[hansvm]

That's probably a decent approximation, but it assumes the distribution of latencies for flagging is the same for you and the aggregate "everyone else."