Most of us are using technical indicators to generate buy & sell signals. And it’s a known fact that most people use more than a single indicator to accomplish that. The most common method is using a few technical indicators that generate buy & sell signals if they all meet a certain condition at the same time (for example: lines crossing/crossovers): and that occurs, most of the times, if all indicators used within a system point in the same direction (bullish or bearish).
As we all know, a trading system consisting of more than one technical indicators has its main purpose of filtering ‘false’ signals.
It’s widely accepted that technical analysis can be used to forecast the market. I use to call technical analysis a discipline to “forecast the past” (as funny & ironic as it sounds) but I won’t discuss its inherent attribute in this article, so I take it as it is. It is also widely accepted that using more than one indicator would increase the likelihood of filtering ‘false’ signals.
There are two kind of indicators:
1. indicators derived from price (correlated) and
2. indicators not derived from price (not correlated)
Let’s define a technical indicator:
A technical indicator is a result of mathematical calculations based on indications of price. Therefore, most indicators provided by a trading platform are derived from price.
But there are many trading systems consisting of more than one/several indicators – including those which are not mainly derived from price.
If you allow me, I’d like to call an “indicator” not only a technical calculation result but also a non-chart related factor, whatever that is, at least for strengthening the example purpose in this article.
So, a non-price derived indicator could be anything else people use to generate trading signals – be that a specific time of day, astrological event, economic data, coin tossing etc. (again, I prefer to provide some non-price related examples here to strengthen my argument)
But what are trading signals all about? Money? Trading performance? Knowledge? Skills? Ego? No – trading signals are all about probabilities.
And since they’re all about probabilities, we’ll use probability calculations to reach a conclusion.
Case I - using more than one or several indicators derived from price:
This is pretty straight forward: since all (used) indicators are basically calculations of (same) price, they’re all behaving in the same way. Which means: if price goes up -> indicator A goes up -> and indicator B goes up -> and indicator N+1 goes up -> which means they all “agree” on the market direction, so it’s a buy signal
So basically we’re using several indicators ‘saying’ the same thing, guiding us in the same direction, but we’re using a bunch of them believing they’re all adding value to the outcome.
The reasonable question is: Why not using only one? … since they’re all pointing toward the same outcome. It’s like using a single GPS device in your car instead of having several attached to your car dash.
Case II – using more than one/several indicators NOT derived from price
There’s a common mistake people do when using indicators not derived from price (not correlated). From what I’ve seen, people use to think that using more than a single indicator would increase the chances that the final result would be more accurate (profitable).
Let’s say you have a signal/trading system consisting of an indicator with an accuracy of 50% and another one with an accuracy of 30%. So, if you use these two indicators in your trading system, would that make it more successful than using only the first one? Would you use the second indicator to filter the ‘bad’ signals generated by the first one?
The main mistake in this case is that people believe that using these two indicators would be better than using only the first one. Why? Because traders use to calculate second indicator’s accuracy in addition to the first one’s. So, they believe that in the end, combining the two, the signal will be 80% accurate (50% + 30%).
That is totally false – because adding one to another is not the way to calculate probabilities. Non-price derived indicators are independent events. Two (or more) events are independent if the outcome of one does not affect the likelihood of other(s) event(s). So in this case, we’re dealing with independent events.
In order to calculate probabilities of independent events, we’re using a simple multiplication formula, not addition. We’re not adding probabilities but multiplying probabilities. So if we’re using two indicators, one having an accuracy of 50% and the second one having an accuracy of 30%, the result would be:
1 1 1
— x — = —
2 3 6
And there’s more: let’s say you have two indicators of 50% accuracy each:
1 1 1
— x — = —
2 2 4
In a nutshell: using multiple non-correlated indicators would drastically reduce the final accuracy.
So, using two indicators which are accurate half of the time would result in a system with 25% accuracy.
And the problem gets even worse if there are more than two indicators used in a system. Let’s look, for instance, at a system consisting of 3 indicators of 50% accuracy rate each:
1 1 1 1
— x — x — = —
2 2 2 8
See? combining 3 indicators which are accurate half of the time would result in a totally unprofitable trading system.
Not sure if I make myself clear enough. Let me repeat myself: people have a bad habit of adding more indicators to their system expecting to make better filters. So if they have a system which is accurate half of the time (and btw, that’s one hell of a great system) they will look to “tweak” it by adding more indicators expecting it to become even more accurate. Sorry, but simple math does not agree.
Case 3 – two or more indicators derived from price are turning into non-correlated (non-price derived) indicators (see II)
This is the tricky part. Not all indicators work in the same way. What matters the most is the time when the signal change / hence providing trade opportunities. But even if several indicators are derived from price, they do not react to price changes the same way, although within a specific time-frame they point toward the same direction.
Different, price-derived indicators do not provide signals at exact the same time, due to lag and settings related to previous-bars calculations. As a result, in the end one or more indicators will be even too fast or too slow and that will result in a non-correlated outcome, turning two/or more price-derived indicators into a system consisting of, basically, non-price derived indicators.
My conclusion is: even if you use price-derived/correlated indicators, if the trading system consists of mixing them either to ‘filter’ bad signals or ‘strengthen’ the system’s accuracy, there’s a good chance that you’ll end up using non-correlated (independent events) signals, therefore applying the probability multiplication formula which would drastically reduce the system’s accuracy.
Personally I prefer to keep my trading systems as simple as possible. Not mixing several indicators on a chart helps me to have a broad view of what’s going on. Also as a driver, I don’t need several GPS navigation systems on my car dash – one is enough.
Related reading: Figuring the Odds (Probability Puzzles)