Keltner Bands Trading Strategy Explained (Keltner Channels Systems)
Last Updated on October 22, 2022
Keltner Bands Trading Strategy (Keltner Channels) is a lesser-known sibling of the Bollinger Bands which are widely used and pretty famous. Despite being rather unknown, Keltner Bands are still included in most software packages.
In this article, we take a look at Keltner Bands to see what it is, how they are calculated, how they differ from Bollinger Bands, and if it’s possible to make money on Keltner Bands. We test a Keltner Channel trading strategy.
Keltner Bands explained: What are Keltner Bands/Channels?
Despite being not widely known, the Keltner Bands were developed before John Bollinger made the more famous Bollinger Bands. The Keltner Bands were first introduced in the 1960s and are renamed after its innovator, Chester Keltner.
Both bands are based on volatility and based on three separate lines (see below for explanation). The middle line sets the basis for the upper and lower bands that are based on the ATR. The bands expand and contract as volatility (measured by ATR) goes up and down.
Keltner Bands are also called Keltner Channels. The reason for the latter is simple: They are made up of three bands, just like in the chart below:
As you can see, most of the price action is inside the bands.
How are Keltner Bands calculated? The Keltner Channel formula
There have been many variations of the bands and channels over the years, but as of now the bands are calculated this way (to our knowledge):
- In the middle of the bands (or in the middle of the channel) we have a line showing the “typical” price or “normal” price. The “typical price” is the high, low, and close divided by three. This deviates from the Bollinger Bands that only use the closing price.
- Based on the “typical” price in number one, two bands above and under (upper and lower bands) are calculated by adding and subtracting the Average True Range (ATR).
- Both numbers one and two are based on x number of days (the period).
(Further below we have included how the formula is calculated in Amibroker.)
What is the difference between Bollinger Bands and Keltner channels?
The chart below shows both a 15-day Keltner Bands and Bollinger Bands:
The grey shaded is the 15-day Bollinger Bands using two standard deviations, while the orange lines are the 15-day Keltner Bands added and deducted two ATRs.
As we can see, the Bollinger Bands are more responsive to changes and contract and expand much more than the Keltner Bands.
Why Are Bollinger Bands more responsive? Most likely because they use the closing price as “the base”, while Keltner Bands use three daily observations as the base.
The Amibroker code for Keltner Bands/Keltner Channels
Keltner Bands are included in Amibroker, but to better understand the calculations we believe the code explains better than plain English what the bands are all about:
Period = 15;
TypicalPrice = MA((H+L+C) / 3,period);
UpperBand = TypicalPrice + 2*(ATR(period));
LowerBand = TypicalPrice – 2*(ATR(period));
Plot(TypicalPrice,”TypicalPrice”,colorRed,styleLine);
Plot(UpperBand,”UpperBand”,colorRed,styleLine);
Plot(LowerBand,”LowerBand”,colorRed,styleLine);
You can of course change the period to any period you prefer.
What is the best setting for Keltner Channel?
It is, of course, no exact setting that works best for the Keltner Bands. That’s because every asset has its own characteristics and moves in different ways. There is no reason to expect a commodity to move in the same way as Facebook, for example.
Further below in this article, we test different quantified rules for Keltner Bands.
What is the Keltner channel used for? How does it work?
Keltner Bands and other volatility channels can be used in different ways.
You can use it as a mean reversion by buying when the price breaks below the bottom band, or you can do the opposite and go long when the price breaks above the upper band.
There is no right or wrong. You just need to find what makes sense and test if it works.
What is the Keltner Channel strategy?
We did a search on the internet and we found very few quantified tests based on the Keltner Bands. We decided to perform our own:
Keltner Bands strategies (quantified):
We have never looked into the Keltner Bands until we came across the bands just a couple of weeks back. However, we briefly tested some ideas:
Keltner Bands as mean reversion:
Criteria:
- The close closes below the lower band. Enter at the close.
- Sell when the close is above the “typical price”. Exit at the close.
- Optimize by using a period from 2 to 25 days.
- We optimize by using different ATRs from 1 to 2.
The best results are clustered around 6-10 days (periods) and an ATR from 1 to 1.5.
If we test by using six days and 1.3 ATR we get this equity curve on the S&P 500:
There are 205 trades, a win ratio of 80%, a CAGR of 6.14%, the time spent in the market is 12.6% and the profit factor is 2.08. However, the strategy has performed poorly from 2016 with basically no returns.
It seems to work slightly better on Nasdaq (QQQ).
Keltner Bands as momentum indicator:
Let’s flip some of the rules:
- The close is above the upper band. Enter at the close.
- Sell when the close is below the “typical price”. Exit at the close.
- Optimize by using a period from 2 to 50 days.
- We optimize by using different ATRs from 1 to 2.
We get the best results by using long periods: around 30 days produces the best results with an ATR of around one to 1.3.
If we use 30 days and an ATR of 1.1 we get this equity curve on the S&P 500 (SPY):
It’s 158 trades and a CAGR of 4.7%.
Does Keltner momentum work better on other ETFs?
We tested on GLD (gold) which produced about 1% per trade but not something we would like to trade. However, one interesting feature was its profit distribution:
Many losers and a low win ratio, but the few big winners offset the losers. This is a typical trend-following distribution, the opposite of a mean-reversion distribution with a lot of winners but the occasional big loser. Trading is about trade-offs!
Conclusion about the Keltner Bands:
Keltner Bands are one of the many volatility bands, but unfortunately, we found the bands not very useful. However, we didn’t spend much time studying the bands, and some readers might offer better ideas to test.