The long recorded history of earthquakes in Japan affords an opportunity to forecast seismic shaking exclusively from past shaking. We calculate the time-averaged (Poisson) probability of severe shaking by using more than 10,000 intensity observations recorded since AD 1600 in a 350 km-wide box centered on Tokyo. Unlike other hazard-assessment methods, source and site effects are included without modeling, and we do not need to know the size or location of any earthquake nor the location and slip rate of any fault. The two key assumptions are that the slope of the observed frequency-intensity relation at every site is the same, and that the 400-year record is long enough to encompass the full range of seismic behavior. Tests we conduct here suggest that both assumptions are sound. The resulting 30-year probability of IJMA ≥ 6 shaking (∼ PGA ≥ 0.4 g or MMI ≥ IX) is 30%-40% in Tokyo, Kawasaki, and Yokohama, and 10% 15% in Chiba and Tsukuba. This result means that there is a 30% chance that 4 million people will be subjected to IJMA ≥ 6 shaking during an average 30-year period. We also produce exceedance maps of PGA for building-code regulations, and calculate short-term hazard associated with a hypothetical catastrophe bond. Our results resemble an independent assessment developed from conventional seismic hazard analysis for greater Tokyo.