Stats models and rankings

[ibeattetris]

One of my favorites was a ranking based on EPA diff but it no longer exists (at least the one I knew of is gone).

I think when it comes to a lot of these rankings, some are trying to determine “who is the best team” and others “who is having the best season”. I prefer the “best team” type of metrics that ignore wins and losses as a parameter, but I think CFP type polls should value wins and losses more (if that makes sense).

Cool thread thanks for @ing me @slugboy

 

[bobongo]

One of my favorites was a ranking based on EPA diff but it no longer exists (at least the one I knew of is gone).

I think when it comes to a lot of these rankings, some are trying to determine “who is the best team” and others “who is having the best season”. I prefer the “best team” type of metrics that ignore wins and losses as a parameter, but I think CFP type polls should value wins and losses more (if that makes sense).

Cool thread thanks for @ing me @slugboy

Indeed. Any poll that ignores wins and losses for the purpose of determining the CFP is useless. Winning is the very object of the game. At the end of the day, there should be three criteria:

W/L record
Strength of schedule
Margin of victory/defeat

That's it. The only stat that counts is what's on the scoreboard at the end of the game, because that's the object of the game.

 

[roadkill]

Indeed. Any poll that ignores wins and losses for the purpose of determining the CFP is useless. Winning is the very object of the game. At the end of the day, there should be three criteria:

W/L record
Strength of schedule
Margin of victory/defeat

That's it. The only stat that counts is what's on the scoreboard at the end of the game, because that's the object of the game.

That's essentially what the Colley matrix does. Except I don't think it accounts for the margin of victory. Strength of schedule is determined by opponent W/L. So it doesn't start making sense until around week 4 or 5.

 

[GetYourBuzzOn]

One of my favorites was a ranking based on EPA diff but it no longer exists (at least the one I knew of is gone).

I think when it comes to a lot of these rankings, some are trying to determine “who is the best team” and others “who is having the best season”. I prefer the “best team” type of metrics that ignore wins and losses as a parameter, but I think CFP type polls should value wins and losses more (if that makes sense).

Cool thread thanks for @ing me @slugboy

I agree with this 100%. I prefer to view power rankings that try to predict who the best team is, vs. who has the best resume. I think it all stemmed from my experience as an 11-year-old in 1990, where I swore at that young age that we would beat COL up and down a neutral field....come to find out that our historic SP+ that year was 17.8 vs. Colorado's 19.1 LOL

 

[slugboy]

These are not advanced stats—they’re just stats from theacc.com.

Offense after 5 games (we’re average to below average for the conference)

Rank	Team	G	TD	FG	XPT	2XP	DXP	SAF	PTS	AVG/G
1	Miami	4	22	7	20	1	0	0	175	43.8
2	Florida State	4	23	4	23	0	0	0	173	43.3
3	Syracuse	5	25	6	23	0	0	0	191	38.2
4	Louisville	5	24	6	23	0	0	0	185	37.0
5	North Carolina	4	18	6	17	0	0	0	143	35.8
6	Clemson	5	24	3	23	0	0	0	176	35.2
7	Duke	5	21	5	20	1	0	0	163	32.6
8	Georgia Tech	5	21	6	18	0	0	0	162	32.4
9	Wake Forest	4	13	9	11	0	0	0	116	29.0
10	Boston College	5	18	5	16	0	0	0	139	27.8
11	NC State	5	16	5	16	0	0	0	127	25.4
12	Virginia Tech	5	15	6	14	0	0	1	124	24.8
13	Pitt	5	15	4	15	0	0	0	117	23.4
14	Virginia	5	13	5	12	1	0	0	107	21.4

Defense through 5 games (like the first chart, 4 for some)

Rank	Team	G	TD	FG	XPT	2XP	DXP	SAF	PTS	AVG/G
1	Duke	5	7	2	6	1	0	0	56	11.2
2	Miami	4	6	3	5	0	0	0	50	12.5
3	Syracuse	5	10	2	8	0	0	0	74	14.8
4	Louisville	5	12	1	11	0	0	0	86	17.2
5	NC State	5	12	5	11	1	0	0	100	20.0
6	Clemson	5	13	4	12	1	0	0	104	20.8
7	North Carolina	4	10	6	10	0	0	0	88	22.0
8	Florida State	4	11	5	9	0	0	0	90	22.5
9	Wake Forest	4	11	5	10	0	0	0	91	22.8
10	Virginia Tech	5	16	3	16	0	0	0	121	24.2
11	Pitt	5	16	6	16	0	0	0	130	26.0
12	Georgia Tech	5	19	8	16	0	0	0	154	30.8
13	Boston College	5	23	2	22	0	0	0	166	33.2
14	Virginia	5	24	4	22	0	0	0	178	35.6

 

[roadkill]

These are not advanced stats—they’re just stats from theacc.com.

Offense after 5 games (we’re average to below average for the conference)

Rank	Team	G	TD	FG	XPT	2XP	DXP	SAF	PTS	AVG/G
1	Miami	4	22	7	20	1	0	0	175	43.8
2	Florida State	4	23	4	23	0	0	0	173	43.3
3	Syracuse	5	25	6	23	0	0	0	191	38.2
4	Louisville	5	24	6	23	0	0	0	185	37.0
5	North Carolina	4	18	6	17	0	0	0	143	35.8
6	Clemson	5	24	3	23	0	0	0	176	35.2
7	Duke	5	21	5	20	1	0	0	163	32.6
8	Georgia Tech	5	21	6	18	0	0	0	162	32.4
9	Wake Forest	4	13	9	11	0	0	0	116	29.0
10	Boston College	5	18	5	16	0	0	0	139	27.8
11	NC State	5	16	5	16	0	0	0	127	25.4
12	Virginia Tech	5	15	6	14	0	0	1	124	24.8
13	Pitt	5	15	4	15	0	0	0	117	23.4
14	Virginia	5	13	5	12	1	0	0	107	21.4

Defense through 5 games (like the first chart, 4 for some)

Rank	Team	G	TD	FG	XPT	2XP	DXP	SAF	PTS	AVG/G
1	Duke	5	7	2	6	1	0	0	56	11.2
2	Miami	4	6	3	5	0	0	0	50	12.5
3	Syracuse	5	10	2	8	0	0	0	74	14.8
4	Louisville	5	12	1	11	0	0	0	86	17.2
5	NC State	5	12	5	11	1	0	0	100	20.0
6	Clemson	5	13	4	12	1	0	0	104	20.8
7	North Carolina	4	10	6	10	0	0	0	88	22.0
8	Florida State	4	11	5	9	0	0	0	90	22.5
9	Wake Forest	4	11	5	10	0	0	0	91	22.8
10	Virginia Tech	5	16	3	16	0	0	0	121	24.2
11	Pitt	5	16	6	16	0	0	0	130	26.0
12	Georgia Tech	5	19	8	16	0	0	0	154	30.8
13	Boston College	5	23	2	22	0	0	0	166	33.2
14	Virginia	5	24	4	22	0	0	0	178	35.6

I watched the Duke-ND game last night, and these stats seem about right. Duke's defense is elite, despite not having any superstars. While they may have a bit more talent than us at a few key positions, it's not that significant on paper. My takeaway after yesterday is that we should throw money that we don't have at their D-Coordinator to get him to come here.

 

[slugboy]

I watched the Duke-ND game last night, and these stats seem about right. Duke's defense is elite, despite not having any superstars. While they may have a bit more talent than us at a few key positions, it's not that significant on paper. My takeaway after yesterday is that we should throw money that we don't have at their D-Coordinator to get him to come here.

I’m amazed BC and UVA are below us.

 

[roadkill]

I’m amazed BC and UVA are below us.

Indeed. I haven't watched any of their games except the BC-FSU one, which apparently was a fluke.

 

[takethepoints]

Here’s the composite that I liked last year, for week 5:


I like the per-conference graph he puts together here. GT has fought their way into the middle (ending the Ole Miss game in the third quarter would have probably moved us up a lot)


View attachment 14793

We’re at #67 overall, and you can see and download the numbers here: https://github.com/natemanzo/composite_ratings/blob/main/composite_ratings_2023-09-26.csv.

When you were watching football in the 70’s and 80’s, seeing the AP and the Coach’s poll was a big thing. You’d check the paper on Tuesday or Wednesday to see if your team had made the top 20. Now, there’s a wealth of info.

That is one really bad graphic. The print is too small, the arrows don't appear to lead much of anywhere (I know they were meant to, but …), and the comparisons are hard to make.

Well, these guys are football nerds, not real data analysts.

 

[takethepoints]

In general, I take football indices of all kinds with a boulder - a big one - of salt. I know how dicey indices can be; I've developed a few myself and I've never been satisfied. Since I never bet on football games I have never had much use for them. When I look at a game I'm guided by the teams records and who they played. Sometimes that's an easy call; I knew Oregon would beat Colorado, but I didn't know how bad.

So call me a luddite.

 

[Techster]

Confirms what we've seen so far from our team. Good offense, but bad defense.

 

[leatherneckjacket]

Confirms what we've seen so far from our team. Good offense, but bad defense.

Damn! Our defense stinks.

 

[LongforDodd]

Bright side…we’re scoring 1.6 more points per game than we’re giving up.

 

[takethepoints]

Confirms what we've seen so far from our team. Good offense, but bad defense.

Interesting, especially since having a good O is a poor predictor of having a good D. Good display too.

 

[roadkill]

Interesting, especially since having a good O is a poor predictor of having a good D. Good display too.

A positive I took away from this chart is that several of the teams we still have yet to play this season are in the upper left quadrant.

 

[iceeater1969]

A positive I took away from this chart is that several of the teams we still have yet to play this season are in the upper left quadrant.

This is a really nice clear chart. Would be great if there was a way add a 'tail" of last couple of weeks so u could see a trend. Probably get to congested if doing nation wide, but would be great for acc and remaing


However (Slug would need to be the judge) , aren't uga and mia are "way" into the good o and good d, while most of other remaing opponents are near the abcissa. I think that means they are average.



Imo (no ststs) our d can improve a lot with some minor improvements in player making right first step - too slow to begin closing on point of attsck.
After Mia, things will look a lot sslower when we play the average teams.

 

[slugboy]

This is a really nice clear chart. Would be great if there was a way add a 'tail" of last couple of weeks so u could see a trend. Probably get to congested if doing nation wide, but would be great for acc and remaing


However (Slug would need to be the judge) , aren't uga and mia are "way" into the good o and good d, while most of other remaing opponents are near the abcissa. I think that means they are average.



Imo (no ststs) our d can improve a lot with some minor improvements in player making right first step - too slow to begin closing on point of attsck.
After Mia, things will look a lot sslower when we play the average teams.

It’s not opponent adjusted, but Michigan looks good bottom right is where you want to be.

 

[slugboy]

For your Sunday morning:

And this is fun:

 

[slugboy]

And here’s this:
Miami still in the ACC title race (they only have one loss)


Upset (we’ll show up in “did we really lose that bad?” this week, on the other side of the for Miami)

 

[slugboy]

While we’re indulging in weird stats:


For future reference, it’s the GT-UM game, Miami’s odds of winning the game, in incredible whiplash.