What really is divergence?

You can tell me that I needn't know how an engine works to drive my car. True enough. Does this forbid me to be curious?
Divergence is this kind of curious thing I'd like to understand.

You can find zillions of sites that explain you how to spot the four types of divergences. Never more than a copy'n'paste of each other. A lot of web sites and youtube videos show you how to trade the divergence. Each one adds his own pinch of salt to the recipe with no explanations.

I searched the web up and down I could find absolutely not a single site, document or video that explains what produces (or is supposed to produce) the divergence phenomenon.

There are traders using the divergence --sucessfully or not-- but I feel like none of them understand what they really do beyond waiting the phenomenon to happen. They argue for which is the perferct oscillator to use but how can they evaluate the best oscillator if they can't explain what an oscillator is supposed to do to produce the divergence?

At first I though the divergence was some mean to measure the acceleration of the price because the difference between two points on a momentum oscillator is an approxiamtion of an acceleration. But it seems that the volatility also plays some role. E.g. a congestion gives some time to the indicator to catch up its lag.

My questions are
- What does produce the divergence between the price and the oscillator?
- Why is a regular divergence supposed to announce a change of the trend?
- Why is a hidden divergence supposed to announce a continuation of the trend beyond the simple fact that a trend tends by itself to continue?
- Are the regular divergence and the hidden divergence the same or two completely different things?
- Does the absence of divergence carry some information?
- Can we use these answers to measure or estimate the divergence effect continuously and adaptatively instead of only at the peaks and valleys with assumed universal parameters?

Just to be clear:
1- I needn't you to cut and paste a upteenth time how to spot a divergence.
2- I can't care less of a screenshot that shows how well a divergence worked in the past.
3- I can't care less of a screenshot that shows how a divergence failed in the past.
4- I don't give one of which oscillator you prefer to spot divergence because without a proper explanation of the phenomenon that's just a value judgement.
5- I'm not interested to know if you believe it works or not because without a proper explanation of the phenomenon that's just a value judgement.
6- I don't care you're successful trading divergence.
7- I don't care you failed at trying and based on that experience you decided it doesn't work.
8- I know any causal filter lags. That's not the point of this thread.

Here is an intelligent answer to the regular divergence question, by gspajon.
Your structured approach to try to decipher this is very interesting. It would be nice for example to simulate (or use the actual) data and see how different (with respect to their period), say the simplest - "stochastic" oscillators, behave around the turning points. I guess that's what Spud's MTF stochastics thingy was about...

k

Hi

- Why is a regular divergence supposed to announce a change of the trend?
When the correlation of the oscillator with price changes degrades divergence occur.So lot of traders believe in that the correlation coeff of the oscillator/ price changes
will revert to 1.
As far as I know all the divergence indicators are lagging because they try to spot the peaks or valleys.Maybe some kind of pip difference or percent change difference between oscillator/ price changes has some potential,I dont' know.
Just a thought maybe mutual information of oscillator/ price changes may give some interesting results.

[quote=PipMeUp;8656473]You can tell me that I needn't know how an engine works to drive my car. True enough. Does this forbid me to be curious? Divergence is this kind of curious thing I'd like to understand. You can find zillions of sites that explain you how to spot the four types of divergences. Never more than a copy'n'paste of each other. A lot of web sites and youtube videos show you how to trade the divergence.


I think divergence shouldn't be calculated or pinpointed with the suggested methods in forexfactory. People generally try to figure out divergence between price and the oscillator. In my belief it is comparing apples and oranges.

Price continues to evolve freely but any oscillator you put has a limited length that the trader selects (it forgets the past history).

If divergence should work, I believe it should be calculated with oscillator and a moving average of sorts, both having the same length or lookback.

Trading isn't like driving a car - designing and trading a profitable system is more like, building a car yourself and then driving it across the country. Yeah, you kind of do need to know how the engine works. And to design and trade a profitable system, you kind of need to know how every component of that system works.

If you can't pick up a pencil and calculate and draw your own technical indicator, you have no business using that indicator on your charts - because if you don't know how it is calculated you don't really know what it is telling you, therefore it doesn't really add any meaningful information to your trading. If you can pick up a pencil and draw your own stochastic, or RSI, or MACD indicator, then you understand what the indicator is telling you, and you will understand what divergence really means.

There is nothing secret or magical about the popular divergence concept.
It's just a way to detect and to measure changes in momentum.

Divergence = slowdown in momentum (potential)
Hidden divergence = overextended trend correction.

When you measure the distance between moving average and the price you can easily spot changes in momentum.
New price extreme+ smaller distance between price and MA compared to the previous extreme = divergence. SIMPLE!
For the hidden divergence is exactly the same but you measure it backwards
In uptrend for example you measure the distance between the lows and MA.

Another way to search for divergence is to measure the average length of the waves.

Also keep in mind that indicators and oscillators are not necessary. You can spot the divergences even in a naked chart. (when you know what to look for).

hello,
how to understand the logic of indicators ?

By understanding where the formula is coming from. What is computed. For instance RSI estimates where the price is in the horizontal range of the N(=14) previous bars. Knowing this makes you immediately understand that the RSI expects a ranging market. In a trend it is shifted by a value proportional to the trend. Now you can correct it: rsi14 - sma14(rsi14) + 50 gives you a detrended RSI. You can also see that RSI stretches the price range to the range 0 -> 100. When the range is very narrow RSI becomes very choppy. This happens when the price runs quickly with no wiggles. With the same kind of logic you can reverse engineer the MACD and show that the histogram is a measure of the trend up to a multiplicative constant which depends on the period you choose. You can take Bollinger bands and show how wrong the computation is. etc. But I could not make the divergence appear in the formulas.

Thanks to the people who answered.

I like very much gspajon's explanation. It makes sense in a DSP point of view: if the price rises quickly the oscillator (which is a bandpass filter) will attenuate it more and we get a lower high when price makes a higher high. We can also expect that if the price rises normally but previously rose slowly we'll also get a divergence. Yet in this case it's not an exhaution rally but rather a rally after an accumulation. Fading this move is much dangerous.

I think divergence is merely a by-product of the way an indicator is constructed.

P. 66 : "As the RSI moves lower, it is encountering increasing methematical resistance. Remember that the RSI is a logaritmic line and for it to move lower when it is below the 30 level is much more difficult than when it was at 50." (RSI The Complete Guide by John Hayden)

So, an uptrend will exhibit bearish divergence and hidden bullish divergence.

Perhaps you can find more information in the book I referenced.


Best,

This is how I understand it.