How to make the overbought VIX trade that much better
Our long national CBOE Volatility Index (VIX) nightmare is over: We're overbought again! By overbought, I mean VIX closed 25.6% above its 10-day simple moving average (SMA). I use >20% over the SMA as a threshold. Just to refresh/update, here's how my chart of overbought VIXes since 2009 looks:

Sept. 28 was actually the best buy signal in this series, up a whopping 9.93% in a month! Furthermore, the SPDR S&P 500 ETF Trust (SPY) bottomed on Day 1.
Fading overbought VIX continues to do well. Buying and holding for one month has a median return of 3.33% vs. a randomly timed median return of 1.52%. It's mostly worked, as you can see, but we're just off a pretty bad one in August -- which I hope highlights that this idea, like most trading ideas, will sometimes fail badly. I'm going to take a wild guess that when we look back on longs initiated Friday, they will return something in between the last two instances.
I cut a lot of studies off in 2009, mainly because 2008 numbers look an awful lot like outliers, and thus tend to distort findings. But there's always a danger in picking endpoints. It doesn't change an awful lot in this case, though. If I go all the way back to the beginning of SPY time in 1993, the one-month hold strategy shows a median return of 2.16% vs. a median one-month return of 1.12% on randomly timed one-month holds.
I did notice something else that was interesting. VIX vs. SMA as a signal works best when VIX is relatively modestly overbought. The correlation of VIX vs. 10-day and future one-month SPY returns is near zero since 1993. If we only look at days where VIX closed 20% above the 10-day, it's still near zero. But if I look at days where VIX closed >15% above the 10-day, but <27% above, the correlation jumps to 16.6. That's not terribly high, but it does suggest that there's a sweet spot of overbought in there that's worth looking further into.
The flaw, of course, is that overbought in that range can expand to severe overbought. That suggests the best idea is to fade (go long) with a lower threshold than 20% and then use stops. We'll take a look in the not-too-distant future.
Disclaimer: Mr. Warner's opinions expressed above do not necessarily represent the views of Schaeffer's Investment Research.