I would have thought the same, but wanted to know for sure.
So I've run each version of the query over and over again, both for short durations and for long. The variability of server responses while everyone else is running queries makes this hard to test. However, when running them simultaneously, they are mostly in a dead heat (~977 seconds over 30 days, ~10-15 seconds over 4 hours).
When running one of them over and over again, and then doing the same with the other, the average performance of the eventstats is slightly better. I think what might be going on is that while eventstats is using more memory during the query, this doesn't cost much in the way of performance. Stats values, on the other hand, might be wasting performance in upkeep while it throws things away.
... View more