I believe that when Amir places an audio device on his test bench, he goes into "lab tech" mode, where sound only exists to confirm the perfection of his measurements.
Seriously, improvement is a hard thing to quantify. If everybody agrees A > B, life is easy. But what if you're the only one who thinks so?
In theory, this is a great advantage for the objectivist. Instead of days of pain-staking listening, they spend a few hours running a battery of tests. But, wait - how do you evaluate the results of each test? How do the tests correlate to the SQ characteristics most important to you?
Some people need that kind of assurance, so ASR definitely serves a purpose. However, if you're trying to decide if you'll like the SQ of a specific component in your particular setup playing your kind of music, I doubt that ASR will be much help.