Case Study: Do Indicators Really Work?

Case Study: Do Indicators Really Work?

June 13th, 2018

Full disclosure: I love indicators.

I love that, years ago, a smart person saw something interesting in the markets and used smart-person math to numerically give us all a way to make money.

There are ways to make money trading without an indicator, of course, but our minds are so undisciplined that most of the time we sabotage our own chances before we ever bank the profits.

Indicators fix all that.

They give us a definite mathematical entry and a definite mathematical exit. They give us logical reasons to make profitable trades and their calculations provide the discipline to execute properly.

Just stand on the shoulders of mathematical geniuses and follow the indicators to the promised land.

I love that.

But indicators are taking a lot of heat lately.

They’re lagging.

They’re based on math that’s outdated.

They’re a crutch only used by amateur traders. 

Is this true? Is it pointless to use indicators?

To answer, we’re going to need some rules that make perfect sense. We’re going to need rules that are so simple that the conclusion we draw is darn near inescapable.

Otherwise arguing about indicators is just a bunch of sound and fury signifying nothing.

First, we need to use an instrument that’s been around a while. Putting an indicator on bitcoin futures, for example, is not going to help us. So what’s been around the longest?

The stock market.

But we can’t pick individual stocks. That would be too biased. It could be argued that an indicator happens to work best on a particular kind of stock.

No, we need the whole stock market. That incorporates a bunch of stocks and can give us a good picture.

Second, let’s use the futures market, specifically @ES. I suppose we could also use something like SPY, but I’ll take the leverage of @ES instead.

Third, let’s only go long. The stock market has an upward bias, so let’s use that and also stay very specific. The more variables we use, the less valid our conclusions. Plus, the long side of the market is very different than the short side, so we need to run those experiments separately.

Last, what is the definition of “working”? If we say an indicator “works” what does that mean?

It means that if an indicator generates a Long signal, the market will go up. If we get a buy signal today, we should then be able to walk away and come back later and see a trade that’s made money.

If we put a stoploss in, we’re tampering. We’re saying that the indicator probably doesn’t work that great, so let’s put in a safeguard. That’s a shaky signal.

If we put a target in, we’re also tampering. We’re saying that the signal doesn’t have much strength, so we better get out before anything bad happens.

If we’re truly trying to find out if indicators work, then we need to keep it strictly black and white.

Get a signal, take the signal, and come back later and see profit.

The question then becomes: how long should we wait? Should that signal last years?

I don’t think that’s a good test. The indicator was based on math that was intended to foreshadow an imminent change in our favor. I don’t think any indicator was made to predict where we’ll be five years from now.

So how about three weeks? One week seems too short for this test. A month seems slightly too long. This is debatable, of course, but three weeks seems reasonable. It gives the indicator time to work while not forcing it to do too much.

Three weeks it is. If we find anything exciting, we can change this to a different time span and test some more.

But that’s not all we need. If we’re going to believe in indicators, that indicator needs to give us signals that, if followed, beat the market. These signals should help us do better than if we just got in at the beginning and held for the next several years (or decades).

If our famous indicator makers did their jobs, then those signals should do better than a trader who had no indicator at all.

Finally, we need an indicator. I’m biased toward RSI and CCI, but the very first indicator I ever looked at was Stochastic, so that’s where we’ll start. And, to avoid any funny business, we’ll just use the default settings given to us by George Lane himself (or C. Ralph Dystant).

Those settings have a Length of 14 and an Oversold reading of 20.

Rules:

  1. Using a Slow Stochastic with a Length of 14, we’re going to wait for price on the ES to go below 20 on a Daily Chart.
  2. Then we’re going to wait for the Fast stochastic line to cross over and close above the Slow stochastic line.
  3. Then we’re going to hold the trade for 15 days (3 trading weeks) and exit at the open of the next bar. We won’t count the entry bar. So that’s 15 bars after the entry and then exit at the open of the 16th bar.

Here’s a picture of a winning trade from March 28, 2018: https://www.screencast.com/t/vLXYmS72x

What were the overall results?

Starting with a $20,000 (and including all trading costs such as slippage and commission), using a Stochastic indicator generated a profit of $56,039 and a close-to-close max drawdown of $15,091. The testing period was 1998 to June, 2018. The test used a trade size of 1 contract on every trade (no compounding).

According to the Tradestation Performance Report, that’s a return on initial capital of 280%.

If we simply bought and held 1 contract of the ES during that same time period, we would only have achieved a return of 122%.

By the way, the max drawdown if we bought and held the ES would’ve been $55,625, which is more than 3 times as much as Stochastic’s was and would have taken our whole account and shoved it down the toilet.

The conclusion seems pretty clear: Indicators work. They provide a substantial edge over buy-and-hold.

Well, at least Stochastic does.

But does it work only under those conditions?

We’ll find out in Thursdays YouTube video.