If we have sample mean of $n$ observations as $\bar{x}$. It is given that maximum value of the whole set is $M$. How many more observations we need so that the estimate for average we have differs from the true average by not more than 10 with 99% confidence?