This has been the problem with nearly every model used by the experts so far. The models originally predicted 2.2 million US deaths based on the best assumptions they could come up with. Flash forward a couple weeks after seeing what happened in other countries they changed it to 100K - 200K deaths. Flash forward another week of observation and they changed it to 80K deaths. All the assumptions used in the models have been proved incorrect at each step. Meanwhile those that are strictly observing trends with other countries and matching ours to those countries, like CSoul, are seeing much more promising results that have remained consistent and lower than the predictive models.