This opinion piece from the NFL model's developer discusses their system, ELWAY, and its forecast for the 2026 season. The key numbers mentioned are 4-13, referring to the Patriots' record two years ago. The article takes an analytical angle, explaining how ELWAY's performance ratings combine with roster ratings, as well as how it accounts for strength of schedule through simulations.
Written by the local model on 2026-09-10,
using this article's own text rather than the other coverage of the
same event (that is the story summary below).
Story summary
The NFL season has started with an unusual schedule, beginning on a Wednesday with a Super Bowl rematch between the Patriots and Seahawks. The following night, the Rams will play the 49ers in Melbourne, Australia. Nate Silver's NFL forecasting model, ELWAY, has been revamped for the new season, incorporating improvements to its quarterback rating system, QBERT. Despite some misses last year, ELWAY made smart calls on teams like the Seahawks and Patriots, leading Silver to introduce a "Model Talk" approach to explaining his forecasts. A detailed preseason forecast is available on the ELWAY landing page, providing an in-depth look at what's expected from each team this season.
Written for “John Elway's NFL Predictions” on 2026-09-10,
grounded in this article and the 0 other(s) covering the same event.
No, I’m not sure why the NFL is starting its season on a Wednesday.
asserted
NFL → ’m → Wednesday
But I won’t complain because it should be a hell of a lot more fun than watching the Republican midterm convention.
asserted
it → complain → convention
The NFL is back tonight with a Super Bowl rematch between the Patriots and Seahawks, and then the Rams will play the 49ers tomorrow night U.S. time in Melbourne, Australia.
asserted
Rams → play → Melbourne
Over the weekend, we relaunched our NFL model, ELWAY, and our quarterback rating system, QBERT.
asserted
we → relaunche → model
ELWAY made some smart calls last year — it was bullish on the Seahawks and the Pats, for instance — but also had its share of misses, clinging to hope for the Chiefs, Lions and Ravens until it was far too late.
asserted
it → make → Chiefs
Never satisfied, we’ve made quite a few improvements to ELWAY and QBERT this year and also upgraded our charts and visual presentation.
asserted
we → make → charts
Since ELWAY is still a relatively new system, I thought it would be better to show you the season through the model’s eyes, rather than giving you some sort of traditional, 32-team X’s-and-O’s preview.
asserted
it → think → preview
Therefore, we’re officially dubbing this a rare sports Model Talk.
asserted
we → dub → this
So see the ELWAY landing page for our final, extremely detailed preseason forecast.1 This newsletter is intended as a complement, a hybrid between a season preview and an explainer.
asserted
newsletter → see → preview
But a lot of this should interest even medium-nerdy football fans.
asserted
lot → interest → fans
For example, we have a fun way to account for strength of schedule by scrambling the positions of all 32 teams in each simulation.
asserted
we → have → simulation
Here’s what we’ll cover:
How ELWAY compares to Vegas;
Which teams are helped and hurt the most by the schedule;
The impact of injuries;
Roster ratings versus performance ratings and how ELWAY combines them;
Finally, the impact of quarterbacks, who are especially important in ELWAY because of QBERT.
asserted
who → ’ → QBERT
How ELWAY compares to Vegas
A projection model will never solve the problem that a 17-game NFL season with imbalanced schedules is inherently going to introduce a degree of randomness.
asserted
season → compare → randomness
The best teams in 2025 had little correlation with the elite sides of 2024: the Patriots were 4-13 two years ago.
asserted
Patriots → have → 2024
Few teams are on a sustainable, multi-year dynastic streak like we used to see with Brady’s Patriots or the Peak Mahomes Chiefs.
asserted
we → use → Patriots
Josh Allen is probably the best QB in the sport, but he’s coming off his lowest passing yardage total since 2019.
asserted
he → come → 2019
Lamar Jackson’s year was a lost cause.
asserted
year → lose → ?
Drake Maye has an argument for being the most valuable long-term asset in the league, but the adjustments we made to QBERT this summer — and his middling performance in the playoffs — hurt him.
asserted
we → have → him
Mahomes has a chance for a remarkable recovery, but he’s coming off an ACL/MCL tear.
asserted
he → have → tear
Still, one thing that helps when working on a sports model is that you can use Vegas odds2 as a sanity check.
asserted
you → help → check
That doesn’t mean Vegas is necessarily right, and we think ELWAY has a few insights that typical sports bettors might lack, like about the importance of injuries.
uncertain
bettors → mean → injuries
Moreover, it would be boring if models perfectly aligned with Vegas on every game.
asserted
models → align → game
But if your model shows profound, systematic differences from Vegas, it’s probably doing something wrong.
asserted
it → show → something
So here’s a comparison table between ELWAY’s season win totals and Vegas’s.
asserted
table → ’ → totals
We think ELWAY is a really cool model.
asserted
ELWAY → think → ?
But before we go behind the paywall, keep in mind that NFL betting lines are notoriously tough to beat.
asserted
lines → go → mind
We wouldn’t necessarily expect ELWAY alone to be enough.
asserted
ELWAY → expect → ?
Late-breaking injuries and weather can also materially affect betting lines after ELWAY posts its weekly update.
asserted
ELWAY → break → update
ELWAY + football knowledge + some betting street smarts might bring you closer.
uncertain
ELWAY → bring → you
But we’re mostly using this as a diagnostic exercise.
asserted
we → use → exercise
The differences between ELWAY and Vegas can reveal how the model “thinks” about football.
asserted
model → reveal → football
