ESPN Analytics Rank WRs, QBs, and Teams

[ivotuk]

Not real interesting, but a fun distraction for a few minutes.

The Best NFL Receivers, According to ESPN Analytics

These ratings, updated weekly, use player-tracking data from NFL Next Gen Stats to evaluate every route a pass catcher runs and scores his performance in three phases of the game, from 0 to 99

The Best NFL Receivers, According To ESPN Analytics

Receiver ratings from ESPN Analytics use player-tracking data to evaluate every route a pass catcher runs and scores his performance in three phases of the game.

projects.fivethirtyeight.com

2022 NFL Predictions For the regular season and playoffs, updated after every game.

Projections: Record Point Diff Make Playoffs Win Division
Seahawks 5-3 NFC West 10-7 +19.2 62% 34%

2022 NFL Predictions

FiveThirtyEight's 2022 NFL forecast uses an Elo-based model to calculate each team's chances of advancing to the playoffs and winning the Super Bowl.

projects.fivethirtyeight.com

Quarterback ratings

2022 NFL Predictions

FiveThirtyEight's 2022 NFL forecast uses an Elo-based model to calculate each team's chances of advancing to the playoffs and winning the Super Bowl.

projects.fivethirtyeight.com

 

[AgentDib]

It's an interesting area that people will continue to work on, but there are still some gaping holes in this publicly available stuff.

Receiver scoring is about 50% catch rate and YAC, which make sense and is relatively easy to track with stats and our current data sets. Unfortunately, the other 50% is "open rate" which looks at player position data but does not know about defensive schemes. Lockett is way ahead of DK in this metric, but how much of that is because DK is getting bracketed more by the defense, or because DK is lining up in the X and fighting through press coverage vs. getting a free release? There's still a long way to go to make this a meaningful stat IMO.

I'll also include my usual disclaimer about 538 simulations spanning multiple weeks. Simulations that assume NFL games are independent outcomes become increasingly inaccurate when there is more than a week or two left in the season. In the real world football outcomes are highly dependent on previous outcomes which makes the extremes much more likely.

For example, 538 has the Eagles at >99% likely to make the playoffs, but there's a greater than 0.1% chance that Jalen Hurts gets injured for the season this week and then all bets would be off. On the other side of things, they have the Rams as just 23% likely to make the playoffs, which seems fair based on their season so far. However, I'm sure none of us would be that surprised if they figured things out and improved as the season went on.

 

[olyfan63]

It's an interesting area that people will continue to work on, but there are still some gaping holes in this publicly available stuff.

Receiver scoring is about 50% catch rate and YAC, which make sense and is relatively easy to track with stats and our current data sets. Unfortunately, the other 50% is "open rate" which looks at player position data but does not know about defensive schemes. Lockett is way ahead of DK in this metric, but how much of that is because DK is getting bracketed more by the defense, or because DK is lining up in the X and fighting through press coverage vs. getting a free release? There's still a long way to go to make this a meaningful stat IMO.

I'll also include my usual disclaimer about 538 simulations spanning multiple weeks. Simulations that assume NFL games are independent outcomes become increasingly inaccurate when there is more than a week or two left in the season. In the real world football outcomes are highly dependent on previous outcomes which makes the extremes much more likely.

For example, 538 has the Eagles at >99% likely to make the playoffs, but there's a greater than 0.1% chance that Jalen Hurts gets injured for the season this week and then all bets would be off. On the other side of things, they have the Rams as just 23% likely to make the playoffs, which seems fair based on their season so far. However, I'm sure none of us would be that surprised if they figured things out and improved as the season went on.

Tyler just freaking gets open, against anyone and everyone, always has. So not a surprise he has a high rating there, league-leading. Agreed it's subjective, but this one passes the eye test. It's also pretty impressive he's also at the top for "makes the catch". No surprise he's not even on the chart for "Yards After Catch", I think after his broken leg injury a few years back, he makes more business decisions, but no complaints here on that.

 

[DarkVictory23]

For example, 538 has the Eagles at >99% likely to make the playoffs, but there's a greater than 0.1% chance that Jalen Hurts gets injured for the season this week and then all bets would be off. On the other side of things, they have the Rams as just 23% likely to make the playoffs, which seems fair based on their season so far. However, I'm sure none of us would be that surprised if they figured things out and improved as the season went on.

538's model does incorporate injury possibility to Hurts--but only Hurts. No other position has an individual adjustment. In fact, their model doesn't ignore previous outcomes in its simulations at all. Every victory or loss adjusts their ELO rating going forward in the simulation.

And I think the fact that in NFL history that no team has ever gone 7-0 and missed the playoffs add some credence to it. Is there a greater than .1% chance Hurts gets injured? Yeah, but I don't think that means all bets are off.

That said, I do find their ELO system is far too dependent on preseason ratings and is slow to adjust. I'm sure there are stats that 538 has seen showing the model is more successful in general by doing this but in specific circumstances (for example a team having a journeyman backup QB as their starter who plays at a top 5 level unexpectedly) you get a situation like now where guys like Kyler Murray have a better rating right now than Geno.