CHANNELS: EMOTIONAL EXTREMES
A moving average may give some idea of consensus, value, and relative rest. Properly drawn channels define excursions from that area of consensus and may give some insight into where the crowd’s emotional extremes lie.
In most applications, this is the one and only purpose of channels: to define meaningful extensions. In order to do this well and consistently, the channels must be able to flex and to adapt to the changing volatility conditions of the underlying market. It should be obvious, but channels do not provide any kind of actual barrier to prices.
The idea of buying at a band or selling at a band, with a few exceptions in highly mean-reverting instruments, is conceptually flawed. A better way for discretionary traders to use bands is to look for touches of or excursions out-side the bands, and then to carefully monitor price action following those events, treating the band as an alert level.
Probably the most commonly used channels are Bollinger bands, which are set at a multiple of the standard deviation of price around an average of price. (Default settings are usually two standard deviations around a 20-period SMA).
As the market becomes more volatile (as measured by standard deviation of price), the bands automatically widen to accommodate the larger swings. Most books on Bollinger bands (for instance, Kirkpatrick, 2006) claim that 68 percent of the values should fall within one standard deviation of the average, about 96 percent within two standard deviations, and that virtually all prices should fall within three standard deviations.
Though this claim has been repeated in book after book, simple empirical observation shows that this is not true. First, prices are nowhere near normally distributed, so the rule of thumb for standard deviations does not apply; over a large sample of markets, approximately 88 percent (not 96 percent) of closes are within two Bollinger standard deviations of a 20-period average. Second, standard deviation of price is not a meaningful measure, so the width of the Bollinger bands does not track most accepted measures of volatility well.
Keltner channels were created by Chester Keltner, a grain trader in Chicago, but most traders today use a modification that is actually much closer to the Stoller Average Range Channels (STARC). Regardless of terminology, the advantage of these channels is that they respond to a simpler and more consistent measure of volatility: the range of each bar. Also, since these channels use true range rather than simple range, they can also be applied to data that are close-only (e.g., economic data, funds, or some calculated indexes).
They are simple, robust, and consistent: big bars = more volatility = wider bands, avoiding many of the potential issues with Bollinger bands. Figures 7.6 and 7.7 show Bollinger bands and Keltner channels applied to the same weekly chart of Wheat futures. Notice the typical “Bollinger balloon” effect of a large price change on the Bollinger bands; the Keltner channels also respond to the increasing volatility, but they do it in a much more measured and controlled way. As volatility contracts, Bollinger bands tend to tighten very aggressively, and so they will give alerts on small price movements out of these environments.
This may or may not be desirable behavior, and traders who use these bands must know how to adjust for these quirks. Whichever type of channel is used, the critical behavior occurs when the price bars engage the bands. It is important, then, that the bands be set at a level that is meaningful—too close and they will give insignificant signals, but too far and they are never touched.
The Keltner channels that I use are set at a level that contains between 85 and 90 percent of all trading activity (the range of the bar, not just the close) across a wide range of markets. Table 7.1 verifies this claim empirically with an analysis of every bar from a large test universe. For each of the 2,403,774 bars in that universe, two statistics were recorded: “Inside” measures the percentage of the total range of all bars that was inside the bands. “Free bars” are bars that are pushed to such a dramatic extreme that the low of the bar is above the upper band, or the high of the bar is below the bottom channel—the entire bar is outside the channel. These are rare, but potentially significant, events that can indicate a condition of extreme imbalance in the market.
