Bubbles are the most over-narrated event in finance. After the fact there is always a tidy story: the valuations were absurd, the enthusiasm was irrational, anyone paying attention could see it. The story is satisfying and mostly useless, because it is assembled from information that did not exist at the time.
What follows is an attempt to describe the shape of a bubble using real numbers rather than the narrative — and then to be honest about why recognising the shape is much harder than it looks.
Almost every major bubble starts with something real. The internet did reshape commerce. Remote work did arrive suddenly in 2020. Bitcoin did represent a novel idea about money. Bubbles are rarely built on nothing; they are built on something true, extrapolated too far.
At some point the asset detaches from whatever justified it. Buyers arrive because the price is rising, and the price rises because buyers are arriving. The original thesis is still quoted, but it is no longer what's setting the price.
Tops are not announced. They are simply the last day more money arrived than left. At the peak, sentiment is near-uniformly positive — that is why it's the peak, because everyone inclined to buy already has.
Declines are usually faster than the ascent, because leveraged holders are forced to sell into falling prices, and because the marginal buyer disappears exactly when needed most.
Computed from the daily closing series embedded in this site's Mystery mode:
| Episode | Run-up to peak | Peak-to-trough |
|---|---|---|
| Bitcoin, 2017–18 | +1,704.5% | −64.3% |
| GameStop, 2021 | +1,998.6% | −88.3% |
| Zoom, 2020 | +548.3% | −18.9%* |
| Amazon, 2000–01 | +51.7% | −86.9% |
| Peloton, 2021–22 | +14.7% | −87.9% |
| Netflix, 2011 | +10.3% | −78.7% |
*Zoom's decline within this particular window only; its larger fall came later.
Two things stand out. First, the scale of the collapses: four of these six lost more than three quarters of their value. Second, and more instructive, the run-up column is wildly inconsistent. Bitcoin and GameStop went vertical. Amazon, Peloton and Netflix barely rose at all within their windows before falling apart — because the mania had already happened before the window opened.
This is the practical problem in one table. There is no characteristic pre-crash pattern. Sometimes the top is a spike; sometimes it is a quiet drift that simply stops.
Consider Amazon. It fell roughly 87% in the dot-com bust, and at the time serious people questioned whether it would survive at all. With hindsight, "buy Amazon in 2001" is the most obvious trade in history.
But live, an investor holding through that decline had no way to distinguish Amazon from the hundreds of dot-coms that fell just as hard and never came back. The information that makes the decision obvious — that this particular company would become one of the most valuable on earth — is precisely the information that did not exist yet.
The same applies to Peloton and Zoom. Both were rational responses to a genuine shift in how people lived. The error wasn't believing the shift was real — it was real. The error was assuming it would continue at the same rate indefinitely.
No reliable method exists for calling tops. What can be done is preparing so that being wrong isn't terminal:
The trouble with reading about bubbles is that you read about them knowing they were bubbles. That knowledge cannot be unlearned, which makes self-assessment nearly impossible.
Mystery mode exists for this reason. It plays a real historical chart day by day with the company's name hidden and the price rebased so the raw level cannot give it away. You are inside one of these episodes with no idea which, no hindsight, and a decision to make on each new day.
Most people discover they hold too long into declines and sell too early in recoveries — the same two errors that make bubbles work. Finding that out costs nothing here.
Play a mystery chart →