Well, that is sort of the nature of a real experiment, no? To actually learn things you do not already know.
But this specific subject is covered at great length in Podcast Episode 40, just released. They go into how they saw the nature of the experiment, what their data and analysis showed about the academic performance of recent admits, what happened to applications and who chose to submit test scores, and so on. Quinlan is the guest on that Episode, and he repeats those basic findings, but there is much, much more conversation on the subject.
Not surprisingly, the emphasis is really similar to what Dartmouth came out with, as we discussed above. The correlation between test scores–including apparently APs–and grades is relevant background, but the real problem they are trying to solve is underreporting of scores that would actually help specific applicants. Indeed, Quinlan explains for most of their applicants/admits, they were actually happy with the recent results. But there are specific groups they thought they were underadmitting relative to highly-resourced students, because they were not submitting what would have been helpful scores (SAT/ACT and AP), and they are now convinced they cannot communicate the right message to those people within a test optional policy.