But with several weeks worth of lag. The number of deaths today should be divided by the number of cases N>10 days ago. The number of cases was much more severely underestimated N days ago than today. The death rate in Italy is basically unknown because the denominator is just noise with a time lag.