One problem of particular concern for many budding statisticians is the difference in the way Student's t-distribution and the Normal distribution are handled.

One problem of particular concern for many budding statisticians is the difference in the way Student's t-distribution and the Normal distribution are handled. The Excel functions for the t-distribution split the probability into two tails, but the functions for the Normal distribution did not. This made it quite a pain when doing some calculations with both large and small sample sizes. Newer versions of Excel have "fixed" this by adding a flag to the t-distribution functions that let you choose whether you wanted to use single tailed probability or two tailed probability. But this flag has not appeared on the functions for the normal distribution and thus the functions for these two similar distributions are still inconsistent. Another annoying issue was the autocorrelation in the random number generator (a very bad thing for a random number generator). The main rand() function is now corrected but the problem lives on in the randbetween() function, which shows that these two related functions are using separate RNGs instead of them both tapping into a single good RNG. It's very frustrating if you want to do any real work in Excel because you can't trust the built in functions.