It looks like it warned you before every big spike. Twenty-two years of data shows that is the base rate fooling you, not a signal.
It is one of the most shared chart setups in volatility trading. You overlay MACD on the VIX, wait for the MACD line to cross up through zero (the moment VIX's 12-day average climbs above its 26-day average), and you call it an early warning: volatility momentum has turned, a spike is coming. The screenshots look devastating. There is a clean cross a couple weeks before every major blow-up.
So we tested it properly. We took every VIX MACD zero-cross-up since 2004, all 138 of them, and measured what the VIX and the S&P 500 actually did over the next 10, 20, and 40 trading days. Then we compared those outcomes to a plain baseline: what happens after a randomly chosen day. If the cross carries information, the after-cross numbers should beat the baseline. They do not.
| Forward window | VIX peak after cross | VIX peak, any day | 5%+ drawdown after cross | 5%+ drawdown, any day |
|---|---|---|---|---|
| 10 days | +14% | +15% | 11% | 11% |
| 20 days | +26% | +23% | 22% | 20% |
| 40 days | +35% | +34% | 31% | 31% |
Read across any row. The cross adds a point or two at best, well inside the noise, and by 40 days it is a dead heat (it is actually a hair worse on the broader drawdown counts). Conditioning on the signal tells you almost nothing you did not already know from the calendar. As a forward edge, it is not there.
This is the part that fools everyone, so it is worth slowing down. Yes, a zero-cross fired before each of the famous unwinds: 14 trading days before Volmageddon in February 2018, 19 days before the COVID crash, 13 days before the August 2024 carry unwind. Three for three. Uncanny.
Now count how often the signal fires at all: 138 times in 22 years, roughly once every two months. When something fires that often, there is almost always a recent one sitting just before any event you care to point at. The cross did not select the spikes. It fired constantly, and the spikes happened to follow a handful of them. The other roughly 130 crosses fired into nothing.
This is survivorship bias wearing a disguise. Look only at the crosses that preceded a disaster and you see a genius indicator. Look at all of them and you see a metronome that ticks through calm and chaos alike. The big spikes are in the "after-cross" bucket only because nearly every day is.
Same honest setup we use on every signal: define it, count every instance, compare to baseline, no cherry-picking.
Forward VIX gains and S&P drawdowns after a cross land within a point or two of a random day at every horizon tested. The signal does not move your odds.
138 crosses in 22 years means there is almost always a recent one before any event. "It called the big one" is what you would expect from a signal that calls nearly everything.
By the time the MACD line crosses zero, VIX has already risen a median 16% over the prior two weeks. The move you wanted to catch is mostly behind you.
MACD is the gap between two moving averages of VIX's own past prices. Smoothing delays by construction, so a leading warning out of it is not possible in principle.
The reason is baked into the indicator. MACD is the difference between two exponential moving averages of VIX's own price (the 12-day and the 26-day). Moving averages exist to smooth and therefore to delay. For the shorter average to climb above the longer one, VIX has to have been rising for a while already. By the time the cross prints, you are not early. You are confirming a trend that is well underway.
Our sample makes that concrete: in the 10 trading days before the cross, VIX had already gained a median 16%. You can tune the parameters all you like (a signal-line cross instead of a zero-line cross, faster or slower lengths), but you cannot tune away the lag. It is the price of using moving averages of a number that has already moved.
No black box, and nothing you cannot reproduce.
This was never really about MACD. It is about a whole family of signals.
We ran the identical test on two other popular reads: a sharp one-day jump in VIX's RSI, and speculator positioning in VIX futures. Both came up with no usable edge, for the same reason the MACD cross does.
Every one of these is a transform of VIX's own past price. By the time the indicator changes, the move is already in the tape. You cannot squeeze a leading signal out of a lagging construction, no matter how you dress it up.
The things that actually move before VIX are different in kind: the shape of the VIX futures curve, crowded positioning that has to unwind, cross-asset stress. Those at least have a chance, because they are not just yesterday's VIX in a new coat.
The shape of the VIX futures curve moves with positioning and forward expectations, not just yesterday's spot. See it live in the term structure tool.
Open the VIX Term Structure
Comments
Loading comments…