The Moving Average Spectrum - Part 2

This is a follow up to my post The Moving Average Spectrum. Like the first, this one is a bit on the complicated side, but I think important to understanding how to trade these unruly markets.

My last post demonstrated the idea that:

Short-term (let’s say, 1 day to 1 week) and intermediate (1 week to 1 month) indicators tend to be contrarian (i.e. revert to the mean). Longer-term indicators tend to be momentum/trend-driven. The contrarian tendency in short and intermediate-term indicators (especially the short-term) is a recent phenomenon.

In this post, I want to expand on that idea and show that:

As we move up the spectrum towards shorter and shorter-term indicators, they tend to be more predictive of next-day returns, but also less stable over time (think RSI(2) versus 50/200-day MA crossovers), and better candidates for an adaptive approach.

Recall a similar table to the one below in my first post. It shows next-day performance, broken down by decade, when the S&P 500 closed above (top table) or below (bottom table) a moving average of length X. A higher positive number indicates greater bullishness during that decade, and a lower negative number greater bearishness.

click to enlarge

Geek notes: Like the first post, the table has been adjusted by (a), limiting daily changes to +/- three standard deviations to reduce the impact of a handful of large market moves, and (b), removing the impact of the broader trend by subtracting from each decade’s returns the (geometric) average return of that decade. However, unlike the first post, I have NOT adjusted returns to account for volatility.

Warning: New Geekery Ahead

In the next graph, I’ve plotted the average absolute value for each column. In other words, I’ve ignored the + or - sign. Here I don’t care if something was bullish or bearish, only that it was predictive in that decade. Average next-day return is on the y-axis and the length of the MA on the x-axis.

click to enlarge

The graph above shows that very short-term indicators have been far and away the most predictive of next day returns (ignoring direction).

But that’s only part of the story.

The next graph looks at just the top set of data in the original table (i.e. closes above each moving average) and plots the difference between the maximum and minimum values in each column. Here I care about how unstable results have been.

click to enlarge

This second graph shows that, despite being the most predictive, very short-term indicators have also been the least stable (i.e. the difference in performance across decades was the greatest).

This test only includes one very simple indicator (that I would not suggest anyone trade), but I think the conclusion is generally applicable to ALL indicators that are based purely on asset price.

At any given time, short-term indicators (ex. daily follow-through, RSI(2)) tend to be much more predictive than longer ones (ex. 50/200-day MA crossovers), but are also more likely to fundamentally change over time.

We see this in how stable things like 50/200-day crossovers have been historically, and how very unstable (but also more powerful) something like RSI(2) has been.

Warning: Personal Opinion Ahead

I think the optimal solution is two-fold: First, treat longer-term indicators as a sort of foundation for the trade. Like in the State of the Market report, they might also be used to bias the trader’s position at any given time.

However, I don’t think it makes sense (except perhaps for extremely lazy or inept traders) to buy or sell based on long-term indicators alone. They are just so much less powerful than shorter-term ones.

So on top of that long-term bias, trade shorter-term indicators, preferably using an adaptive approach (like I’ve written about a number of times on this blog and we also use in our YK and Scotty strategies) that adjusts with the market as it inevitably evolves.

Just my $0.02.

You can read Part 1 here.

Comments

[oldfox:]

What a bucketfull of garbage.

M

Jan 19 09:35 AM

[oldfox:]

Opium dream.

M

Jan 19 09:37 AM

[Michael S:]

Oldfox - Your anecdotal shot from the hip is appreciated for what it is...an anecdotal shot from the hip. With all due respect - click through to the blog - checkout the independently-audited real-time returns of our programs (which are in large part built off of these fundamental concepts) and then tell me (preferably empirically or at least thoughtfully) how this is a garbage idea. ms

Jan 19 10:44 AM

[d_teller:]

Since 1986, I used the same process to add funds to my 403(b) account when I, too, perceived a relatively bearish environment was persisting...and meanwhile I "Milked" the over-performers to a cash subaccount. The overall return was 9.28% from 1986 through 2007. However, this is a value-based, not a time-based (fixed-interval), method of "dollar-cost-averaging...
The longer term moving averages were useful for noting both "out-performance" of specialized and sector funds, and "under-performance" as well. I appreciate all the time it took you to generate these spectra. I just used multicycle log-paper graphs to plot the S&P and MSCI or other general sectors, and bought when the 200-day moving avg, was more than a bit below the general trend; + sold when the holding was similarly above the expected trend.
With Mutual funds in a tax-sheltered environment, this allows for a greater purchase of shares when they're relatively cheap, and avoids blindly following a calendar-based purchase of expensive shares. At the beginning of 2008, the swans were beginning to darken, and I moved 40% to cash, despite the low interest rates, and in May 2008, sold off all but two highly-valued (over-priced) REITs from my separate IRA.

Despite the current valuation carnage, I'm comfortable with my holdings and plan to add beaten-down energy + materials securities during the current commodity sell off.

Jan 19 08:45 PM