Creating Your Own Trading Indicator (by TheInflationist)

Hi again fellow slopers. Thank you for your kind comments in our previous post (despite the poorly timed call of 11300 based on 1937 comparison with markets dropping 105 points the same day).

There are just so many indicators out there - most are widely available and are built into charting softwares and quoted by every technical blog. How would they work if every trader has in their disposal the same indicator to warn them when to long or short? Does it work simply by being a self-fulfilling prophecy where more people jump in the bandwagon - which in the course of time the overcrowding would lead to a game of musical chair (ie the last to bail when the music stops spots the bill). For an indicator to work, we believe it has to be a custom indicator not commonly known to all.

We have experimented on a few indicators (yes, its an experiment and work in progress) and are quite happy so far with the one we have. It is very much like devising a medical diagnostic test. Most of the readings of the test come in a spectrum. Let’s take the pregnancy kit for example. From a consumer (non medical) point of view, this test appears to be a all or nothing (positive or negative) test.

If you see a “green +” it means you are pregnant or a “red -” it means you are not pregnant. The truth is it is not that simple. There is a spectrum of “positiveness” and the manufacturers had to pick a cut off which hopes to include as close to 100% of true positive results without having too many false positive ones. Same with negative results. The lower the cut off in beta hCG levels, the more likely it is that the user is not pregnant.

Setting it too low though would risk false negative results, ie patients who are pregnant with beta hcg levels above the cut off selected by the manufacturer of the test. The pregnancy test is not the best example since it is a VERY accurate test. But most other diagnostic tests are NOT. Apologies for the long winded medical analogy - we cannot help applying medicine (our day job) in trading. So back to our indicator.

The “overvalued” or “undervalued” zone/cut off is arbitrary. One could pick any cut off. The more extreme, the more sensitive it is. Compromise and lower(or increase) your cut off and you start getting false results (ie false short signals and false long signals). And it is easy to change your bars to fit what you want it to fit for that period of time. We have been using this indicator on a more macro level. Below is a zoom in snapshot of what the indicator is saying now.

At the moment, our indicator is surprisingly flat lining. It is not telling us that markets are overvalued or undervalued despite the relentless rise. That does not mean markets cannot crash. For our readers who are statisticians or of medical and science background, its all about sensitivities and specificities. A highly sensitive test is great at ruling OUT (ie a negative result has a higher predictive value) something whilst a highly specific test is great at ruling it IN (ie a positive test is of higher predictive value). We are still working on our indicator, but so far so good.

Current Levels

Back testing of our indicator in 2001-2002: