Since I tested Bollinger Bands , what’s natural to think of testing next?
Rubber bands? Rock bands?
No, you silly goose! Keltner Channels…
What’s the Difference Between Keltner Channels and Bollinger Bands?
Bollinger Bands are based on chunks (called “standard deviations”) away from an average.
Keltner Channel bands are based on multiples of the average true range (ATR) away from the average.
Alrighty then. Let’s get into our test setup…
The Trading Truth Test Setup
We’re keeping this the same as our Bollinger Bands test, except for the ATR period.
Since we’re using ATR to determine the Keltner Channels, we’re going with the same period as the moving average. Market: the S&P 500 index (using SPY to trade it, assuming SPY is exactly 1/10th the S&P 500 Index price)
Timeframe: Jan 2, 2008 to March 28, 2023
Bar interval: 1 hour
Moving averages: 50 bars (simple moving averages, meaning every bar gets equal weight, unlike with exponential)
Average true range: 50 bars
Starting Equity: $ 25,000
Max % of Equity Per Trade: 3%
Commissions, fees and taxes. To keep things super simple, we’re assuming these are all zero.
Our 2 Tests
Test A:
We’re using Keltner Channel bands 2 ATRs away from the moving average.
Any time a high pierces the upper Keltner band and then a high is below the band, go short (if we’re not already in a trade).
Any time a low pierces the lower Keltner band and then a low is above the band, go short (if we’re not already in a trade).
This way, we’re waiting for a cross back over the band after it gets pierced.
We didn’t do this for our Bollinger Bands tests, which makes this not a direct performance comparison.
Test B:
The same as Test A, except we’ll use Keltner bands 3 ATRs away from the MA instead of 2.
The Test Results
Test A's equity ended at $ 38,137.63, up 52.6%. The biggest loss from the initial $ 25,000 deposit was $ 843.16, a 3.4% loss. The maximum losing streak was $ 1,805.75 or 6.36%.
Test B's equity ended with $ 34,435.71, up 37.7%. The biggest loss from the initial $ 25,000 deposit was $ 3,027.42, a 12.1% loss. The maximum losing streak was $ 2,080.71 or 8.22%.
The worse results 3 ATRs away from the MA is surprising. Seems like that’d give us higher-quality trades.
Even Test A’s numbers don’t come close to what plain-Jane buy and hold did: ending up 173.1%.
Note: I did this analysis in a spreadsheet, with exported TradingView data. If you see any errors, please let me know.
What Test Tweaks We Could Make
Some traders wait for when Bollinger Bands or Keltner Channels get pinched (narrower in width). That’d be interesting to test vs bounces.
One well-known trader, John Carter, looks for when Bollinger Bands go inside (“squeeze into”) Keltner Channels. He sees that as an indicator of bigger-than-normal moves.
What would you test? And what else would you like to see tested?
@AlmadeCubano, haha looks like it, if following this strategy. That said, it didn't beat buy and hold, so I'd keep working on the strategy before trading it.
DaddySawbucks
⋅
IDK how you get 173% for buy and hold for this period SPY ended 28 Mar only $10 above Jan 02 price, 388 > 398 ~2.5%
TradingTruthseeker
⋅
Hi @DaddySawbucks! Thanks for your comment. The test period started on Jan 2, 2008.
DaddySawbucks
⋅
@TradingTruthseeker, ahhha how our minds play tricks I saw the date but did not see it! yes ofc B&H best strat in long bull, 15 years was a good run.
TradingTruthseeker
⋅
Haha all good, @DaddySawbucks. And yeah, buy and hold beats many strategies over that time!