No, it is not possible to estimate the size of a compressed version of a file based purely on its character count. Different strings can be compressed at different levels of efficiency; a string made purely of one character will be much more easily compressed than a string of purely randomly generated characters.
In information theory, there is a concept of Kolmogorov complexity, which is (more or less) the smallest amount of information necessary to reconstruct a string. Not all strings an be compressed into smaller strings, and it is impossible to build a general algorithm to find the Kolmogorov complexity of an arbitrary string. Moreover, it's impossible to prove that you have found the optimal encoding for a string once the string ets sufficiently long.
Hope this helps!