Comments on Advanced NFL Stats: Thanks for the Memories, Brett

Stats don't do these justice. Stats don't...

2009-09-23T00:49:50.524-04:00

Stats don't do these justice. Stats don't show his reads, the blocking, the receiver's routes, audibles, etc. For example, you have his 11/6/00 pass to Antonio Freeman and say how it was a 43 yard touchdown pass to win in overtime. But that doesn't say anything about how Freeman caught it off the tip after already landing on his back out of attempt to get it before the tip. It actually was a bad play on Favre's part, but probably Freeman's best ever plays to make the catch and save Favre after he had thrown it into coverage and had it tipped. Several of those plays were just great plays by the receiver/line and a good throw from Favre. Those all were great plays that you have, but they weren't necessarily his greatest plays.

Brian- Thanks for the reply. I know that it's ...

2009-08-03T11:22:10.891-04:00

Brian- Thanks for the reply. I know that it's impossible to model the exact significance of each game, but I had a look at game results for the past 7 seasons recently to see whether there is a trend in records and their probability of making the playoffs (of course, sometimes worse records end up making the playoffs see San Diego / New England last season). It seems as though there does seem to be an applicable table that would model the probability of a team making the playoffs given their record, but this wouldn't be able to model a player blowing a game when their team is 9-6 and needs to win to make the playoffs any more differently than a team that is 9-6 but cannot possibly make the playoffs, in which case it wouldn't matter what the players did.

Looking at the data and just using raw numbers (no smoothing), and assuming two idenitical plays that have 0.5 WPA, increasing the WP from 0.25 to 0.75.
In instance one, a team is 9-5 (the probability of a 10-5 making the playoffs is ~95%, and a 9-6 team has ~50%). The playoff-prbability added (PPA) by the play is then around 23%.
In the second instance, a team is 1-1 (probability of making playoffs from 2-1 is ~50% and from 1-2 ~27%). In this case the play has a PPA of 12%.
It's not a perfect system as it doesn't feature last-game scenarios, but it could be used to help with 'play of the year' decisions, depending on how 'play of the year' is defined.

Ian-Yes. I did almost exactly that during the play...

2009-08-02T00:28:59.055-04:00

Ian-Yes. I did almost exactly that during the playoffs/Super Bowl last year when trying to identify the "play of the year." It's easy to calculate during the playoffs, but during the year, the significance of each game is uncertain until the end of the year.

Thanks. That should read OWN 18.

2009-07-30T08:31:39.051-04:00

Thanks. That should read OWN 18.

Sorry for the double post but thinking out loud, i...

2009-07-30T08:28:24.241-04:00

Sorry for the double post but thinking out loud, in terms of trying to account for the context around a game is it possible to build in some kind of 'probability of winning Superbowl' algorithm by taking the probability a team has of winning the Superbowl given their record at the time i.e. costing your team the game in the first week of the season isn't as important as doing so when you are 9-6 heading into the last game of the season, but costing your team the win when you're 14-0 will have pretty much no effect as you're already in the playoffs.

Being nitpicky, but his best play 1-10-OWN 8, pass...

2009-07-30T08:19:55.256-04:00

Being nitpicky, but his best play 1-10-OWN 8, pass for 82 yards, TOUCHDOWN doesn't add up to a whole field length. A TD from your own 8 is a 92 yard play.