As promised last week this kicks off a series of studies that seek to refine and enhance the 6 ETF Gamma model posted Friday. In this instance we'll look at the dynamics between TLT (20 year treasuries and its ultra inverse TBT). TLT is an IShares product while TBT is a Proshares product and its important to keep in mind that TBT is an ultra (x2 short).
We've discussed previously how ETFs and their inverses don't necessarily always work in true inverse fashion and that's certainly the case with these two.
First, here a look at TLT and TBT using the MOSAIC pairs Analytics version of the ETFP platform :
What the analytics reveal are clear pockets of opportunity when the inverse skew between these 2 gets imbalanced. What follows that imbalance has always been a return to parity (the zero line) and I won't even begin to speculate on the factors that periodically create the skew.....the point is simply to recognize it and take advantage of it's profit taking potential. For the purposes of this study I've set the time stop at 8 days.... just because I like to cut my risk exposure short.
What's encouraging about the risk profile is the 81% linearity of the equity curve over a period of 6 months....better odds than you'll get with most trades and the 20 trades over the last 6 months produced a 46% return or an average of about 2% for every 8 days of exposure.
We can look at the TLT / TBT skew another way using the M3 platform:
There is no ghost input for this model because we're just trying to see how these 3 inputs play against each other. In times of market weakness we expect TBT to rank highest, whereas in times of market strength we expect SPY to outrank TLT.
This is a top 1 sort and, as always, we need to play close attention to the short term ALERT, the P6 and the RSQ line. These are the results just using the AUTO STOP turned on....not a barn burner on the 6 month metrics but attractive on the shorter term times frames....and a nice upslope equity curve.
The M3 version doesn't get us close to the pairs modelling returns of 46% over 6 months, but we're also looking at the skew in an entirely different way using an entirely different momentum algorithm and timing framework plus adding SPY to the mix to further confuse things so the differences in returns are not unexpected.
Going forward we'll look at how we might use this correlation of the TLT and TBT (and other divergence pairs) to improve the Gamma model returns.
