I'd prefer to use stats that are explainable. Passer rating may be arbitrary and flawed, but at least we know how it's calculated. ANYA is a great stat- it's simple to calculate and bases it's formula off of studies that determined the actual value of an interception. I see any stat that we can't calculate ourselves as being a "mystery stat."
Football outsiders and Advanced NFL stats have a lot of mystery stats, but at least they pass the eyeball test. Seattle's high scores on DVOA and individual scores for DYAR make sense, there are no results that feel wildly at odds with reality. Same thing with Advanced NFL stats individual metrics like success rate.
But QBR almost never seems to pass the eyeball test. If you watched this game, you knew that Wilson played very well. He bought time, he made perfect throws when he had to, and he was smart with the football- most of his incompletions were throw-aways. His 133 passer rating was one of the highest of the week, but his QBR was under 50, saying that he had a below average game overall. QBR might take fumbles into consideration, but if it doesn't and still thinks that, then holy crap what a worthless stat.
Just as bad is Luck's 93.7 QBR, meaning a 93.7th percentile performance, on Thursday. His line? 18/26, 227 yards, 0 TD, 1 INT, 80.1 rating. It's a high YPA performance and probably under-rated by his QB rating, but there is just no way that he had a 94th percentile performance. If Indy was 75% on 3rd downs or something that might explain things, but they were 3/11, which is below average. Really, I can't figure it out, and it makes me think their system is too flawed to take seriously.
In fairness, FO pretty much agrees with QBR with Luck v. Wilson. Luck has a big lead in DYAR and a slim lead in DVOA. The lead in DYAR makes sense- since DYAR is a counting stat- more reps = more opportunity for DYAR. The DVOA number does surprise me though. Wilson only rates +1.2%, meaning that they see him as almost dead average. That's pretty close to QBR which has him at 56.9/100 for the season, with 50 being average.
It should be noted though that play tracking is very much a subjective science. With QBR they have to make decisions about who gets the blame for a drop and also awards extra points for an undefined "clutch factor." It's not hard at all to see bias coming into play with those numbers.