Sports Hernia":1qem2hz4 said:
Though I'm sure Santa Clara fans will attempt to pick it apart.
Here I am!!! :lol:
Literally not a single complaint about the conclusion that Wilson is a damn good QB, but three things I noticed about (what I think is a generally good article), in descending order of importance:
1) It's incredibly hard to compare QBs over time for a whole host of reasons. I think it's very, very commendable that the author of the piece is trying to do it, but I don't think he successfully pulled it off.
If you've got 70 years or so of quarterback play and according to your metric 25% of the best QBs of all time are all of the last decade or so, there's a VERY GOOD CHANCE you haven't been able to flatten out across eras.
TBF the data it would take to do this aren't available across eras, and without those data you have to rely on ranks by era (which I *think* is what the author is doing), but what happens if you try to do that with something like DVOA within eras to compare across eras you miss that QB DVOA is (I would guess) of higher variance in modern eras given the nature of how the game has changed, which over-privileges modern QBs precisely because of that higher variance.
I could very well be wrong about the mechanics under the hood here, but if we were accurately comparing across eras and controlling for eras we'd expect a normal distribution of QBs across those eras, and looking at the top 10, 20, 30, and 40 across eras instead the ranking appears to be highly skewed toward precisely when the explosion in QB stats happened.
2) From what I know the author is correct that YPA correlates well with winning, but it explains less of the variance in winning than does QBR or QB Rating. If you want to talk about QB stats as they correlate to winning you should pick the stats that explain the most variance, not just one that does explain variance. To be clear I SINCERELY don't know if he picked YPA because it's more favorable to Wilson than other measures which are better predictors of winning, but it's weird to pick a less effective measure over a more effective one (e.g. if you want to predict income education is a good measure, but it's a pretty bad measure if you choose to report the correlation between education and income while never mentioning that you're sitting on data on industry and job title, which is an even better one).
3) This is VERY NITPICKY, but a direct comparison between Wilson and Brady in "Head-to-Head Matchups" isn't that meaningful, because they've never actually played each other in any statistically meaningful way.
It looks more meaningful when it's framed the way it is, but doesn't seem as meaningful if you present it as what it actually is:
"In six games Russell Wilson has performed better against the Patriots' defense than Tom Brady has performed against the Seahawks' defense."
Once you explain what it actually is, you realize that a "head-to-head matchup" isn't actually a direct comparison between two players. It is, instead, a comparison of how two players have performed against two different defenses.
In team sports head-to-head matchups can only happen at the team-level. They can't happen at the individual player level.
If you wanted to try to approximate a "head-to-head" comparison for individual players in a sport like football, you have to get rid of "match-up" to actually match it up. You'd do it this way: You'd look at performance against COMMON opponents by year, spanning over a series of years. You'd want to aggregate across as many years as you can to up your N so that you're not misled by the random noise of a small sample size.
Again, I have absolutely NO IDEA if that would favor Wilson or Brady, but it's the only way you could reasonably approximate the type of comparison you're going for.
Overall, though, I thought this was a good read, and also think Wilson is a really good QB.
These are just the nitpicky things I noticed while reading. :2thumbs: