My experience with the expected wait time calculation is that it is typically inaccurate and can have large seings when dealing with a samm number of agents and/or calls. I have not seen any official caluclation for EWT, but I believe it is something like:
Calls In Queue * Average Handle Time / Logged In Agents
Since the AHT calculation can have large swings when the calculation is based upon small sample sizes, the calculation is typically unreliable. I prefer to use position in queue rather than time based indicators for callers when I know the sample size is going to be small.