Technical analysts assert that the market can move in an often predictable fashion. Why else do we study the patterns on a chart if not to find occasions where a pattern could be repeated. In reality, though, the only thing predictable about the market is that it will brutally squash all those deluded enough to believe it is systematically predictable. In fact, it really doesn’t matter what investment or trading approach is applied - even the most rigorous market pricing model of cause and effect is too clumsy to time the market much better than a coin toss would (check out this forecast study: It's Official! Gurus Can't Accurately Predict Markets). Perhaps the only real merit to all of the resources that go toward modeling the market has much more to do with risk management optimization - systems and processes that help insure against bad predictions. The best offense is a good defense.
Enough about reality and portfolio management principles, though. We’re market timing believers because 1) flipping a coin isn’t as satisfying as scientific method, and 2) even a normal probability distribution of a random statistic shows there’s room for someone in the 95th percentile. Let’s just say most market timers aspire to be positive outliers – smarter than the average investor and richer for it. Stock Trends hopefully will help you be in this “heavy-tail” company and makes a proposition that the key to improved probability of market timing success is, in fact, simplicity: define a price trend, learn how the distribution of past returns of those trends favour a trade or not, and then execute an order within a trading plan that emphasizes prudent money management. Successful market timing is not so much about being a high rolling hare; it's the parsimonious trading turtle who builds wealth.
The statistical language used here may be familiar and elementary for some, but I’ll try to keep all Stock Trends commentaries as plain and clear as possible. In the past Stock Trends editorial was largely typical technical analysis storyboarding – finding existing market trends and isolating particular stocks that are in a ‘good’ technical position for a trade. Now the emphasis will be on learning if and how the Stock Trends indicators predict probable outcomes. We will seek to quantify probable outcomes based on particular observations and the combination of variables recorded. That process of discovery will be an incremental learning exercise. Let's begin!
The Stock Trends Variables
Stock Trends is essentially a handful of discrete and continuous variables. You can see them in every Stock Trends Report on the website. They include the trend (,,,,,) and volume (,) icons, RSI values, and trend counters. I will be providing more explanatory commentary on these variables and discovering underlying relationships between the Stock Trends indicators and future price performance. After all, technical analysis may be rooted in historical data, but it is future results we would like to predict.
The most important variable is the trend indicator. The trend indicator is a factor variable with 6 possible values:
Bullish Crossover (also referred to as Newly Bullish)
Bullish (also referred to as Strong Bullish)
Bearish Crossover (also referred to as Newly Bearish)
Bearish (also referred to as Strong Bearish)
All issues and indexes with at least 40-weeks of trading data in its time series are assigned one of these values every week. These six indicators are defined here and in the Stock Trends Handbook, Chapter 4 - Guide to Stock Trends Symbols and Indicators.
However, the major trend categories are Bullish or Bearish. Each stock - even if it were more accurately modeled to be defined as a ‘flat trend’ – must be either Stock Trends Bullish or Stock Trends Bearish, depending upon the relationship between the 13- and 40-week average price. While there are problems that go with this binary denotation, it is a simple method of grouping our weekly observations. Once we separate observations into subsets we can begin to understand the meaningfulness of the grouping. Of course, its meaningfulness can only be evaluated if the division shows different results. In this case, does the future performance of stocks categorized as Bullish (,,) differ from that of Bearish (,,) stocks? Statistical analysis of our historical weekly data tells us the answer.
Differences in Central Tendency
As an initial step toward understanding the relationship between the indicators and future performance let’s take a sample from 13-weeks ago – October 12. Since then the market has retreated, then rallied – with the S&P 500 index now up 3% from its level on October 12th. We can now analyze a basic statistic provided weekly by Stock Trends for all issues: the 13-week price change. How does the mean (average) price change vary for each trend variable? That is the simple question we will answer.
For reasons we will examine at another time, this analysis will be limited to issues valued at $5 and greater. Controlling for only those issues of common stock and trust units that had a trend indicator 13-weeks ago, there are 5,134 observations (we will ignore ETFs and indexes for different reasons). Of course, the range of price change is large – from a painful -92% (Petrobank Energy and Resources PBG-T) to 187% (Uni-Pixel UNXL-Q). The mean (average) percentage change of this sample is 4.5%. This is obviously higher than the 3% advance by the market benchmark indexes, but remember that the mean is not the only measure of central tendency. The median percentage change over this period – the value that sits in the middle of all observations – is 2.7%. Our question has to do with how one measure of the statistic (13-week price per cent change) varies between different subsets of the sample that are determined by the Stock Trends indicators.
If our binary grouping of Bullish and Bearish stocks is meaningful, there should be some variability in the median values of each group. What do we find?
Of the total 5,134 observations 3,254 (63%) had Stock Trends Bullish trend indicators – , , or . The mean percentage change in the 13-weeks for these Stock Trends Bullish major trend stocks is 4.2% (median 2.3%), while the mean percentage change for Stock Trends Bearish major trend stocks (,, or ) is 5.2% (median 4.2%). Here we do observe there is a difference in the performance of the two groups (Bullish and Bearish), although it is not a result we might have expected.
Certainly, we should sample other sets of records from different times and markets to see how the two major trend categories perform relatively. Of particular interest, too, are measures of variance (like the standard deviation) that quantify the distribution of the results. But for now we’ll simply drill down into these observations from October 12 and stick with the mean statistic (average price change) as a comparison for the Stock Trends trend indicators.
Why would Bearish stocks perform better than Bullish stocks in this sample? We should have a look at the breakdown of the statistic for the six minor trend categories (,,,,,) :
|Breakdown of minor trend categories (October 12, 2012)|
# of issues
The most notable statistic here is the performance of Bearish () stocks – clearly the difference maker in the results recorded in the period. During the period this particular group of stocks did better than Bullish trending stocks, and better than the Weak Bearish () stocks, a group we focus on for our Stock Trends Picks of the Week selections – although a Bearish stock generally travels through a Weak Bearish category before its Bullish Crossover () and would still qualify for a Picks of the Week selection subsequent to observations on October 12. (Take note, though, of the median values and the way the differences between the mean and the median say something about the distributions of returns in each group. This is something to be evaluated at another time.)
Regardless of the reason for the impressive move of these Bearish () stocks, we should break down this subgroup further to see if there are particular characteristics recorded by the other Stock Trends variables that might indicate commonality. Those characteristics would be something we would look for in other samples.
Stock Trends keeps track of the time period of the major trend – number of weeks an issue or index has been categorized in a Bullish or Bearish major trend – as well as the time period of the minor trend. (See Trend Counters) The minor trend counter is the number of weeks an issue or index has been in its current trend indicator (,,,,,). The minor trend is more sensitive to price movements and is both an indicator of strong, as well as weak and changing trends. How does the length of the major trend and minor trend indicate the post-observation results from this subset of Bearish () stocks from October 12?
Separating the major trend counter into two groups offers some immediate information.
Of the 1,342 (Strong) Bearish () observations, 263 had Bearish major trends over 26-weeks long. Most of the (strong) Bearish () observations on October 12 had been in a Bearish major trend for less than 26-weeks.
|# of issues||Length of major trend||median||mean|
|263||Bearish major trend at least 26-weeks long||0.4||1.5|
|1079||Bearish major trend less than 26-weeks long||5.6||6.4|
Those results again tell us something interesting: the better performing stocks were relatively short-term Bearish stocks. The mean 13-week percentage change of Bearish () stocks in a Bearish major trend for less than 26 weeks was 6.4% - much better than the 1.5% recorded by those above 26-weeks long in their Bearish major trend.
Here we see that entrenched long-term Bearish trends are not stocks that this analysis would favour, and that a contrarian long trade has long odds. And some Bearish stocks are not as entrenched in a trend as others.
That a trend ends is no surprise. They always do… eventually. Nor is it surprising that some trends are not as pronounced or as forceful as others. How can we evaluate the character of a trend? We can drill down further into the trend counters to see if there is something about the quality of these Bearish trends that perhaps would provide a cue to the change ahead.
The minor trend counter is an indicator of the quality of a long-term trend. If the minor trend counter is relatively low compared to the major trend counter, we know that the major trend has included shifts into the “Weak” iteration of the trend – here a Weak Bearish () trend. A low trend counter for this group indicates a recent switch back to (Strong) Bearish (). This obviously shows that the long-term trend is not as entrenched as those with a minor trend counter that is relatively high.
What does the data from October 12th show? The best performance from these Bearish () stocks (major trend < 26 weeks) are those that have been in a Bearish minor trend for a relatively short time.
Bearish stocks with major trend >= 26 weeks and minor Trend < 4 weeks
|# of issues||Minimum||Median||Mean||Maximum|
Bearish stocks with major trend >= 26 weeks and minor Trend >= 4 weeks
|# of issues||Minimum||Median||Mean||Maximum|
Bearish trends bend
We’ve laboured through this elementary statistical exercise to illustrate that while Bearish trends do reverse, that stocks eventually complete their downward drift, transformations from bear trends are often part of a process that is akin to a bending of a line, rather than a breaking of it. Finding potential breakout stocks in the Bearish () group is risky, though - regardless of how statistically appealing this analysis of observations presents. This is the reason Stock Trends focuses on Weak Bearish () and Bullish Crossover () categories in our stock selections and trading strategies. They are transformational categories by definition. However, Weak Bullish () stocks are also transformational and certainly of interest since many of these stocks rally back to their strong Bullish trend (). We'll take a statistical look at that category soon.