CEB
Helluva Engineer
- Messages
- 2,621
Yeah… if it ain’t a turnover it’s a touchback!
Stats models and rankings
- Thread starter slugboy
- Start date
leatherneckjacket
Helluva Engineer
- Messages
- 2,081
- Location
- Atlanta, GA
We had three drives start in the short field territory against Miami and had a punt, touchdown and a FG.
Against BC, we had one and punted.
Against Bowling Green, we had one and scored a TD.
Against Wake, we had four with three FGs and downs.
No drive data for Ole Miss or Louisville.
- Messages
- 11,517
Against UNC we blocked a punt, then threw a pick. Was it BC where we made a pick on their side of the field and kicked a field goal?
leatherneckjacket
Helluva Engineer
- Messages
- 2,081
- Location
- Atlanta, GA
Against Ole Miss, we had one and punted.
Against Louisville, we had three and punted once and missed two FGs.
So on 13 drives, we have scored 26 points, or 2.00 PPD, which is atrocious and worse than our overall PPD.
ibeattetris
Helluva Engineer
- Messages
- 3,604
At least according to bcftoys, Ole Miss didn't have any https://www.bcftoys.com/2023-ps-w03#137
I don't see any of our "start" being less than 60.
ibeattetris
Helluva Engineer
- Messages
- 3,604
Wake we forced a fumble inside the 20 and got a fg
I think Miami is what you were thinking of
- Messages
- 11,517
It was. Those were both offensive failures.
Van Dyke making the tackle in the Miami game saved 4 points instead of postponing them.
Lotta Booze
Ramblin' Wreck
- Messages
- 779
- Messages
- 11,517
The Eckel matchups are a strange version of the immovable object vs the irresistible force
“I can’t reliably get a first down” vs “I won’t stop you”
leatherneckjacket
Helluva Engineer
- Messages
- 2,081
- Location
- Atlanta, GA
Yeah, not sure how I screwed that up.
I can’t vouch for the accuracy of this info. that I picked up on the Internet, but very weird if it’s the truth!!
Allegedly,,,,
Georgia Tech is having one of the strangest seasons I’ve ever seen.
First off, the Yellow Jackets have only won in their even numbered games (their wins have come in games 2, 4, 6, and 8). Game 10 is Clemson, game 12 is Georgia.
Second, they’ve played three ranked teams. All three have been AP #17 (#17 Ole Miss, #17 Miami, #17 North Carolina). They are 2-1 vs. the AP #17 team.
Those two wins over ranked teams are the most of any team in the state of Georgia.
Their last four games:
• 11-point loss to BOWLING GREEN
• Victory over #17 Miami
• 15-point loss to Boston College
• Victory over #17 North Carolina
Their win over Miami was absurd enough, as the Canes had them beat but chose to run the ball instead of taking a knee with 0:26 seconds left.
Going back to the only winning in even numbered games stat, last night they only had success in even numbered QUARTERS.
They had 0 points in the 1st and 3rd quarters. All 46 of their points came in the 2nd and 4th quarters (0 in the 1st, 24 in the 2nd, 0 in the 3rd, 22 in the 4th).
Here’s possibly the weirdest part:
Georgia Tech has scored 254 points.
Georgia Tech has allowed 254 points.
Georgia Tech has scored 33 touchdowns.
Georgia Tech has allowed 33 touchdowns.
In other words, as a team they have scored 31.8 points per game and allowed 31.8 points per game.
The Yellow Jackets have also led by two possessions in two of their losses, and trailed by two possessions in two of their wins.
• Led Louisville by 15, lost
• Led Bowling Green by 14, lost
• Trailed Miami by 10, won
• Trailed UNC by 14, won
Allegedly,,,,
Georgia Tech is having one of the strangest seasons I’ve ever seen.
First off, the Yellow Jackets have only won in their even numbered games (their wins have come in games 2, 4, 6, and 8). Game 10 is Clemson, game 12 is Georgia.
Second, they’ve played three ranked teams. All three have been AP #17 (#17 Ole Miss, #17 Miami, #17 North Carolina). They are 2-1 vs. the AP #17 team.
Those two wins over ranked teams are the most of any team in the state of Georgia.
Their last four games:
• 11-point loss to BOWLING GREEN
• Victory over #17 Miami
• 15-point loss to Boston College
• Victory over #17 North Carolina
Their win over Miami was absurd enough, as the Canes had them beat but chose to run the ball instead of taking a knee with 0:26 seconds left.
Going back to the only winning in even numbered games stat, last night they only had success in even numbered QUARTERS.
They had 0 points in the 1st and 3rd quarters. All 46 of their points came in the 2nd and 4th quarters (0 in the 1st, 24 in the 2nd, 0 in the 3rd, 22 in the 4th).
Here’s possibly the weirdest part:
Georgia Tech has scored 254 points.
Georgia Tech has allowed 254 points.
Georgia Tech has scored 33 touchdowns.
Georgia Tech has allowed 33 touchdowns.
In other words, as a team they have scored 31.8 points per game and allowed 31.8 points per game.
The Yellow Jackets have also led by two possessions in two of their losses, and trailed by two possessions in two of their wins.
• Led Louisville by 15, lost
• Led Bowling Green by 14, lost
• Trailed Miami by 10, won
• Trailed UNC by 14, won
leatherneckjacket
Helluva Engineer
- Messages
- 2,081
- Location
- Atlanta, GA
Lithium kicks in only for even numbered games, I guess.
ThatGuy
Ramblin' Wreck
- Messages
- 977
- Location
- Evergreen, CO
Jumping in to point out that we made our debut on the Sickos Committee November poll at #3, with a bullet:
(The Sickos are essentially CFB fans who look for the absurd in college football. Schadenfreude, strange occurrences, teams that beat teams they aren't supposed to...that's what they're all about. They poll 1500 Sickos members every month, and rank the results. And the results can include anything from CFB teams to political figures to Kentucky Fried Chicken. Whichever gets the most votes rises to the top.)
(The Sickos are essentially CFB fans who look for the absurd in college football. Schadenfreude, strange occurrences, teams that beat teams they aren't supposed to...that's what they're all about. They poll 1500 Sickos members every month, and rank the results. And the results can include anything from CFB teams to political figures to Kentucky Fried Chicken. Whichever gets the most votes rises to the top.)
roadkill
Helluva Engineer
- Messages
- 1,837
Rather than continue the SoS discussion tangent in the ACC Schedule thread, I decided to bring my question here.
I don't know the source of @Randy Carson's SoS numbers (where we are slotted at 42), but after checking various sites, they are all over the map. FEI provides 3 different rankings which can be widely apart. Sagarin has us at 50, Congrove has us at 33, Massey at 64, Colley is at 16 but is just for games played, similarly, CPI ratings have us at 10.
My question to the stats gurus on here is, do you have a go-to SoS source? The only reason I ask is that invariably there will be SoS discussions on this board even though our record so far this season indicates you can through SoS out the window.
I don't know the source of @Randy Carson's SoS numbers (where we are slotted at 42), but after checking various sites, they are all over the map. FEI provides 3 different rankings which can be widely apart. Sagarin has us at 50, Congrove has us at 33, Massey at 64, Colley is at 16 but is just for games played, similarly, CPI ratings have us at 10.
My question to the stats gurus on here is, do you have a go-to SoS source? The only reason I ask is that invariably there will be SoS discussions on this board even though our record so far this season indicates you can through SoS out the window.
- Messages
- 11,517
If I was looking for one number, I’d pick the strength of schedule from the rating system that I thought did the best job so far this year.
It’s a pretty hard job to do well. To get the strength of schedule, you rate all the teams and look at who a team plays, and assign a score to the sum of all their opponents. To rate all the teams, you need to rate who they play, so you have a nice circular dependency.
Fortunately, you can use common opponents and the stats against them to normalize what you’re doing, but it depends on the stats you use. Sagarin is pretty much using points and wins, so they’re easy, but SP+ and FEI are using a lot of different stats, so they’re more complicated. Depending on how much you care about efficiency vs big plays, you get a much different number.
Most of the models are pretty close to 50% against the point spread. They do pretty well spread over all 131 teams. They aren’t doing well with Georgia Tech this year.
What do I do? I look at three or four models that make sense, use those to give a range (“we’re somewhere between the 20th toughest schedule and the 60th”), and I don’t get super worked up about it being exact.
If I have two people working with me, and I see one has an IQ of 135 and the other has an IQ of 120, I’ll judge for myself who can figure out a problem better—some things are a mess to assign a single number to. Same with college football.
(The Bowling Green loss was still ridiculous. These numbers can be variable and vague, and I can still know we should win those games at least 19 times of of 20 if we have our minds right)
I also think that most of the top 30 teams play a tough enough schedule to know what they are. UGA could beat anyone in the country, but they’re more vulnerable this year than last
It’s a pretty hard job to do well. To get the strength of schedule, you rate all the teams and look at who a team plays, and assign a score to the sum of all their opponents. To rate all the teams, you need to rate who they play, so you have a nice circular dependency.
Fortunately, you can use common opponents and the stats against them to normalize what you’re doing, but it depends on the stats you use. Sagarin is pretty much using points and wins, so they’re easy, but SP+ and FEI are using a lot of different stats, so they’re more complicated. Depending on how much you care about efficiency vs big plays, you get a much different number.
Most of the models are pretty close to 50% against the point spread. They do pretty well spread over all 131 teams. They aren’t doing well with Georgia Tech this year.
What do I do? I look at three or four models that make sense, use those to give a range (“we’re somewhere between the 20th toughest schedule and the 60th”), and I don’t get super worked up about it being exact.
If I have two people working with me, and I see one has an IQ of 135 and the other has an IQ of 120, I’ll judge for myself who can figure out a problem better—some things are a mess to assign a single number to. Same with college football.
(The Bowling Green loss was still ridiculous. These numbers can be variable and vague, and I can still know we should win those games at least 19 times of of 20 if we have our minds right)
I also think that most of the top 30 teams play a tough enough schedule to know what they are. UGA could beat anyone in the country, but they’re more vulnerable this year than last
Randy Carson
Helluva Engineer
- Messages
- 1,316
- Location
- Apex, NC
roadkill
Helluva Engineer
- Messages
- 1,837
Thanks. That seems as good as any, and is in the middle of the range of some of the other rankings. Our SoS at the beginning of the season was likely higher but has dropped some due to the generally poor performance of the majority of ACC teams we play, including Clemson.
ibeattetris
Helluva Engineer
- Messages
- 3,604
Apart from everything already mentioned, there is also the fact that some SOS are only team's you have currently played, and some are the strength of all teams on your schedule (preseason before anyone has played a game still have SOS). Any SOS that includes all games for GT will obviously be pretty impacted by the last game on our schedule. ESPN FPI Resume differentiates this. According to them, we've had the 35th hardest schedule so far, and have the 28th hardest going forward. Things like FEI only consider the games that have been played. It is also important that there is more to it than just "how strong on average were the teams I played" (this is further explained in the final link down below). The "expected number of wins" for good/great teams is different from the average strength across all opponents.
This is GT's different SOS ranks:
Expected losses elite team: 0.48 (71)
Expected losses good team: 2.04 (46)
Expected losses average team: 4.45 (38)
So according to this, an elite team would have a pretty easy go with our schedule, but an average team would be performing exactly as we are +/- half a game.
Massey has some interesting comments about SOS on his pages.
This is how he defines a very rudimentary ranking system to work:
To get the feel for how computer ratings work, you may want to try this iterative procedure(https://masseyratings.com/faq.php):
- set each team's rating to zero
- calculate each team's SOS to be the average rating of their opponents
- calculate each team's rating to be their average net margin of victory plus their SOS
- go back to step 2 and repeat until the ratings converge
More info:
Massey Ratings Description
masseyratings.com
stinger78
Helluva Engineer
- Messages
- 4,401
I like the simplicity of Massey. Though some may question its validity, I think it's a good way to look at ranking teams.