleys of price.
This is what we often get and often fail to see…
My logic for this is:
1. Price is infinitely more important than an indicator, so the points that are plotted are based on price and the oscillator has to accept those points where ever they land on the oscillator.
2. In plotting divergence, we are looking for the differential between the two values on the oscillator that correspond with two specific points in price. I’m not using oscillator action as I would use price action, nor am I using oscillator structure as I would use market structure.
3. By allowing the oscillator’s line to be drawn through, you find divergence that might otherwise be missed if strictly using peaks and valleys of the oscillator that match peaks and valleys of price.
This is just my opinion based on my observations.
If you use divergence as a confluence in your trading, I’m interested in hearing how you draw divergence and your thoughts on this.
(The oscillator pictured is Stochastic RSI with only the %K line enabled)