That SP+ makes WAY more sense for all the reasons I listed earlier about the FPI being faulty. I certainly don't think we are deserving of being #9, but if you take the homerism out of the equation it makes sense for Miss State to be above all the schools I mentioned earlier: coaching and roster stability. Returning starting QB. Etc.A significant chunk of our prior performance variable in their model is the Pruitt years, i.e., no offense. It's just math. They had us starting around 53 last year and we were in their top 25 until after the Purdue game. We ended the year with them at #26.
So all of the teams that are overrated/underrated is because primarily they're looking at how they were over the last 4 years. 3 years ago LSU was national champs. Time has moved on but ESPN's model to a degree by design is stuck in the past. jmo.
SP+ seems to rely less on prior performance sort of like past performance is no guarantee of future success. Connelly goes in depth with every team looking at all the little details. ESPN FPI is different. They don't do a lot of common sense work with FPI. Basically, they're lazy. jmo.
edit to add: Maybe a simpler way to understand this is FPI says our offense last year (for Tennessee) was an outlier. Connelly looks at our offense last year and then looks at Heupel's offenses at UCF and says, no, it's real. jmo.
View attachment 447449
A significant chunk of our prior performance variable in their model is the Pruitt years, i.e., no offense. It's just math. They had us starting around 53 last year and we were in their top 25 until after the Purdue game. We ended the year with them at #26.
So all of the teams that are overrated/underrated is because primarily they're looking at how they were over the last 4 years. 3 years ago LSU was national champs. Time has moved on but ESPN's model to a degree by design is stuck in the past. jmo.
SP+ seems to rely less on prior performance sort of like past performance is no guarantee of future success. Connelly goes in depth with every team looking at all the little details. ESPN FPI is different. They don't do a lot of common sense work with FPI. Basically, they're lazy. jmo.
edit to add: Maybe a simpler way to understand this is FPI says our offense last year (for Tennessee) was an outlier. Connelly looks at our offense last year and then looks at Heupel's offenses at UCF and says, no, it's real. jmo.
View attachment 447449
View attachment 447406Cat missilery can be deadly…
Yes but FPI is "quick and dirty" analytics. Not a human poll.LSU in any poll? Didn't their roster get gutted?
All good and helpful info.These models use a collection of variables to try to predict mostly offensive scoring and defensive scoring. Some also include special teams.
For example, SP+ says that we will have the 3rd best offense in the nation and average 41.6 points per game. It says we will have the 42nd best defense and give up 22.9 points per game on average. Subtract defense from offense, 41.6 - 22.9 = 18.7. That 18.7 is our SP+ rating. It is a handicapping tool. If we're playing a team that has a rating of 10.0 then on a neutral field we would be favored by 8.7 points.. The models rank us by our rating not by how many games we're expected to win.
Everyone has their own models so they have different approaches of trying to figure out how many points our offense will score and how many points/game our defense will give up. FPI gives a team 3.3 "extra" points for a returning QB. They also factor in stability at the head coach position. They factor in how many days of rest (bye week) and how far a team has to travel. They claim all their variables lead to a more accurate handicapping tool. Models include returning production and recruiting/talent among other things. Most models also now include transfer players added to a roster (usually at a discounted value). Again, all they're trying to do is predict the outcome of individual games.
Most models, and especially ESPN's FPI, see tremendous movement with a lot of teams after week 1 of a new season. Teams can move as much as 20 +/- spots up or down depending on how week 1 plays out across the college football landscape and strength of schedules are updated with real time results. The largest movements happen in week 1 and movement in the ranking is increasingly less as the season progresses.
I think they are probably close to the same number for our defense as Connelly. I think where SP+ differs from FPI in our case is projecting our scoring offense. So for example say we scored an average of 40 points per game last year and the 3 prior years on offense we scored an average of 20 points. Most recent year accounts for more than the 3 prior years but the 3 prior years still count. So 40*0.67 + 20*0.33 = 33.4 points per game on average. I think that's essentially the difference this year for us with FPI vs SP+. jmo.
FPI has our rating at 9.7. Last year we finished at 9.3. In 2020 we finished at 3.0. In 2019 we finished at 7.8 and in 2018 we finished at 1.2. Right now FPI believes our offensive scoring last year was an outlier based on recent history. They think we'll be marginally better this year than last year but not by much. I disagree but whatever. They gon' learn. jmo.
FPI is garbage ... until they rank the Vols #1. Then, maybe, it will be respectable.These models use a collection of variables to try to predict mostly offensive scoring and defensive scoring. Some also include special teams.
For example, SP+ says that we will have the 3rd best offense in the nation and average 41.6 points per game. It says we will have the 42nd best defense and give up 22.9 points per game on average. Subtract defense from offense, 41.6 - 22.9 = 18.7. That 18.7 is our SP+ rating. It is a handicapping tool. If we're playing a team that has a rating of 10.0 then on a neutral field we would be favored by 8.7 points.. The models rank us by our rating not by how many games we're expected to win.
Everyone has their own models so they have different approaches of trying to figure out how many points our offense will score and how many points/game our defense will give up. FPI gives a team 3.3 "extra" points for a returning QB. They also factor in stability at the head coach position. They factor in how many days of rest (bye week) and how far a team has to travel. They claim all their variables lead to a more accurate handicapping tool. Models include returning production and recruiting/talent among other things. Most models also now include transfer players added to a roster (usually at a discounted value). Again, all they're trying to do is predict the outcome of individual games.
Most models, and especially ESPN's FPI, see tremendous movement with a lot of teams after week 1 of a new season. Teams can move as much as 20 +/- spots up or down depending on how week 1 plays out across the college football landscape and strength of schedules are updated with real time results. The largest movements happen in week 1 and movement in the ranking is increasingly less as the season progresses.
I think they are probably close to the same number for our defense as Connelly. I think where SP+ differs from FPI in our case is projecting our scoring offense. So for example say we scored an average of 40 points per game last year and the 3 prior years on offense we scored an average of 20 points. Most recent year accounts for more than the 3 prior years but the 3 prior years still count. So 40*0.67 + 20*0.33 = 33.4 points per game on average. I think that's essentially the difference this year for us with FPI vs SP+. jmo.
FPI has our rating at 9.7. Last year we finished at 9.3. In 2020 we finished at 3.0. In 2019 we finished at 7.8 and in 2018 we finished at 1.2. Right now FPI believes our offensive scoring last year was an outlier based on recent history. They think we'll be marginally better this year than last year but not by much. I disagree but whatever. They gon' learn. jmo.