The past looks safest just before the break.
Taleb's turkey illustration highlights the danger of inferring safety from repeated benign observations while missing the process generating them.
Inside this model
Daily outcome is +1 before the chosen break and −100 from the break onward. A deliberately naive next-day forecast uses (positive days + 1)/(observed days + 2), as if days were exchangeable Bernoulli trials. The plotted history never shows days beyond the reveal control.
Out in the world
A practical use
A supplier can deliver reliably until a single unmodeled dependency fails. Studying that dependency may matter more than extending a streak chart.
A useful lens. Not a universal law.
- The break is set by you, not predicted. This deterministic story does not estimate real rare-event probabilities or show that Bayesian reasoning itself is invalid; it exposes a misspecified stationary model.
Associated thinkers
Further reading
Explore the original research or the teaching reference behind this experiment.
Taleb and Blyth — The Black Swan of Cairo ↗