A committee of experts — well-regarded mathematicians and infectious disease experts — appointed by the Department of Science and Technology to use mathematical modelling and forecast the course of the pandemic has brought good tidings. By their estimate, India passed its COVID-19 peak in September and the decline in the overall caseload being observed for nearly a month now is to continue. Active cases, about 7.5 lakh now, are expected to drop below 50,000 by December, and by February, the pandemic is likely to extinguish itself with only 'minimal' (not zero) infections. While it is reasonable to assume that the seven-member committee has been scrupulous, the caveat is that this is still a forecast based on mathematical modelling. There are some strong assumptions. The decline will continue only if there are no major mutations during winter, protective antibodies are durable, and current restrictions are maintained. There would be no significant gains from a strict lockdown beyond the district level, and current containment measures would suffice, except if there are local outbreaks that threaten to overwhelm health-care facilities there. Their calculation also showed a peak by July latest, with anything from six to 15 times the existing infections had there been no lockdown or if it had been delayed to April. The purpose of pandemic modelling is to generate a probabilistic overview of the future and mathematical modelling has become a popular, creative exercise, with several models and forecasts being made available on pre-print servers and pending peer-review. The latest model is expected to be published in the Indian Journal of Medical Research this week, but it appears to be a quotidian exercise. The datasets it has relied on are publicly available and the modelling employs a category of models called SEIR that estimates, within a population, those Susceptible, Exposed, Infected and Recovered. It is extremely dependent on the quality of data that is used as an input and relies as much on simplifying assumptions that sacrifice complexity for comprehension but there is nothing to suggest, from what is known about the exercise, that it is more likely to be true than similar estimates from scores of models the world over that subscribe to a certain degree of rigour. Experts associated with the pandemic have reiterated many times that mathematical modelling ought not to be taken literally. The latest assessment too should then be used not to critique or justify past decisions but dwell more on the future. For instance, if the model suggests that the pandemic would extinguish by February with a dramatic dip by December, then should the accelerated clinical trials of potential vaccines be top priority? Mathematical models, to be useful, must induce policy or behavioural change to avoid their own worst-case scenarios and this latest assessment must be seen — no more, no less — as a tool to this end.