This question originally appeared on Quora, the knowledge-sharing network where compelling questions are answered by people with unique insights. You can follow Quora on Twitter, Facebook, and Google+.

MORE FROM QUORA: What does the analytics team at ESPN do?

Answer by Brian Burke, ESPN senior analytics specialist and creator of advanced football analytics:

QBR is far from perfect, but it’s light years ahead of what came before it. I didn’t invent QBR, but I innovated most of the concepts it’s based on.

The traditional QB rating is a statistical abomination. It double counts some things, ignores others. The weights that go into the formula are completely arbitrary. It’s a disaster. Imagine rating cars with a formula like 1.7 * MPG + 3.14 * max speed + max speed / # of seatbelts. The critics of QBR often blindly accept goofball stats like that, but think ESPN has some kind of agenda behind its own stat.

MORE FROM QUORA: Are analytics more important to some sports than others?

Here’s how QBR works: It measures the impact of every play a QB is centrally involved in using Expected Points. EP measures the likely impact on the net score (the lead/deficit). A 10-yard gain on first-and-10 from the 50 historically impacts the score by an average of 0.6 points, so the Expected Points Added of that play is +0.6. It’s a bit more complicated than that, but that’s the idea. We can measure EP for any combination of down, distance and field position.

QBR then apportions credit for each play according to several factors. A short pass where the wide receiver runs for a lot of yards after catch (YAC) gives less credit to the QB than a deep pass. QBs get more credit for plays when there are more pass rushers. Turnovers, runs and scrambles are considered as well.

“Trash time” is also considered. Plays that occur when the game is still on the line are fully weighted, but plays that occur when the game is largely decided are given little weight.

Finally, the result is turned into a per-play “rate” stat. It is then presented as normalized on a 0-to-100 scale, where 50 is average and 100 would be near-perfect.

So it does a lot of great things that no other stat can do:

Accounts for all QB contributions. Accounts for drive situation (down, distance, yard line). Apportions credit based on difficulty of the play. Accounts for game situation (time and score). It’s easy to see who’s above or below average. It’s a rate statistic rather than a total volume statistic.

The one criticism that I won’t disagree with is that it’s complex. But so is quarterbacking. If we want to consider all the variables inherent in QB performance, we’re going to end up with a fairly complicated and opaque metric. It’s an unavoidable trade-off.