Defining the McClellan Oscillator - Part III

This is a multi-part series poking around the edges of the McClellan Oscillator (MO), a breadth indicator using the number of advancing and declining issues to measure the degree of participation in market moves (read more).

In my previous post, I looked at rising/falling and high/low MO readings and showed that what they’ve said about the markets has evolved over time, much the same way that similar indicators based purely on price have.

In this post I want to look at how the market (specifically the S&P 500) has reacted to extreme MO readings. We’ll look at three variations. I don’t purport that they are the best, or that they are what McClellan originally intended, but they are similar in spirit to how I usually see investors using the indicator. (Click to enlarge:)

[logarithmically-scaled]

A quick note before we start. Recall the graph above from the previous post showing the result of trading the S&P 500 index frictionless from 1950 based on whether the MO closed above (green) or below (red) zero.

Recall too that somewhere in the 1990’s, the relationship flipped from being momentum driven (i.e. positive MO readings led to positive gains) to being not predictive or even a bit contrarian. I want this post to be about what’s working now, so all of these tests will only cover from 2000 (to 04/17/2009).

Variation #1: Long when the MO is Low, Short when the MO is High

(Click to enlarge:)

[logarithmically-scaled]

The graph above shows the result of going long the S&P 500 index at Thursday’s close if Thursday’s MO reading closed below -50 and short at Thursday’s close if Thursday’s MO reading closed above 50 in red, versus buy and hold (blue) from 2000.

Note: all of these results are frictionless (i.e. do not account for transaction costs or slippage), but for all intents and purposes, could be duplicated using mutual funds designed for active trading.

After I show the three variations, I’ll provide some stats and my own thoughts.

Variation #2: Long when the MO is Low, Short when the MO is High, Hold Until MO Recovers Enough

(Click to enlarge:)

[logarithmically-scaled]

The variation above is long at Thursday’s close if Thursday’s MO reading closed below -50 and holding until it crosses above -40, and short at Thursday’s close if Thursday’s MO reading closed above 50 and holding until it crosses below 40.

Variation #3: Long when the MO is Low but Rising, Short when the MO is High but Falling, Hold Until MO Recovers Enough

(Click to enlarge:)

[logarithmically-scaled]

And finally, this variation is long the S&P 500 index at Thursday’s close if Thursday or Wednesday’s MO reading closed below -50, but rose Thursday and held until it crosses above -40; and short at Thursday’s close if Thursday’s MO reading closed above 50 but fell and is held until it crosses below 40.

Summary Stats

Thoughts

Does the McClellan Oscillator (all by itself) do a reasonable job in today’s market identifying when the market is extremely overbought/oversold? Sure.

But I think you would be hard pressed to say that these results are better than (or even as good as) other similar OB/OS indicators based purely on price such as RSI(2) (of course, I am very open to being proven wrong).

Below I show why the MO doesn’t tell us much beyond what price already tells us, and in the post after that I’ll try to put my own spin on the indicator and turn it into something a bit more useful.

P.S. I know that some MO-proponents out there are saying, “yes michael, but it’s about the divergences between the MO and price, not just the MO”…

Does the MO give us info beyond price indicators?

It's going to seem like I’m picking on the McClellan Oscillator (MO), but I’m really not – I’m trying to make a bigger point about technical indicators in general.

This week we’ve been looking at the MO indicator and what it says about the broader market (read part one and two).). So far, the MO has said a lot, but not much beyond what simpler indicators based purely on price have.

In this humble developer’s opinion, the reason is that, despite the fact that the MO is based on advancing and declining issues (rather than price like most indicators), the end result is more or less the same.

To illustrate, first we need to do a quick tutorial on calculating the MO (we’re using the NYSE ratio-adjusted version). To begin, calculate the ratio of “net advancing issues” each day [(advancing – declining) / (advancing + declining) * 1000]. Next, calculate a 19-day and 39-day EMA of that ratio. And finally, subtract the 39-day EMA from the 19-day EMA to arrive at the daily indicator value.

Getting past all the geekery, the core thing being measured by the formula is “net advancing issues”, or advancers versus decliners. And as you might expect, that number is highly correlated to market price (bullish markets tend to have more advancers than do bearish ones).

From 1966 to the present, the correlation between that daily “net advancing issues” and the plain ol’ daily % change of the NYSE Composite Index (Yahoo ticker ^NYA) has been +83.7%. (For the non-statistically minded, that’s pretty darn high).

When we add the additional smoothing of the moving averages, correlation increases even further. If we create a MO-like oscillator of daily % changes on the NYSE Composite Index (by subtracting a 19-day EMA from a 39-day EMA), the resulting oscillator has a +87.0% daily correlation to the McClellan Oscillator.

Click to enlarge:

The graph above shows the two together, the MO and our MO-like daily % change oscillator, year to date…pretty darn similar.

The Point

The point of all of this isn’t to pick on the McClellan Oscillator. At least it’s making an attempt at using a different source of data (advancers vs decliners).

But my lack of enthusiasm for the MO is akin to my lack of enthusiasm for the VIX. As has been demonstrated ad infinitum here and elsewhere in the blogosphere, like the MO, the VIX is almost totally driven by information we already know and is already reflected by the price.

Treating it either as if it was some unique data point with some unique insight into the market is, in the humble developer’s opinion, a bit silly.

A Word about Divergences

The response I expect to the above diatribe is that the power of these indicators is in “divergences” from price, or put another way, when that small bit of uniqueness in the indicator causes it to drift away from the price, signaling something that the price doesn’t yet reflect.

That’s a nifty concept.

My response would be that I’ve spent probably a hundred hours trying to figure out how to do that (mechanically) with the VIX. I’ve spent a fair amount of time trying to do that (again, mechanically) with the McClellan Oscillator. And at the end of the day, I’ve come up empty-handed.

I’m not saying it can’t be done, but if it can, the recipe for the secret sauce has eluded me.

P.S. I hate going out on a negative note, so I’ll do one final follow up on the MO. And I promise that I’ll have something nice to say this time.

Comments

[Lincoln:]

I support your feeling that mechanical approaches to well-picked-through public data are lacking to empty in predictive value. I use the oscillator in context only, and the context is definitely not mechanical. For example, in the current rally -- that now seems to be fading -- the oscillator can contribute to observations on the strength of the rally, the breadth of the rally, and the particular sectors that participated in the rally. In this case we have relatively low volume rally that was led by weak sectors, and triggered by propaganda and, perhaps, market intervention. The rally also seemed unusual in that volume peaked at the beginning and the end of the trading days in a pattern that repeated for weeks. This seemed to reflect 1) the market's hanging on information that was being released either before or after market hours, and 2) an unusual participation perhaps related to program or day trading.

In this case the SP500 oscillator reflects, I think, something about the broader markets participation in the rally. What I'm looking for is to see the oscillator repeatedly return to the mid-range area. To me this "sputtering" implies weakness, and it confirms the weakness in the rally itself. What I'm trying to say is that I'm using the oscillator as a confirmatory indicator, and not as a predictive one. Does that make sense?

Apr 27 02:23 AM