Draw a support line under VIX, wait for it to drift down and bounce, right on cue. It looks uncanny. The catch is that VIX bounces from almost everywhere, so a line drawn anywhere looks like it works.
It is one of the most satisfying things you can do to a VIX chart. Sketch a support line under the recent lows, watch VIX drift down toward it, and watch it bounce, seemingly right on the line. Do it a few times and you start to believe the line has power. We wanted to know whether it does.
The trouble with trendlines is that they are hand-drawn and subjective. Two people draw two different lines on the same chart, and both can claim the bounces that follow. The only honest way to test the idea is to mechanize it, turn "draw a line and wait for the bounce" into a rule a computer applies the same way every time, and then ask the one question that matters: does the line add anything? We built three objective versions, a horizontal support off prior lows, a sloped regression-channel line, and a downtrend-line break, and ran them over 22 years.
| Within the next... | VIX bounced 10%+ | Bounced 25%+ | Median VIX gain |
|---|---|---|---|
| 10 trading days | 66% | 30% | +15% |
| 20 trading days | 77% | 46% | +23% |
Read that before you draw a single line. On a randomly chosen day, with no setup at all, VIX rises at least 10% within the next month 77% of the time, and at least 25% 46% of the time. VIX is the most mean-reverting, jumpiest number in markets. It is almost always about to pop by some amount. So a "bounce" is not a rare event you need a line to find. It is the default. Any line you draw will sit just under a bounce that was already coming, and will look, in hindsight, like it called it.
Here is the classic version: a horizontal support at a level VIX has bounced from before. When VIX drifts back down to within a few percent of a prior low, do we get a better bounce than usual?
| Next 20 days | Bounced 10%+ | Bounced 25%+ | Median VIX gain |
|---|---|---|---|
| Any random day (baseline) | 77% | 46% | +23% |
| At horizontal support | 81% | 50% | +25% |
| Any day at the same low VIX (~15) | 80% | 53% | +27% |
At support, VIX bounces a little more than the overall baseline, 81% versus 77%. That is the number people point to when they say the line works. But look at the third row. Support touches only happen when VIX is already low, around 15 against an average of 16.6, and low VIX mean-reverts upward harder no matter what. Compare the support touches to a random day at the same low VIX, and the bounce is just as big, in fact a touch bigger (80%, and 53% for the larger move). The support line adds nothing. The only thing predicting the bounce is that VIX is low, and you can read that straight off the number without drawing anything.
Maybe the horizontal line is too crude, and a proper sloped trendline does better. It does not. We fit a mechanical trend channel and flagged the days VIX touched its lower band, the computer's version of "price hit the rising support line." Those days bounced 10%+ in the next month 73% of the time, below the 77% baseline. And the textbook breakout play, VIX breaking up through a descending trendline while in a downtrend, the "it is finally turning, ride it" setup, did worse still: it bounced less than a random day. Neither sloped version even gets the low-VIX tailwind that flatters the horizontal one.
A trendline on VIX is a line drawn around a bounce that was already on its way. The chart makes it look like the line summoned the move. The data says the move was about 77% likely from anywhere, and the line added nothing the VIX level was not already telling you.
Three objective stand-ins for a hand-drawn line, a level control to isolate the line from the level, and a baseline that turns out to be the whole story.
From any random day, VIX is up 10%+ within a month 77% of the time. A "bounce" is the default, so any line gets credit for a move that was coming anyway.
Support touches bounce slightly more than the overall baseline, but match a random day at the same low VIX. Control for the level, and the line's edge vanishes.
The mechanical sloped-channel touch bounced 73% (vs 77%), and the downtrend-line break bounced less than a random day. The fancier versions are worse, not better.
A hand-drawn line can be nudged to fit any past bounce. That is not prediction, it is curve-fitting after the fact on a series that bounces anyway.
Why a line on a VIX chart feels predictive when it is not.
If a bounce happens 77% of the time anyway, almost any rule will look like it works. The bar a signal has to clear is not "was there a bounce," it is "more bounce than you would have gotten doing nothing."
Support touches happen at low VIX, and low VIX bounces harder on its own. The honest test holds the level fixed. Once we did, the line's apparent edge vanished. Always ask what your setup is quietly selecting for.
The flexibility that makes trendlines feel insightful is exactly what makes them untestable by hand. Pin them down so a computer draws them the same way every time, and the magic turns out to be the mean reversion VIX has anyway.
We test popular signals the honest way: every instance counted, every result measured against a plain baseline. See what else held up, and what did not.
Browse Reality Check
Comments
Loading comments…