Subscribe to:
Post Comments (Atom)
Football Insights
Brandon Thorn has the 49ers OL ranked 11th, PFF 10th. With Football Insights 9th ranking it seems everyone is on the same page except for s...
-
The postseason is a year away. Win or lose the only thing they're playing for today is pride. Another opportunity to advertise your wo... -
Good morning everyone and welcome to day 3 of the NFL draft. Let's take a look at who the 49ers were able to add on day 2: Drake Jacks... -
Nope, it's not a misprint. After trading up to the #3 pick from #12, the 49ers have put themselves in a position to select their franch...
3 days until vets report
ReplyDeleteCoach,
ReplyDeleteI may be wrong. But I believe that there is a non-contact rule until at least two weeks after the vets arrive. Once the hitting starts, the wheat begins to be separated from the chaff.
👀
DeleteAES,
DeleteI googled the NFL rules regarding full contact practices in camp. It said teams may conduct full contact practice on the 6th day of practice
EPA on 3rd down in 2025:
ReplyDelete0.51 - Brock Purdy
0.36 - Joe Burrow
0.35 - Jordan Love
Trent Dilfer is a strong believer that EPA should cover the entire team rather than just one player. As he points out, it's an 11 on 11 game.
ReplyDeleteEvery QB stat that anyone has ever measured is contaminated. Yards? Receivers run the routes and make the catches. TDs? OL opened the lane, TE held his block, RB got to the edge. Completion %? Drops, scheme, air yards. There is no such thing as a clean individual stat in football.
DeleteI think the big misnomer with QB EPA is that the external variables are doing the heavy lifting while the QB is along for the ride, or that the QB contribution is on equal footing with the supporting cast. Neither are true. It is a QB-driven stat. Always has been. And yet we rank QBs constantly. We debate Purdy vs Jones. We argue about who's a system QB. We do this with stats that are fundamentally team outputs. So either we stop ranking QBs, or we get smarter about how we use the data.
There's a mathematical operation called the Schur complement. It was developed for linear algebra. It answers one question: what remains of variable A (e.g. QB stats) after you condition out the influence of everything else? In other words what's actually the QBs stats?
Apply it to a QB. Build a matrix of his performance stats. Build a second matrix of his supporting cast quality, OL health, WR availability, RB usage, TE presence. The Schur complement subtracts the cast's contribution from the QB's numbers.
What's left is the signal.
Think of it like noise-canceling headphones. The cast health matrix is the background noise. The Schur operation strips it out. You're left hearing only what the QB is actually generating, independent of who's blocking for him and who's running his routes.
I tested this on the 2025 49ers, the most controlled natural experiment in recent NFL history. Purdy got hurt Week 1. Mac Jones started 8 games. Same OL. Same McCaffrey. Same depleted WR/TE room. Same Shanahan system. Only the QB changed.
Raw stats were almost identical:
Mac Jones: 2,151 yds, 13 TD, 6 INT, 69.6%
Brock Purdy: 2,167 yds, 20 TD, 10 INT, 69.4%
Same yards. Same completion rate. Standard analysis calls this roughly a wash. The math disagrees.
After conditioning out the identical supporting cast, two distinct QB profiles emerge.
Purdy: higher absolute Schur score; genuine elite signal reaching beyond what the system provides.
Jones: higher K-stiffness; meaning a larger fraction of his production was cast-independent.
K-stiffness is the measure of how rigid your performance is relative to your environment. High K = your numbers don't move much when the cast degrades. Jones was actually MORE environment-independent than Purdy, not because he's better, but because he never pushed beyond what the system gave him.
Purdy pushed harder. More TDs, higher EPA, bigger throws. Also more INTs, partly because he was pressing with Jennings as WR1 and did not have Kittle. His performance was more coupled to the cast because he was demanding more from it. The math separates aggression from recklessness.
PFF confirms the absolute ranking; Purdy graded 90.0 (4th in NFL), Jones 77.1 (13th). Purdy is clearly the better QB. But the Schur analysis tells you something PFF can't: HOW they differ, not just by how much. One extends the system. One maximizes it.
This is what "decoding individual stats from team stats" actually looks like. You don't ignore the team context; you model it explicitly, then subtract it. What you're left with is as close to a true individual signal as the data allows. You can't remove the team from football. But you can condition on it.
Every football argument about whether someone is a "system QB" is really an argument about K-stiffness; how much their performance degrades when you degrade the cast around them. We now have a way to measure that directly instead of just arguing about it.
The broader principle: in any complex system, football, biology, finance, the interesting signal is always buried under correlated noise. The math that strips out the noise is the same whether you're evaluating a QB, a drug candidate, or a financial instrument.
Razor,
DeleteThanks for this informative breakdown.
Raw stats were almost identical:
I'm a little confused on these stat lines.
"Mac Jones: 2,151 yds, 13 TD, 6 INT, 69.6%
Brock Purdy: 2,167 yds, 20 TD, 10 INT, 69.4%
Same yards. Same completion rate. Standard analysis calls this roughly a wash. The math disagrees."
With Purdy throwing more Ints than Jones, should Mac' stats be much better than only slightly better?
As I stated yesterday, Purdy is that guy. But, given his familiarity with the Shanahan offense, he should be.
Again, your breakdown on numbers provides a better perspective on how the EPA system works. I must admit that I see football based on the overall team performance and mostly on what I see with my eyes. Call it old school.
74% of all rushing TDs come from inside the 10
ReplyDelete55% of all TDs from inside the 10
44% of all passing TDs come from inside the 10
Brock is top 5, maybe top 3. We are so lucky to have lucked out.
ReplyDeleteAgreed.
DeleteAbsolutely. I often have to remind myself that he was the last pick of the draft.
ReplyDelete