Dot-com fractal v5

In depth

This model asks one question: how closely does today's market resemble historical bubble peaks and crashes? It encodes the market into a multi-dimensional fingerprint — valuation, breadth, concentration, momentum, cross-asset structure — finds its nearest historical neighbors across three decades of data, and reads what happened next. The current reading is stark: today's closest analogs are among the most concentrated, most expensive late-cycle environments on record, with the index's top-handful concentration at the most extreme level the model has ever measured. Crucially, it is a risk lens, not a timing tool — it does not say 'sell,' it says the distribution of forward outcomes is unusually fat-tailed. It informs the operator's discretionary risk posture and never trades on its own. The model has been through five honest iterations (v1 through v5); a sixth was built, failed its robustness check, and was archived — the version you see live is the one that survived.