this is freakin sweet. I wish I could do this (and I hope you're right).
Post Week 8: Yardage Analysis and Predictions + Score Predictor
Just a quick refresher from week 7 game vs. Iowa. Diary can be found here: http://mgoblog.com/diaries/post-week-6-yardage-analysis-and-predictions-... . UM's offense and defense both came to play against Iowa. The offense was predicted to put up 407 yards of offense. They put up 522 yards. This was UMs best offensive day of the season, posting a 200% of Iowa's season average. This moves UM season average up to gaining nearly 159% of what their opponents typically give up.
The defense also had a stellar day giving up 383 of the predicted 454 yards of Iowa offense. Statistically for the defense, this was their best game of the season. They held Iowa to 91% of their season output. This was the third time this season the UM defense held their opponent under their season average.
The score predictor nearly hit a grand slam two weekends ago. I predicted a 35-28 UM loss. The final score was 38-28 Iowa. Despite UM having a great day moving the ball and having a relatively decent day holding Iowa in yardage, they did not succeed where it counts: on the scoreboard. Iowa was scoring 1 point every 12.69 yards on the season. In the Michigan game, they scored 1 point every 10.08 yards; well above their season average. Iowa had damn good field position all game and plays like Sash made on the block FG returning it past midfield, really hurt UMs chances in this ballgame. Not to mention the TWO!!! kickoffs out of bounds.
Where was UM on yards per point? For the season, UM was scoring a point every 13.75 yards. During the Iowa game, UM scored one point every 18.64 yards. A product, no doubt, of penalties and turnovers. UM, after the Iowa game, is score a point every 14.78 yards. Something interesting to note is that UM was scoring a point every 13.03 yards in 2009. Despte that, UM is on pace to score 78 more points this season than last. Go figure.
What does this mean moving forward?
It all has to be positive. I mean, the DEFENSE HAD THEIR BEST STATISTICAL GAME OF THE SEASON!!! That alone is reason to celebrate. On top of that, UMs offense had their best game as well. If UM can limit their turnovers the rest of the season, or they can create some of defense, then I feel comfortable saying that there isn't a team on UMs remaining schedule that they can't beat.
Anyway, let's move on to week 9. Game at PSU...
Charts first? Yes... CHARTS
Well, with the bye week, UMs stats did not change. However, PSU was able to help their cause a little bit. They put up their second highest point total on their fourth lowest yardage total. That increased the predicted point total by 1 point.
UM still remains a favorite to outgain their opponent in all of their remaining games besides OSU. They are a -113 yard dog to OSU. However, UM was a 48 yard dog to Iowa and out gained them by 139 yards, for a 187 yard turnaround, so anything is possible.
UM - 527 yards
PSU - 369 yards
Score Predictor based on statistics...
UM - 35
PSU - 24
My score predictor based on gut feeling and blue shades - UM 42-17
I just took the prediction down the next avenue. I've calculated a percentage error of my predictions and then formulated a high and low for yardage output and their corresponding point totals.
"And the defense also had a stellar day giving up 383 of the predicted 454 yards." Wow, you'd almost think we won. I'm not sure I was yelling "stellar" while I watched the game.
Yeah, stellar for this defense. When your defense is giving up 115% and holds a pretty good offense to 90%, I'd call that a good day from that group.
Also, Iowa is the 4th best offense (yardage total) that UM will face this year. The only better offense they face the rest of the way is OSU at #2.
Princeton is right. The turnovers and special teams gaffes put Iowa on short fields all day; as a result, they didn't have to gain many yards to score. Sadly, there's not much reason to think that our D would have prevented long drives and scores if Iowa's average starting field position has been, say, their 20. So I think the stats are lying here.
That's a fair assessment.
Statistics can lie, as we all know. Based on statistics, yes, we held Iowa to less than 400 yards. However, a big part of the reason for this was that Iowa went to their Prevent offense, which we all know and love from the Carr era, pretty much from the start of the 3rd quarter. At this point they were moving the ball at will, after a rough start. Drive summary:
- 3 and out!!
- 3 and out!! (woohoo, I'm sure I wasn't the only one who was fist-pumping at this stage)
- 84 yards for a TD
- 51 yards, TD
- 52 yards, TD
- First drive of 2nd half - up 21-7, run, safe 4-yard out, run on 3rd down - 3 and out
- Still 21-7, inside zone run, inside zone run, pass - 3 and out
- Ok, let's score again. Pass (34 yards), Pass (11 yards), run, pass, run for 62 yards and a TD.
- Back to prevent offense, but they score! Inside zone run, inside zone run, inside zone run, then a 19 yard slant for a TD on 3rd and 4. 60 yards incl the 15 yard penalty.
- 35-21, inside zone run, inside zone run, false start penalty, zone stretch run. 3 and out.
- 35-28, they need to move the ball and preferrably score. Waggle TE out (25 yards), waggle TE out (10 yards, but on 1st and 15), Inside zone run, Slant pass on 3rd and 9 for 17 yards, zone stretch run, waggle keeper, disaster Avery on dumpoff pass for 26 yards on 3rd and 8, - ok, we're safe, so inside zone run, inside zone run, incomplete pass, FG.
Without the prevent offense, this would've been much uglier. Stats lie in this case.
So the fact that our defense actually slowed down the prevent offense means nothing? What if Iowa was more aggressive and committed a TO on another Stanziball? You can't really deal with with the hypothetical.
You're right, of course, we have to work with actual numbers, not hypothetical numbers. As long as we know their limitations, so we don't abuse the numbers to reach faulty conclusions. In this case, I would argue that the statistics overrate our defensive performance vs Iowa, as Iowa's main offensive goals shifted from scoring as many points as possible to also include chewing up the clock and avoiding costly mistakes.
We managed to limit the yardage production of Iowa's offense. However, our success in doing so was less than it seems on the surface, as the offense pretty much told us exactly what they were going to do and dared us to stop it. Some of the time we did, which is great, but a couple of times we didn't. When they did something besides the obvious, they succeeded at an alarming rate.
Also, we're not even considering field position here. As others have pointed out, we never looked like stopping Iowa on those TD drives in the first half, which were relatively short, limiting their total yardage production.
What conclusions can we draw from this? Mostly just that our "stellar" defensive performance, limiting Iowa to less than their average offensive production, was probably a few notches below stellar. Not groundbreaking news for any of us, I'm sure... :(
I hate statistics with a fiery passion, but I like what these statistics are telling me. Thanks for the hard work!
You have to have data points first. More like a mistress fellatio predictor.
"I am afraid to say anything"
Is hold on to the the DAMN ball. And with the steady improvement in the punting game from Hagerup, we can start to even out the field position battle.
There are at least 3 more wins sitting out there before the calendar flips over to December. Lets enjoy the ride....
the defense out performed. Now if the offense and special teams would not have made mistakes we could have won the IA game, may be even the MSU game!
I just hope the pressure of a big night game on the road, and the pressure of that 6th win will not cause any big mistakes. The team just needs to go out and play hard and have fun! It will be interesting to see how the predictor does this Saturday. I think it will be a much closer score. UM 35 vs PSU 31.
We need to win this game in the first half by getting 300+ yards and 21 or more points and yield no more than 3 to them. Then we get to play prevent defense and milk the clock on offense in the second half. We need a decisive win against a wounded opponent to restore a winning attitude for the following 4 games. I do not foresee wins against Wisconsin and OSU but I do see us staying in these bigger games if we can elevate our confidence level starting this weekend.
Should be a good weekend of football
Take the number of yards predicted for the offense and use the yards/point of the defense and THEN predict the outcome of the game.
(Hope this is clear, I can clarify if you need).
Defenses that create a lot of turnovers will be better modeled with this predictive scheme. This will also account for bend-but-dont-break defenses that give up lots of yards but not too many points. NOTE: Jokes about our "bend-over" defense will not be appreciated.
Your prediction is actually (loosely put) using two random variables without a joint probability term right now.
That we win even if PSU's score is on the high end of the error margin and M's score is on the low end.
love the excel screenshot. excel >>>>> SAS
After this excellent analysis, I am not even going to watch the game! Oh, wait, my in laws are in town ... check that, I will be watching and trying to keep up with a beer shot per michigan point.
my plane leaves in 17 hours. I have began pacing. I need something to make time go faster.
this is awesome info. thanks for all you do. i hope your right. Go Blue!