The Financial Bubble Experiment (FBE) was a scientific test of a hard question: can financial bubbles be diagnosed before they burst? Rather than studying crashes after the fact, the experiment committed to forecasts in advance, then waited for the market to deliver the verdict. It was run by Prof. Didier Sornette and his group at the original Financial Crisis Observatory at ETH Zurich.
Most market “predictions” are really explanations dressed up after the event, or back-tests tuned on history until they fit. The FBE avoided both traps. Forecasts of which assets were in a bubble, and the window in which a regime change was most likely, were sealed and time-stamped in advance, then released only after the predicted period had passed. This ex-ante, pre-registered design is the gold standard for showing that a method has genuine predictive content, not hindsight. It is the same forward-testing philosophy the FCO still advocates today.
in real-time LPPLS bubble forecasts over five years of live testing
Source: the Financial Bubble Experiment, Financial Crisis Observatory, ETH Zurich. Past performance is not a guarantee of future results. See our disclaimer.
The FBE established the principle; the FCO platform operationalizes it. The same LPPLS model that was tested in the experiment now powers daily bubble scores, multi-timescale confidence indicators, and critical-time forecasts across stocks, indices, commodities, currencies, bonds and crypto, translating bubble and Dragon-King diagnostics into decision-useful signals.
The experiment was pre-registered and documented in a series of public arXiv papers by the Financial Crisis Observatory: