A fascinating 2016 paper by Cirillo and Taleb on predicting major conflicts remains deeply relevant today. Their key insight? We consistently fool ourselves about the predictability of extreme events.

Some sobering findings:

• For conflicts with >10M casualties, average waiting time is 136 years
• The “Long Peace” of 70 years isn’t statistically significant
• Historical data actually underestimates true violence by at least half
• Under fat-tailed distributions, the mean is dominated by extreme events

Most striking? They quote an 1860 assessment celebrating an “unprecedented” 40-year peace and declining warfare… right before the bloodiest century in human history.

With current tensions in Europe, Middle East, US-China relations, and emerging climate pressures, are those saying “Humanity has changed! This time is different!” the same mistake of extrapolating recent stability?

Statistical humility matters: For fat-tailed events, we need much more data to claim peace than to claim risk.