RiotVol
I Am Jack's Complete Lack of Surprise
- Joined
- Dec 25, 2011
- Messages
- 3,485
- Likes
- 12,639
Ok, here's what I got...
Using Phil Steele's total experience points, the previous year's total offensive yards per game, and the previous year's Sagarin rating, I was able to create a model to predict last year's results with a r2 of 0.78, or a model that accounts for 78% of the variation for all FBS teams.
When you pull out the SEC and compare what the model predicted versus Sagarin's final ratings, you get this...
This is how it would have shook out as far as predicted wins and losses versus what actually happened last year.
The model predicted that we should have lost every SEC game we played last year, but we ended up pulling out two wins.
When you roll the model forward to next year, you can use it to get a sense of who might be rising rising or falling when you compare it to last year's final ratings. I was a little surprised that the model revises everyone down next year EXCEPT Tennessee.
Finally, when you compare us to the SEC teams on our schedule, what we have is this:
We're still behind a lot of folks, but the gap is closing, especially if you thinking about the differential between us and Auburn last year (about 14 pts on Sagarin's scale). So, the model says that we should expect to beat Vandy, we should be in for close games against Kentucky, Missouri, and South Carolina, and Georgia and Florida are within the distance that was between us and Auburn last year. Bama is Bama.
Looking at this, I put us losing to Florida, Georgia, and Alabama, winning close ones to South Carolina and Kentucky (even though the model has us slightly behind them), and beating Vandy soundly. I don't know what to do with Missouri. They might have a deeper team, but our coaching staff is so much better. There's there's Mississippi State.
I think Mississippi State shows some of the flukiness that comes along with Sagarin's model, which tries to incorporate strength of schedule. In his final rankings, despite going 8-5, his model ranked them as #8 in the country. So, when you project the model forward and use last years final rankings in the prediction for next year, it pushes the prediction for Mississippi State so high despite losing their starting QB and three first round draft picks on defense. As a faculty member, I've gotten cheap season tickets for MSU games, but I didn't renew this year. I just don't see how they lose as much as they did, and do better than the 8-5 season that they posted the year before. If we have another Auburn-grade upset next season, this is the game I'm going with.
Using Phil Steele's total experience points, the previous year's total offensive yards per game, and the previous year's Sagarin rating, I was able to create a model to predict last year's results with a r2 of 0.78, or a model that accounts for 78% of the variation for all FBS teams.
When you pull out the SEC and compare what the model predicted versus Sagarin's final ratings, you get this...
This is how it would have shook out as far as predicted wins and losses versus what actually happened last year.
The model predicted that we should have lost every SEC game we played last year, but we ended up pulling out two wins.
When you roll the model forward to next year, you can use it to get a sense of who might be rising rising or falling when you compare it to last year's final ratings. I was a little surprised that the model revises everyone down next year EXCEPT Tennessee.
Finally, when you compare us to the SEC teams on our schedule, what we have is this:
We're still behind a lot of folks, but the gap is closing, especially if you thinking about the differential between us and Auburn last year (about 14 pts on Sagarin's scale). So, the model says that we should expect to beat Vandy, we should be in for close games against Kentucky, Missouri, and South Carolina, and Georgia and Florida are within the distance that was between us and Auburn last year. Bama is Bama.
Looking at this, I put us losing to Florida, Georgia, and Alabama, winning close ones to South Carolina and Kentucky (even though the model has us slightly behind them), and beating Vandy soundly. I don't know what to do with Missouri. They might have a deeper team, but our coaching staff is so much better. There's there's Mississippi State.
I think Mississippi State shows some of the flukiness that comes along with Sagarin's model, which tries to incorporate strength of schedule. In his final rankings, despite going 8-5, his model ranked them as #8 in the country. So, when you project the model forward and use last years final rankings in the prediction for next year, it pushes the prediction for Mississippi State so high despite losing their starting QB and three first round draft picks on defense. As a faculty member, I've gotten cheap season tickets for MSU games, but I didn't renew this year. I just don't see how they lose as much as they did, and do better than the 8-5 season that they posted the year before. If we have another Auburn-grade upset next season, this is the game I'm going with.
Attachments
Last edited:
