re: SOG Model SEC Football Picks o' the Week brought to you by Jefferson Pilot Sports

SOG pick: A&M 27, Auburn 23
Actual: A&M 20, Auburn 3

SOG pick: Alabama 35, A&M 11
Actual: A&M 41, Alabama 38

quote:

---

SOG pick: A&M 27, Auburn 23
Actual: A&M 20, Auburn 3

SOG pick: Alabama 35, A&M 11
Actual: A&M 41, Alabama 38

---

JP Sports

quote:

---

Well if SIzg is right then you’ll know who to listen to when placing bets

---

Not me

Hey I would like MSU to win by 13 because frick AU but I just didn’t expect it to be that high

quote:

---

Ole Miss is quite beat up right? A healthy Ole Miss team might win but obviously the model can’t take into account injuries. I think A&M wins by 10+

---

Yep - I'd definitely lean A&M. Ole Miss' production has steadily declined the last month with the injuries and A&M is obviously playing better and better.

quote:

---

The MSU-AU one is shocking to me though. Not that MSU would win but by 13?

---

Auburn has struggled against on offense in basically every P5 game except for the 2 against teams that play umbrella coverage (Arkansas/Ole Miss).

@ Penn State - 20 pts, 367 yards, 4.65 YPP
@ LSU - 24 pts, 453 yards, 5.88 YPP
Georgia - 10 pts, 294 yards, 4.14 YPP
@ Texas A&M - 3 pts, 226 yards, 3.23 YPP
AVERAGE : 14.3 PPG, 335 yards

State is really good against the run (gives up 75% of opponent averages) and Auburn's run game is above average but not quite good enough to force their will on a good run defense (2.52 YPA vs A&M, 1.59 YPA vs UGA).

On the flip side, Auburn's pass defense isn't very good and they basically play bend don't break, something State should slice and dice.

My guess is it comes down to how each team performs in the red zone (dynamite analysis, I know). 13 seems pretty high to me, too. If I was just guessing a score I'd probably go 24-20 Auburn or something like that.

quote:

---

Actually his model had A&M losing to Auburn by .2. Effectively a jump ball. Like all models they are only as effective as the data they analyze.

---

Yea - it had it Auburn 21.1 - Texas A&M 20.9.

A&M defense played fantastic, held Auburn well below expectation. 40% under yards, 42% under YPP, 47% under Pass YPA, 33% under Rush YPA. And then, of course, they scored a TD themselves.

quote:

---

Does this model even account for defense? It seems almost entirely guided by offensive yardage regardless of opponent. Just curious.

---

Yea it does - takes offensive YPP vs P5 as a % of opponents YPP vs P5 (broken down by run/pass), then adds in the defense they are playings same numbers, gets an average.

So A&M vs Ole Miss
A&M Run Avg - 5.12 YPA
A&M Opponents Run Avg (not including A&M) - 4.18 YPA
A&M averages 122% opp avg

Ole Miss Defense same formula
Ole Miss gives up 4.22 Rush YPA, 101% of opponents avg

Average Difference 112% * ((4.22+5.12)/2)

A&M expectation = 5.22 YPA


That's the general formula (then same for pass). Then to get a point number I use yards per point broken out by run and pass. Then an average play formula.


All in all the actual yardage and YPP numbers end up being really accurate. The toughest thing to figure out is (a) Points per Yard/Yard Per Point and (b) number of plays. It's just not really something you can do uniformly, IMHO. You have to look at it game by game and make some individual decisions on how you think a game will play out and how each team will approach the game in terms of plan.

The model is not factoring in the progression that happens in a season. Players gel. Things click. Guys get healthy. I wouldn't expect some fun statical model to have this because how the hell would you quantify it on a large scale. But this is the kind of qualitative data that takes a good quanitative model and kicks it up a notch Emeril style.

The same thing happens in the opposite direct too. Injuries, lack of depth, in the middle of killer run in the schedule, etc.