yes you can test a program's impact on system resources etc. ,but you cant possibly test detection.
You want a broad test that gives you the best possible snapshot of a AV's detection performance on a large scale, you don't consider the small amount of samples an end-user comes into contact with any measure of detection. (
at least, thats what I assume)
I'm not disagreeing with that. But so many new infection samples on average
per day, changes the various AV vendors make to their databases....all this and more makes it hard for me to take even these large tests too seriously, because they are periodical (usually quarterly?) and so I mostly discard them as cues to the long-term detection performance of AVs, and just try to focus on what appears to serve me well.
The only test I could think of that would impress me would be something fed real-time, or at least daily, and that seems to me to be a massive undertaking that couldn't get the support it needed to keep running, if it was even possible to start.