Price Short Term

The short version of price is more for shorter-term assessment of trendiness. It is simply looking at the price relationship in the 5 to 21 day range. If you were using multiple price measures on the same index, then this is the one that would turn on first and also turn off first; it is the quickest to respond to changes in price direction. Many times a short-term measure is not actually used in the weight of the evidence calculation, but serves a weight of the evidence model well with an advance warning of things to come.

Price Medium Term

This is another price trend measure that uses a different analytical technique and different look-back period than the Trend Capturing component. This Price Medium measure is basically looking at the price relationships over a three- to four-week period. If the short-term price measure isn’t used, then this is the one that will lead the changed in direction of the index being followed.

Price Long Term

This price trend measure is similar to Price Medium but looking at the price relationship over a four- to eight-week period. Generally the Price Medium indicator will turn on first and if the trend is sustained the Price Long measure will turn on thus providing confirmation of the trend and further building the point total of the cumulative weight of the evidence.

Figure 13.5 is an example of the Price Long measure. Although this is almost too much data on one chart, you can focus on the binary overlay on the top plot and can see that it does a good job of tagging the uptrends, which is all we want it to do.

Figure 13.6 shows the same price measure as the one above, just a smaller time frame. It becomes much clearer that the binary line overlaid on the top price data moves in conjunction with the indicator in the bottom plot, the price long measure. Whenever the price long indicator moves above the horizontal line, the binary moves to the top, and whenever the price long drops below the horizontal line, the binary drops back to the bottom. You can then see that whenever the binary is at the top, it is signaling an uptrend and whenever it is at the bottom it is signaling no uptrend (down trend or sideways). This concept is quite valuable since it allows you to view only the binary to know what the indicator is doing. So is this indicator perfect? Of course not, you can see there was a whipsaw signal (short and wrong, but also quickly reversed) near the right center of the chart where the binary was at the top for only a very short period of time. As I have said before and will no doubt say again, I know these measures will be wrong at times, but they are not going to stay wrong because they react to the market. Because these indicators are all trend following, they will reverse a wrong direction almost as fast as they identify it in the first place. That is exactly what you want them to do and it is also why I use a weight of the evidence approach, which means I rely on a basket of technical measures. Sort of a democratic approach, if you will.

Table 13.1 is an example of the detailed research behind each of the various indicators used in the weight of the evidence. This example is over the period from 1980 to 2012, however, it should be run over many different periods (see Figure 13.4) to find consistency.

Here is an explanation of the various parameters in Table 13.1:

1. Parameter. These are the variable parameters used to test the indicator’s usefulness in identifying trends. They can be discovered by optimization, or just a visual analysis of the indicator. They are then spread about to cover a wide range of analysis for the indicator.
2. Winning years. The percentage of years that the trading ended higher than the close from the previous year.
3. Trades per year. This is the number of closed trades per year.
4. Average return per year. Determined by looking at the total return for the entire run, and then dividing it by the number of years.
5. Total return. The total return from the system using the selected parameter.
6. Compounded annual return. The gain each year that would be required in order to achieve the total gain for the analysis period.
7. Compounded annual return while invested. The annual gain that would be required to achieve the total gain, excluding cash positions, over the period being analyzed.
8. Percentage of time invested. The percentage of time that actual trades were placed and not in cash.
9. Ulcer index. A measure of downside volatility also covered in the ranking measure section of this book.
10. Largest trade loss. A single trade that resulted in the largest loss.
11. Maximum drawdown. The maximum decline of the system measured from the highest level that the system had reached. Keep in mind this is a one-time isolated event.
12. Ulcer performance index. The compounded annual return divided by the Ulcer index, this is a performance measure similar to the Sharp ratio, the Sortino ratio, and the Treynor ratio. All are risk-adjusted measures of performance.

Although the calculation of all the various measures of an indicator’s ability to work over a vast number of trials and time segments, one still has to determine “Where’s the beef?” Of all that data generated, shouldn’t you list the categories from best to worst inasmuch as their contribution to what you looking for? I think so, definitely so. One method and the one I use is to ask some sharp folks who are deeply familiar with the indicator and the output to give me their input as to the order in which the parameter analysis should be viewed. And to keep this as a robust process, it is done each year. Often there isn’t much change, but sometimes someone gets a more involved feeling after working with these almost every day as to which is more important than another. Sometimes the overall relative ranking of these gets changed.

Table 13.2 shows that each column of data is ranked based on its relative importance as determined by the individuals involved in the portfolio management process. This relative input is then weighted based on the level it has reached in the vetting process. There is an old forecasting axiom that says the average of all estimates is probably going to work better than trying to select the single best guess.

Note: The columns with Inverted at the top mean that the smaller the value, the better it is performing.

The relative ranking is then placed alongside the parameter output to see where the current parameter being used and followed immediately by the ranking of the output components shown in the first column of Table 13.2. This process allows you to see how the parameters change over time compared to the one currently in use. Note that the parameters in the first column of Table 13.1 are in numerical order, so while this is a change in the relative ranking, if the change in the parameter is small then generally no changes will be made to the parameter in this weight of the evidence indicator. However, it will be watched over time and if there is an obvious drift away from the parameter being used, a change will be considered.

The above performance statistical information provided for the price long weight of the evidence will not be included in all of the indicators that are used as this section would get overly long. However, when something stands out that will provide additional insight into this process; the information will definitely be shown. However, do not fear; a chart or two will be shown for each measure.

Risk Price Trend

The risk price trend is in the lower plot of Figure 13.7, while the top plot is the Nasdaq Composite and the risk price trend binary. This measure uses the MACD concept with considerably longer parameters for both the short- and long components. You can see from the binary that it does a relatively good job of picking out the trends of the market.

Adaptive Trend

The adaptive trend measure incorporates the most recent 21 days of market data to compute volatility based on an average true range methodology. This process always considers the previous day’s close price in the current day’s high low range to ensure we are using days that gap either up or down to their fullest benefit. When the price is trading above the adaptive trend, a positive signal is generated; and when below, a negative signal is in place. This is clearly shown in Figure 13.8 with the adaptive trend binary overlaid on the price chart at top. Whenever the binary is at the top, it is showing an uptrend, and when at the bottom, a downtrend.

Advances/Declines

The advance/decline component measures the relationship of advancing issues to declining issues. Advancing issues outnumbering declining issues is a positive event as more participation is taking place in the equity markets. However, an inverse relationship (decliners outnumber advancers) would be a clear sign of weakness to a price movement. There is a recent example of the clear divergence in the price movement and market internal movement during much of 2007 and particularly in the fourth quarter of 2007. From early October 2007 to late October 2007 we saw a distinct positive price trend while the advance decline measure was rapidly declining. By late October the markets were reaching all time new highs. What you were reading and hearing in the news was all positive, “The market reached an all-time new high.” According to the price movement things couldn’t have looked better. However, the advance/decline measure was telling us a very different story as declining issues continued to outnumber advancing issues. This forewarning was signaling increased risk as the upward price movement was not being supported by broad participation. At that time, the advance/decline component was negative; hence, it was not contributing to the weight of the evidence total. Divergences can cue us to be prepared in the event the price action reverses rapidly. In fact, that is exactly what happened and it was the beginning of the 2007 to 2009 bear market.

Figure 13.9 shows the deterioration in the advance decline measure in 2007. You can see that while the market was reaching new highs, the advance decline measure never even got back up to its horizontal signal line. Breadth does this time and time again. It almost always leaves the party early and is a great tool to have in your arsenal as a trend follower.

Figure 13.10 is the same indicator but plotted with the same data used throughout this section so cross reference is easier.

Up Volume/Down Volume

This Breadth measure gives us an internal look at the volume behind the price action by looking at the relationship of the volume in advancing issues versus the volume in declining issues. Again, by removing the capitalization weighting of the price movement that can drive the index price, this measure gives us an internal view of what is going on behind a price movement. Generally, if there is strong up volume relative to down volume it is a positive sign that the positive price movement is being supported. However, if there is positive price movement, but down volume is greater than up volume it is a sign that the price movement lacks participating volume. In this case the up volume/down volume measure would not contribute to the weight of the evidence leading to a lower total level signaling increased risk. Figure 13.11 is the up volume/down volume weight of the evidence measure. You might notice that the signal line (horizontal line) is not at zero like many use, but in this case it is at +400. The parameter analysis identified that this was a significantly better signal level. You can see that the up and down volume has been quite weak since and before the peak in prices, which occurred in the middle of the time period shown.

New Highs/New Lows

The High/Low Breadth measure looks at the relationship between issues reaching new 52-week new high values to issues hitting new 52-week new low values. Generally, when the number of issues reaching new 52-week highs are outpacing the number reaching new 52-week lows there is a positive indication in support of positive price action. The flip side of more new lows to new highs is a negative indication, and this lets us know that there is potential risk that the price action and prices could possibly turn quickly in the other direction. When this indicator is on we have internal support for positive price movement and it adds to the point value of the weight of the evidence.

Note: See Appendix C for High Low Validation Measure for expanded information.

This weight of the evidence measure is a little different than the previous ones in that signals are generated by the crossing of the indicator’s moving average shown in Figure 13.12 the bottom plot as the dotted line. You can see the binary in the top plot makes seeing these crossings quite easy.

Breadth Combination Measure

The combination breadth measure (Figure 13.13) uses advancing and declining issues as well as advancing volume and declining volume by weighting the significant volume days inside of the indicator. This component serves to underweight insignificant price movements if the volume is weak while weighting the more significant price movements when volume is significant. This measure was created out of concern for days such as the Friday that follows Thanksgiving, a partial day of trading, and always with very low volume. The downside of using breadth is that it does not matter if the day is shortened or the trading is light, you will always end up with a full complement of breadth data. Remember, breadth does not measure magnitude directly, only direction. In this measure, when the up volume is below a predetermined moving average, then only the advances are used. If the up volume is above this average, then the product of advances and up volume is used. The same concept is applied to the declines and the down volume.

Breadth Is Not Always Internal Data

There are a number of breadth-based measures that do not use internal market data such as advances, declines, up volume, down volume, new highs, or new lows. A concept known as the Percent Participation Index is used by many analysts. The concept is simple yet revealing, it measures the number of stocks in an index as to where they are relative to their moving average. For example, many use a 200-day Participation Index, which shows the percent of stocks that make up the index that are above their own 200-day moving average. Figure 13.14 shows the Nasdaq Composite in the top plot with its 200-day simple moving average. The bottom plot is the 200-day participation index, which shows the percentage of all Nasdaq Composite stocks that are above their respective 200-day simple moving averages.

Figure 13.15 is similar to the previous one except that it shows the 50-day participation index.

Small Cap versus Large Cap Component

Small cap participation is critical for sustained uptrends in the market because small caps reflect speculation and speculation is a requirement for uptrends. For small capitalization issues the Russell 2000 Index is used and for large capitalization issues the venerable S&P 500 Index is used. The math is simply to create the ratio of the small to large, then use technical analysis to provide a normalized trend measure. In Figure 13.19 the typical Nasdaq Composite is in the top plot along with the small cap large cap binary. The middle plot is just the ratio of the small to large issues, and the lower plot is the small cap large cap weight of the evidence measure. In this example you can see that small caps dominated the up move near the center of the chart, but have been relatively weak since. If you think about it, small caps also more closely relate to breadth.

Growth versus Value Component

This component of relative strength measures the difference between growth stocks and value stocks. Now this will shock most folks who have studied the markets, but I use small cap issues for both the growth and value components. I was initially concerned that small cap value was almost an oxymoron, but the data has proved time and time again that it is perfectly valid. Figure 13.20 shows the growth value ratio in the middle with the growth value weight of the evidence measure in the bottom plot and its binary overlaid in the top plot.

Breadth versus Price Component

There is only one more relative strength measure to view, the relationship between breadth and price. For breadth, a relationship between the advancing issues and the declining issues is used, while the Nasdaq Composite is used for price. The middle plot in Figure 13.21 shows the raw ratio of those two, with the price to breadth weight of the evidence measure in the lower plot and its binary in the top plot. This is an excellent example of how technical manipulation of the data can turn that noisy middle plot into something like the bottom plot, and a good trend measure.

Relative Strength Compound Measure

This compound measure (see Chapter 12) is designed to measure market sentiment. Are investors actively taking investment risk? Or are they behaving much more bearishly? Again, there are three indicators that drive this component and are shown above. These indicators look at the relationship between small cap issues versus large cap issues, growth-oriented issues versus value issues, and breadth measurements versus price measurements. For example, when small caps are dominant (small caps outperforming large caps) it is generally a sign of more speculation taking place in the markets as investors are willing to accept more risk. This is generally a good sign and it is during these environments that the markets generally perform well historically. When large caps are dominant it is usually the result of a flight to quality as investors are taking risk off the table. This type of investor sentiment usually results in less favorable market conditions. A similar relationship exists between growth and value. When growth issues are outperforming it is generally because the market is pricing in favorable growth estimates, and conversely, when value issues are dominating it is because the market is no longer pricing in such optimistic positive growth estimates. The breadth versus price measurement is similar to an equal weighted measurement versus a capitalization weighted measurement. Again, we are using three indicators in this component for confirmation purposes. When this relative strength component is on it is adding value to the weight of the evidence of the model by letting me know that the sentiment in the markets is favorable and is viewed more as a trend confirmation measure versus a trend identification measure. Figure 13.22 is a chart showing lots of data but includes all three components of the relative strength measure. The top plot is the Nasdaq Composite with the compound binary measure overlaid. The binary measure for each of the components is also shows overlaid on their respective plots.

Figure 13.23 is the same as the preceding figure, but with less data displayed.

Advance Decline Component

The advance decline component of the trend capturing measure uses the advances and declines difference and then mathematically puts that difference into a relationship similar to MACD. (See Figure 13.25.) The signals are given by the crossing of that formula with its shorter term (10–18 periods) exponential moving average.

Up Volume/Down Volume Component

The up volume/down volume component of the trend capturing measure (see Figure 13.26) uses the up volume and the down volume in a similar manner as the advance and decline measure in Figure 13.25. One could assume that these would be very similar as they are tied to the same price movement, one by daily changes, and one by the amount of volume behind those changes. However, in the up volume/down volume, the parameters used are somewhat longer than in the advance decline component.

Price Component

The price component of the trend capturing measure (Figure 13.27), like the other two components uses a similar relationship, but this one uses parameters that are longer than the other two.

Trend Capturing Compound Measure

This is a major component of my model because it ties directly with the investment philosophy of identifying positive market trends that have a high probability of continuing into the future. I never predict; I only have an expectation that the identified trend will continue. This component is a composite of three technical measurements. One is a price measure, and the other two are breadth measures, one of which uses up and down volume, and the other uses advancing and declining issues (see Figures 13.26 and 13.27). I use multiple indicators for confirmation purposes. For instance, when the first of these three indicators turns positive it is telling us there is a positive price trend developing. When the second of these indicators turns positive it provides confirmation of the trend, and if the third turns positive as well I know I have a very solid trend in place. See Figure 13.28.

Mega Trend Plus

Mega Trend Plus is constructed by selecting 11 major indexes with the longest historical database. One could easily make the case for more or less indices and which indices are to be included. I would suggest using enough indices to give you broad coverage over how you plan to make investments. The Mega Trend Plus is also used in the Long-Term Measure (LTM), but in that case it uses weekly data. Clearly if you are going to have a focus on international securities, you would want to include international indices. The list of indices used in this example is:

1. Nasdaq Composite
2. S&P 500
3. S&P 100
4. Russell 2000
5. Russell 2000 Growth
6. Russell 2000 Value
7. New York Composite
8. Dow Jones Industrials
9. Dow Jones Transports
10. Dow Jones Utilities
11. Value Line Geometric

When the index is greater than its exponential moving average (EMA), it receives a +1, below its EMA, it receives a −1. Then the scores from all 11 indexes are summed and then normalized so the composite will have a range between 0 percent and 100 percent. An uptrending market is called when the composite is greater than or equal to 85 percent. A downtrending market is when the composite is less than or equal to 15 percent. Once a threshold is crossed, the market stance stays until the opposite threshold is penetrated. Hence Mega Trend Plus is a digital meter: either bull or bear. Figure 13.37 shows the S&P 500 in the top plot, Mega Trend Plus in the lower plot, and the Mega Trend Plus binary overlaid on the S&P 500. The binary is at +1 whenever Mega Trend Plus is greater than or equal to 85, at −1 whenever Mega Trend Plus is less than or equal to 15, and at zero when it is between 15 and 85. You can see that is does a reasonable job of trend identification. This is a weight of the evidence approach that is totally related to the concepts in this book.

Trend Strength

Trend Strength is a different composite, which is the sum of 11 ratios. For each market index, a ratio is calculated with this formula: (price/EMA−1) × 100 percent. The EMA lengths are identical to those used in Mega Trend Plus. The ratio depicts how far away (up or down) the index is positioned relative to its EMA. Hence Trend Strength is an oscillating analog meter that measures the momentum of each of the 11 market indexes (see Figure 13.38).

Trend Gauge combines the readings from both the digital and analog meters (Mega Trend Plus and Trend Strength). It represents a weight of evidence approach in determining both the direction (digital) and the strength (analog) of the overall market trend. Figure 13.39 shows the S&P 500 in the top plot and Trend Gauge in the lower plot. Whenever the Trend Gauge is at +1 it has identified an uptrend, when at −1 it has identified a downtrend, and when at 0, there is neither an up- or downtrend (neutral).

Figure 13.40 is the same data as in Figure 13.39, just over a much longer time frame. The beginning date in Figure 13.40 is 1992.

Measuring the market is a significant component to a good trend following, rules-based model. This chapter has introduced many measures that can be used individually or in groups to assist in trend identification. Clearly there were many measures introduced and hopefully you can find a few that will fit your needs. Also, hopefully you will not use all of them as many are similar with minor deviations in their concept.