How do you solve a problem like McFadden, part II
WARNING: ERRONEOUS ZONE. LAWYER PLAYING WITH LOADED MATH.
This morning we looked at the fear that is running rampant through the hills of Tennessee this week concerning Darren McFadden's upcoming visit to Neyland Stadium. The fear's well-founded and based primarily on the following facts: (1) Arkansas' rushing offense is No. 2 in the nation; (2) McFadden and partner McFelix Jones hung 487 yards on Steve Spurrier last week; and (3) Tennessee's rush defense is ranked 74th in the nation.
But here's the thing. The rushing defenses of six of Arkansas' opponents this season are ranked 92nd or worse in the nation. Even allowing for the fact that part of the reason they're ranked so low is that they had to play Arkansas, that's still not very good. When Arkansas played Alabama (38th in rush defense) and Auburn (24th in rush defense), they lost. They also lost to the next worst rushing defense, Kentucky (92). So maybe, just maybe, that No. 2 ranking for rushing offense is a bit inflated.
|
ARKANSAS
|
||||
|
DATE
|
OPPONENT
|
RESULT/TIME
|
RUN DEFENSE RANK
|
RUN DEFENSE YARDS (per game)
|
| 09/01 | Troy | W 46-26 |
103
|
205.78
|
| 09/15 | at Alabama | L 41-38 |
38
|
126.44
|
| 09/22 | #21 Kentucky | L 42-29 |
92
|
192.67
|
| 09/29 | N Texas | W 66-7 |
107
|
209.38
|
| 10/06 | UT-Chatt | W 34-15 |
(I-AA)
|
(I-AA)
|
| 10/13 | #22 Auburn | L 9-7 |
24
|
114
|
| 10/20 | at Ole Miss | W 44-8 |
105
|
208
|
| 10/27 | Fla Int'l | W 58-10 |
99
|
204.11
|
| 11/03 | S Carolina | W 48-36 |
104
|
206.50
|
And what about Tennessee's national rank of 74th in rushing defense?
|
TENNESSEE
|
||||
| DATE | OPPONENT | RESULT/TIME | RUN OFFENSE RANK | RUN OFFENSE YARDS (per game) |
| 09/01 | at #12 California | L 45-31 |
T-50
|
166
|
| 09/08 | So Miss | W 39-19 |
28
|
190.78
|
| 09/15 | at #5 Florida | L 59-20 |
29
|
189.44
|
| 09/22 | Arkansas St | W 48-27 |
37
|
175.33
|
| 10/06 | #12 Georgia | W 35-14 |
40
|
173.67
|
| 10/13 | at Miss St | W 33-21 |
66
|
149.78
|
| 10/20 | at Alabama | L 41-17 |
62
|
153.33
|
| 10/27 | #16 S Carolina | W 27-24 |
95
|
121.60
|
| 11/03 | La Lafayette | W 59-7 |
8
|
241.56
|
Hmm. Again, we've done our part to make those stats what they are, certainly. But is it too far out of line to conclude that our poor run defense ranking is at least partially a result of playing against good running teams while Arkansas' exceptional run offense ranking is at least partially a product of playing against teams with a poor running defense? You tell me. I'm thinking that Arkansas is still an excellent running team -- the best we'll play this year, no doubt -- and that we're still not very good at stopping the run, but the numbers aren't as bad as they seem at first blush and the situation isn't as hopeless as it seemed earlier in the week.
Wouldn't it be wonderul if we could determine how many yards we should expect the Hogs to put up on us this week? That's what I thought. I figured it would be easy enough to look at the yards gained by the Hogs against their opponents and extrapolate the numbers to determine what we should expect the Hogs to gain against our No. 74 rush defense. Eventually, I figured out that I didn't want to "extrapolate" but "interpolate." No problem, I thought to myself, how hard can that be? Well . . .
Okay, then. Look, it's no great secret that lawyers chose law due to acute allergies to mathematics, so let's do the poor man's interpolation, where poor = "dumb at numbers".
Test Case! Tennessee's currently ranked (very!) roughly midway between Alabama and Kentucky in rush defense nationally. Alabama's giving up 126.44 yards per game and Kentucky's giving up 192.67 yards per game. So we should be giving up roughly, what, 166 yard per game? Checking . . . it's actually 167.67.
|
TEAM
|
RANK
|
YARDS
|
| Best (Boston College) | 1 | 58 |
| AR opponent (Auburn) | 24 | 114 |
| AR opponent (Alabama) | 38 | 126.44 |
| Tennessee | 74 |
in sealed envelope . . . 167.67, which is very close to 166 according to my calculator!
|
| AR opponent (Kentucky) | 92 | 192.67 |
| AR opponent (Florida Int'l) | 99 | 204.11 |
| AR opponent (Troy) | 103 | 205.78 |
| AR opponent (South Carolina) | 104 | 206.5 |
| AR opponent (Ole Miss) | 105 | 208 |
| AR opponent (North Texas) | 107 | 209.38 |
| Worst (UAB) | 119 | 248.7 |
Okay, so so far so good, and my, that's a lot of sos. But the system works sort of. So now for the real thing. Alabama gave up 301 yards to Arkansas, and Kentucky gave up 338. And boy am I glad that Kentucky didn't actually give up fewer yards than did Alabama or my head would have exploded and this post would have been entitled: Using Math to Paint Yourself Into a Corner.
Anyway, the magic number for the Vols this weekend appears to be . . . drumroll . . . about 318. As in yards the Razorbacks should get against our defense on the ground. Which is not good, but not entirely not good. The three teams that have rushed for the most yards against us are the Bears (230) (beatdown), the Gators (255) (blowout), and the Ragin' Cajuns (234) (blowout in the other direction). So although it's unlikely, Tennessee could win the game even if Arkansas gets 318 yards on the ground. Remember, Georgia is currently averaging 173.67 yards per game, and we held them to 69. Let's do that again, shall we?
Apologies to hooper, our resident nuclear engineering grad student, who must be banging his head against his monitor right now. Somebody get him some BC.
8 comments
|
0 recs |
Do you like this story?
Comments
It's a convincing argument
One thing to remember
by flahog on Nov 7, 2007 2:23 PM EST reply actions
The way I see it....
I really think the key to winning this game is simply to keep Arkansas' offense on the sideline. That means long, drawn out drives with lots of really good intermediate routes and excellent running. Perhaps some screens or draws. Wait a minute, sorry, I forgot who I was talking about.
Either way, the defense must play exceptionally well to keep McFadJones in check, but if the offense plays like they should, we shouldn't see too much of them anyway.
Oh yeah, and please, for Pete's sake, DOUBLE UP ON MONK. I hope he doesn't become DJ Hall pt. 2.
Here we come
Thanks-
by RockyTopbound on Nov 7, 2007 4:22 PM EST reply actions
Waaaaaaaahhh!
by ghostofneyland on Nov 7, 2007 4:33 PM EST reply actions
Good analysis :)
The geek in me just can't help but point out that Arkansas has an unusual breaking point at 200 rushing yards a game. If their opponents average more than that, they've won. Less than that, they've lost. (The Kentucky game really makes a fine line of it.) If you believe that, then Tennessee is on the golden road.
Another P.O.V. - we must be really concerned about this game if a lawyer is reduced to doing math.
Good Result, Erroneous Math
The assumption made when interpolating is that the data behaves in a linear manner. In the first case you use, run defense ranking vs. rushing yards allowed, this assumption is correct. Plotting this data will show that it behaves linearly until we reach the data at the edges of the set, i.e. the top ten defenses and the bottom ten defenses. However, when plotting the rankings of the defenses Arkansas has played vs. Arkansas rushing yards, it is readily apparent that this is not the case. This can more easily be seen by just looking at the results from Mississippi and Florida International who allowed fewer yards against Arkansas than either Alabama or Kentucky did.
(End Boring Math Explanation)
Now for the good news, or bad news for you Joel because you did all this work, Tennessee allowing a little over 300 yards rushing is a very reasonable estimate. With the limited sample size, it is impossible to use numerical methods. Your best bet would have probably been to exclude the Auburn and South Carolina games, due to them being outliers, and then just calculate the average. Doing this gives you a result of 319.14 yards per game rushing for Arkansas. This simple method gives a result that is just as mathematically significant as interpolation in this case without all of the unnecessary work.
Math fight! Yay!
Even including the ends, the linear model is decent enough (R-square of 0.56 and F-test of 0.03). At any rate, there's simply no getting around the fact that playing with stats on a sample size of 8 is sketchy at best. The ranking is likely to be a poor choice for predictor because it takes a perfectly good continuous predictor (yardage), removes the relative differences, and turns it into a boring ordinal.
So we agree, though I'm more tolerant of the linear model.
by David Hooper on Nov 7, 2007 10:37 PM EST up reply actions

by 
























