When calculating the annual return of an investment (as (1 + return in %)^(365/days)-1) over a very short time period - say 2 days - my annual return explodes due to the exponential term, which I think can later on when calculating average annual returns over few investments lead to a huge bias in the average return as the average is sensitive to outliers. This might be a bit misleading. If I would use a normal linear transformation to calculate the net annual return ((return in %*days)/365), the annual return would not explode as much for a short time interval. What is the usual way to handle these cases? Still using the exponential transformation or something else? Thanks a lot!