Time Average Bands: Qualifying Breakouts with BullWhip and BearCage

Studies with Bands applied have been around for a long time, and studies such as Bollinger and Keltner have been a frequent topic that I have been asked to either comment on or build strategies around. Though they are designed primarily as a means to identify breakouts, with some caveats and additional code it is possible to build mean reverting methods as well. It is in this area that I prefer their application, which is conductive to any market in any timeframe. My own preferred study, in the traditional sense, is a Moving Linear Regression Line with Standard Errors computing the Bands. Moving Linear Regression has an advantage in that it tracks a trend, however steep or shallow it is, and therefore makes the Standard Error extreme move relevant.

It is as a breakout methodology that I have had problems, with especially in my core area of Futures and FX. There are simply too many false breaks. This can also be a problem with stocks, albeit to a lesser extent.

Last year’s sideways trend in Microsoft (MSFT) highlights the issue. Bollingers, in their default setting, produced four false breaks before signaling a correct one.

Percentage Envelopes around a Moving Average produced three.

Changing this to Smoothed Envelopes made no difference.

Keltner uses a filter with Average True Range, which should be more dynamic to price action. It prevented false breakouts, but, when the breakout did come, it was late in signaling it.

Time Average Bands are my attempt to improve the concept of not only what constitutes a valid breakout, but also a false one. That is calculated by the studies BullWhip and BearCage. A major issue I have always had with traditional studies is the fixation that the value is calculated based on the close. This means that it is far more difficult to be proactive and, in times of big moves and volatility, waiting for the close can seem like a lifetime. Most of my studies are based on the Opening, meaning the value is fixed and allow time for impartial analysis. Time Average Bands are no different. For me, the relationship between close to close is a limited one. A lot can happen and the closes could be identical. It therefore makes sense to look at the relationship between range to range. Another issue in the traditional sense is that, if the indicator values don’t produce good breakouts, the temptation is to adjust the Average value until they do. The danger here is that you are curve-fitting the Indicator to the data. 

Time Average Bands compensate for this problem by having a variable Moving Average element that will become more sensitive if range expands and less so if it contracts. The lowest value is a 3-period and the highest is 21. The other contrast is that it places up to 4 standard deviations around price, which, in addition to qualifying breakouts, allows for an interpretation of price in relationship to those Bands. The chart below shows how price was contained within the Bands in sideways before correctly signaling the breakout. I have set it to 20 days to match the other studies, although the default is my preferred number of 22, which connects with the settings on Range Deviation Pivots and Volatility Time Bands.

In truth, the study is ugly, made more so by the fact that it is important to connect Breakouts with Support and Resistance via Energy and Expansion. In order to make it easier to visualise when false breaks could be happening, BullWhip and BearCage mark bars when the following criteria is met. For BullWhip, price must have closed below the 4th deviation down on the previous bar, then closed back within the bands on the current one. It’s the opposite for BearCage and will mark the bar with Green or Red dots, respectively. The chart below of Micron (MU) shows an example where this occurred directly within a major support area of both Energy and Expansion, thus indicating a false break.

Returning to Microsoft, we can see how important that process of qualification is. The reality is that, in its raw form, potential false breakouts can happen at its most frequent. This BearCage signal is false, as it is occurring directly into Expansion-based support (circled) and therefore would be ignored unless price made a daily close below that level.

In order to further limit occasions when the Pattern occurs, there are two extra variables within: Trend Confirmation and Expansion. Both use proprietary code, Expansion being the most important; when activated, the BearCage signal disappears.

BullWhip and BearCage also have uses if you wish to be be more active in managing positions. The Apple (AAPL) chart below shows a BearCage signal at the final Short-Term Expansion and Energy-based resistance. Price retreats to its first logical area, which is the confluence of levels circled. A Hammer Doji at support and use of the Time Volume study qualifies the support point as a place to re-enter, as volume is over 150% of what its average is for the time of day.

I can be contacted at shaun.downey@aol.com to answer any questions.

Shaun Downey