Tuesday, May 08, 2007

Comparing Apples and Onions

Apples are different from Onions. But both have a ball form, so why can't we compare them? Well every child will tell you, you can't, they are different. Still in TA we all have done it at times. Place two oscillators within one Window and see the Holy Grail. The crosses of these two oscillators look great, let's trade them

Here you see the TRIX oscillator against the Ergodic within 1 subwindow

That's not too bad on a first look. Unfortunately these 2 Oscillators have nothing in common. One is an apple, the other an onion. To compare 2 Oscillators, the least you need to do is making sure that both oscillators have the same range.

EG: Both have swings from 0 to 100 

But the TRIX and the Ergodic are calculated from price and are not normalized. Meaning the value range these 2 Oscillators swing in, is not fixed.

But there is a way to compare Apples and Onions! Just make them comparable. And the tool to do that is the Stochastic Oscillator. The Stochastic Oscillator normally uses the Close Price as input and it tells you where that close price is relative to the range of the last # bars.

EG: A Stochastic 14,1,1 takes the range of the last 14 bars and tells you where the last price is within that range on a percentage scale from 0 to 100% (0% being the low and 100% being the high of that range)

Now we know that any oscillator swings. Even when prices trend an oscillator tends to go flat. What we don't know with not normalized oscillators like the Ergodic or the Trix oscillator is the range these oscillators swing in. Still there is a way to change that. Ensign Software allows me to plot a Stochastic Indicator which uses not the Close Price but the Ergodic Oscillator Value as Input.

Use a very long range (as seen above I use a 400 bar range) and you will cover the relevant range the Ergodic moved in for the last few hours or days depending on your timeframe. The Stochastic will still be very sensitive, as the return value is the exact percentage value, where the current Ergodic value is within the range of the last 400 bars. Do the same with the Trix value and you get something very similar but not identical to the chart I showed above.

As you can see, while the arbitrary placing of the 2 oscillators within 1 window gives you a cross of both lines on the last bar, the placement within a fixed range of 0 to 100 and the correct placement of each oscillator within it's own range shows you no cross to the downside took place on the lbar prior to the last bar.

I can't tell you, if it makes any sense at all to trade crosses of the Ergodic with the Trix oscillator or slings of one oscillator against the other. The reason I wrote this article was to show you (and myself) a way to make apples comparable with onions, by placing them both within a common frame and then doing the comparison.


adam said...

I presume you can take one idnicator from the other and make a histogram, whcih can then be used to develop various trading signals, or are you going to trade with the bollingers, ie when both oscialotrs are up and the top band is touched go long, if one of the oscialitrs goes up and the other down (range) and the top bollinger is touched then short (for a range trade).

Your ideas?

Globetrader said...

Actually I'm looking into adaptations of this rule by using my favorite indicator, which is the CCI and not the STO. In the article I referred to the STO as it is wider used and has a clear range, while the CCI just places 95% of it's results in the +/-200 range. Instead of using a CCI with a longer bar length I'm using a shorter CCI based on a different input price than the one usually used. This gives me something like a floating zeroline on the CCI. I will write about this adaptation, if the method has merrits.

Dave said...

Chris, can you post your ensign template(s) for those?

Dave said...

Well...for that matter, would you be open to posting any of your templates you use or have used?


mg said...

Instead to comparing two normalized oscillators, you can just compare their unnormalized versions. The unnormalized version for CCI is uCCI=price - SMA(period), for STO it is uSTOC=price - L(period). By comparing their unnormalized versions with any kind of normalization method you like, double normalization is avoided.