NELSON'S LAST SIGNAL TO THE FLEET AT TRAFALGAR "ENGAGE THE ENEMY MORE CLOSELY"
The great naval battles of the 16th-18th century were fought with a large number of huge sailing ships all moving in a straight line, firing batteries of cannon broadside at each other at short range. Hence, the biggest warships were called line-of-battle-ships. The commanding Admiral could only use flags to signal maneuvers to his fleet, through all the din and smoke, and hence information compression took on a whole new role (see more about Trafalgar later). Naturally, the ship that bore the Admiral was called the flagship, and even today, the largest ships are called battleships. If I were to use flags, the name Trend Check would appear as shown in Chart 1.
Chart 1: There are so many rules on composing flag messages at sea, that I almost certainly have it wrong, but here I am trying to say "Trend Check".
At sea, signal flags compress information, just as information compression is a key theme of the Trend Check blog. Since we are still in the start-up phase, today I will introduce the Trend Check Carpet, in which we will summarize the trend status of key indexes across four time periods and market breadth varying from 30 to over 2000 stocks. The purpose is to help you grasp what is happening in the market at a quick glance. It will look complicated, and will take some time to become familiar, but if it makes you feel any better, even today, the NATO Navy Flag signal rules fill many 3" thick binders.
THREE-PHASE TREND-FOLLOWING MODELS
I want a three-phase model: one that can be long, flat or short, unlike moving average cross over models that are only long or short (i.e. two-phase). First, we start with daily data. Second, we use Bollinger bands with band width of half a standard deviation. Third, we combine the Bollinger bands with simple moving averages (of the close). When the moving average is above the upper band, the model is long. Conversely, when the moving average is below the lower band, the model is short. So, it follows that when the moving average is between the upper and lower band, the model is flat. Thus, we have a three-phase model, which can be long, flat or short. One of the advantages of the 3-phase model is that we do not get whipsawed by repeated moving average cross-overs in a short period of time. However, we become vulnerable to false breakouts, where the market moves enough in one direction to confirm a change in direction, only to promptly reverse. Thus, the three-phase models have strengths and weaknesses, just like two-phase models.
We then have to choose the length and band widths for the models. I chose completely arbitrary combinations of moving average and band lengths to define four models as shown in Chart 2 below. In other words, these are not optimized in any way.
Chart 2: The Specification of Trend-Following Models
Observe that for each time period, the length of the calculation period for the band is 5 times the length of the time period of the reference moving average. The time period length doubles at each step from short-term through long-term. Thus, the bands scale more or less consistently across time. Besides, this allows us to use the 50- and 200-day time periods so popular across the charting universe. Naturally, our classification of any particular length as short-term or medium-term is also arbitrary, and you might prefer some other designation. We can now apply these models across key indexes and summarize the data in a single chart or carpet.
TREND CHECK CARPET
Chart 3: The Trend Check carpet covers four time periods (from top to bottom). Market breadth, or the number of stocks included in the indexes, increases from left to right, rising from 30 in the Dow Industrials to over 2000 in the NYSE composite index.
First, the four time periods are stacked from shortest at the top to longest at the bottom. Thus, in any vertical column, you can see how each index is trending from a short- to a long- time frame. In any time slice, you can read from left to right how the market is doing as the number of stocks in the index (i.e., market breadth) increases.
We only classify the trend into long, short or flat, without regard for the trend strength. At one extreme, the market could be falling across all time periods and index breadth. At the other extreme, the market could be rising across all time periods and index breadth. Most of the time, the market will be mixed. As a strong down-trend begins to reverse, the uptrends will appear at shorter intervals and narrow breadth and spread throughout. Conversely, after strong uptrends, down-drafts will first occur at wider index breadth and shorter periods and then diffuse through the picture.
As a user, you may choose to focus on just one given time period (say the long-term) or one market breadth (say the Russell 1000) as best fits your trading style or portfolio.
CONSISTENT WITH OTHER BLOGS
The market carpet is broadly consistent with other blogs on StockCharts.com. For example, the Trend Check Carpet is similar to the Erin Heim's Decision Point Alert Scoreboard. I cover one extra time period and few extra indexes in Chart 3, and the underlying models are three-phase versus two phase. I just want to summarize the trend picture, rather than highlight individual signals. As you become familiar with the layout and review it regularly, you will see the dynamics of the market play out before you.
The current summary is also broadly consistent with Arthur Hill's Art's Charts blog. He recently wrote that he feels we are still in a corrective phase in the market. The flat or down trend status for the major indexes (except QQQ) over the short-term and medium-term is consistent with the markets being in a corrective phase. Here we can quantify the state of the market over time and across market breadth, to indirectly support his analysis. Tom Bowley wrote a piece why this is not 2007, and the summary in Chart 3 at the intermediate and long term time interval shows the indexes trending higher quite consistently across the breadth of the markets.
I have gone on too long already. Until next time, remember, there is strength in numbers.
INFORMATION COMPRESSION AT TRAFALGAR
In the heat of battle of Trafalgar on 21st October 1805, Nelson first ordered his Signals Officer, Lt. Pasco to hoist "England confides that every man will do his duty". However, the word 'confides' was not in the code book, and each letter would have to be spelt out, taking time and precious space on the halyards. Lt. Pasco asked permission to replace 'confides' with 'expects', a word that was already in the signal book. Nelson agreed, giving us the unforgettable message, "England expects that every man will do his duty". Admiral Nelson's final instruction had just three flags: the telegraphic flag and the numbers one and six, "Engage the enemy more closely". How's that for information compression?