|
Post by jetstream23 on Dec 31, 2014 12:35:26 GMT -5
Jets played 5 games in December and were +8 in point differential overall. Scored 106, gave up 98 to end the season.
They lost close games to the following playoff teams (NE -2, NE -1, Detroit -7, Packers -7) and beat Pittsburgh. We are what our record says we are....but I didn't think this team was performing horribly given the lack of talent and schizophrenic nature of our QB.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Dec 31, 2014 13:52:33 GMT -5
The Pythagorean model doesn't work very well in football because you don't have the number of games necessary to develop a sufficient sample size to make an accurate prediction. That's why the model, applied to the NFL, is all over the map every year in its predictions. It works better in the NBA and MLB because they play far more games in a season. The effect of outlier games is sufficiently reduced. The 2013 Jets are a good example of how the low sample size allows outliers to skew the results. In 2013 each of our wins came on a handful of points but our losses ranged from a three point loss to a forty point loss. Those three games where we were demolished skewed the results. You can also look at how much we relied on FGs to put up points on the board and how few points we scored in several games. That's why the model showed us being a 3-13 team. The purpose of the statistical analysis is not to create a new model for looking at this year's performance but to forecast next year's performance. If this year's analysis says we are a 5-11 team then that really means the model predicts we will be a 5-11 or 6-10 team next year (where teams underperform their Pythagorean wins they usually overperform by one win the following year). This is where the model is most accurate but still the model has no function to account for a change in management, coaching, or whatever roster changes are sure to follow this offseason. Pro Football Focus has a model that allows you to calculate next year's expected wins based on current year's actual vs. expected wins. The regressions aren't perfect, but it's better than simply looking at the current year's numbers and making a prediction based on that alone. For instance, the model weighs Pythagorean wins 3x that of actual wins since the former reflects point differential and the latter does not. Either way, you've got a great point in that there is significantly more variance in the NFL than other sports due to small sample size. The "adjusted wins" metric I mentioned earlier tries to account for the variance problem, but it can't do anything about the short NFL season. If you want to be on the safe side, I'd say the expected wins for an NFL team would have to carry a large margin of error, maybe something like 4-5 wins either direction.
|
|
|
Post by Ff2 on Dec 31, 2014 15:49:41 GMT -5
What bunch of hooey
|
|
|
Post by rexneffect on Dec 31, 2014 16:48:18 GMT -5
Either way, you've got a great point in that there is significantly more variance in the NFL than other sports due to small sample size. The "adjusted wins" metric I mentioned earlier tries to account for the variance problem, but it can't do anything about the short NFL season. If you want to be on the safe side, I'd say the expected wins for an NFL team would have to carry a large margin of error, maybe something like 4-5 wins either direction. If you are leaving a margin of +/- 4 wins then you're talking about a team projected at 8-8 to be as low as 4-12 and a total shit pile or as high as 12-4 and running the conference. With that range considered acceptable results you might as well say it's completely worthless, especially when the difference of a single win is usually the difference between making the playoffs and sitting at home in January and, as with the Jets this year, a single loss can be the difference between a good draft position and a great draft position.
|
|
|
Post by Hotman on Dec 31, 2014 18:14:42 GMT -5
Well Dude, I never really thought about it that way
|
|
|
Post by cleatmarks on Jan 1, 2015 19:02:38 GMT -5
"Which is better?" Well, If moral victories are an important stat, then I guess point differential is. How about total offense/defense too? Maybe that's important lol.
The only thing important at the end of the day is wins and loses. Everything else is "Losers Lament."
|
|
|
Post by bxjetfan on Jan 1, 2015 20:52:34 GMT -5
I miss Cookie Monsta.
|
|
|
Post by JB1089 on Jan 2, 2015 7:55:19 GMT -5
Wins are what count in standings but point differential will give you a clearer picture of how "good" a team actually is. Simple.
When projecting for the next year. Use point differential as a base to begin a more subjective analysis. For example, when people were doing projections for the 2014 Jets based on the 2013 season:
DO NOT: This is an 8 win team whose offseason moves should result in +/- X wins.
DO: This is a 5 win team whose offseason moves should result in +/- X wins.
|
|
|
Post by rangerous on Jan 2, 2015 17:49:52 GMT -5
when points differential is what gets to the playoffs and superbowls then it's more important than wins and losses. geno had a 158 qbr in one game. think that will ever, ever happen again?
|
|