Shaq: Come in early, leave late. Learn how to study film from Luke and how to be a pro from TD. Funchess: Come in early, leave late. Get extra work in with Cam and KB. Stay in Proehl's pocket. Lee Ward: Coime in early, leave late. Don't start killing people until the first pre-season game. May lord have mercy on their souls. Shula: Come in late, leave early. No eye contact and no talking.
I work with math and models (not the Igo kind, the nerd kind) occasionally at my job. Here are my issues with this article:
Uses Pro Football Reference's Approximate Value statistic as a be-all end-all metric. Then immediately from the description of Approximate Value: "AV is not meant to be a be-all end-all metric. Football stat lines just do not come close to capturing all the contributions of a player the way they do in baseball and basketball." Does not account for team depth by choosing relatively random players. The top 4 front seven? Really? Why not the top 7 front seven? Does not account for injuriesRookies have the same value as backups. That means Devin Funchess has the same value right now as Jason Avant or Marcus Lucas? Ran linear regression with last years data for Elo ratings and found a high correlation which means fart noises pfffffffffffffffttttttttttttttt who caresNo seriously the last point is where statistically he's correct but logically he's really flawed. Naturally you want to test your model against historical data, but the results aren't really surprising. It's highly correlated...it should be! He's taking the final AV of players (which accounts for team success in some way) and is saying teams with the most AV win more. The more successful teams are successful? Well...yes, its right there in the definition of AV.
Using it to project forward is essentially worthless, especially when you're cherry picking your observations. If he did this based on the entire offensive or defensive rosters and came up with projected rankings I would consider it a little more but it's nothing more than a July fluff piece with some statistical bologna thrown in there. Not shitting on the guys statistics, they seem to be pretty sound, but the lack of common sense in the application is hard to swallow.