I thought that myself when I read that article that talked about a Data Scientist(tm)
Okay, just a quick recap of the week that just passed. My predictor indicated UM would gain 522 yards of total offense. The offense gained 574 yards of offense, nearly 11% better than the predictor. The IU defense was predicted to gain 415 yards of total offense. They gained 568 yards, netting the UM defense at 136% of their projected output. Statistically, this wasn't their worst day (ND @ 144%), but it was still bad enough. Before the Indiana game, UMs defense was keeping teams to 99.34% of their normal yards on the season. That number is a fairly decent one. It took a huge hit this week and currently sits at 111.85%
On the brightside, UM's offense got better. Before the IU game, they were sitting at 154% total offense compared to what their opponents' defenses were giving up. With 574 yards of total offense, at 169% of IU's norm, the season average gained just over 6%. This margin kept UM above their opponents (MSU, Iowa) in the predictor. UM was also able to knock 12 yards off OSU's "lead" in their predictor.
As you can see above, I added the next metric into the equation: a scoring predictor. I've calculated two scores based on differing material. The hybrid yards/point includes all of 2009 and the completed games of 2010. One thing to note in the comparison between 2009 and 2010 for MSU is that their offense is scoring more frequently than last year. On the flip side, despite UM gaining a ton more yardage this year, they are actually scoring at a slower pace than 2009. Well, does this mean UMs offense is less explosive and MSUs offense is more explosive than they were a year ago? Probably not. In 2009, UM was one of the top 10 team in net punting. This enabled UM to have better starting position on each drive. Makes sense. UM 2010 has had pretty bad field position for most of the year. The positive with this is that if the defense can improve and got off the field without giving up field position, the offensive numbers could skyrocket.
My prediction based on limited stats and the hope that UMs defense shows up this week:
Interesting rank metric of teams UM will play/have played:
If you remember from last week's diary, (http://mgoblog.com/diaries/early-um-statistical-analysis), I predicted UM to finish the game with 763 yards of total offense, while at the same time holding BGSU to 343 yards of offense. My predictor proved to be fairly accurate despite starting with a small sample size. According to the results, UMs defense played particularly well, holding BGSU to 82% of their normal game output. This mark is the best of the season edging out their performance against UConn at 84%.
Let's move on to the matchup this week vs. Indiana.
Still to note is that UM is predicted to outgain all but one opponent on their remaining schedule. 5 of the 8 Big Ten teams on UMs schedule had their best offensive day this week, which closed the gap in several matchups including Iowa, MSU, and Wisconsin. Four of UMs Big Ten opponents had their best defensive day. If it weren't for UMs gaudy stat day, some of those teams may have overtaken UM on the predictions.
Despite UM having their best defensive day, stat wise, their overall percentage had a net -1.04% change. What effect will playing on the road have on UMs defense this weekend? Well, Ums defense has held every opponent under their season averages except for one; Yep, their lone away game against Notre Dame. Based on the percentage UM gave up against ND (137%), UM will give up over 570 yards against IU. Honestly, I don’t see that happening. Based on the opponent IU has played, I think their offense may be a bit overvalued. I’ll stick with my predictor though and go with…
UM - 522
IU - 415
Okay, so at first this diary started off as just another post at a different forum. But, one thing led to another, and here we are. The original post started off as a way to statistically justify why UMs defense was ranked #100 in the country. Without further ado, here we go.
These "rankings" are going to change. Especially when competition becomes tougher for the higher "ranked" defenses. For example, you've got a team like OSU sitting at #7 in total defense. They are a great defense, but their ranking is based on play against the #95 offense in Marshall, #65 offense in Miami, and #119 offense in Ohio. (rankings from ncaa.org)
On the flip side, UM faced the #57 offense in UConn, #24 in Notre Dame, and UMass, if placed in with the FBS schools, ranks #17 in the nation at 467 yds/game.
So, on average, UM faced the #33 offense in the nation, while OSU faced the #93 offense.
At this point, you're probably saying to yourself some of the same things I was questioning. Well, of course Notre Dame's offense is going to look good because they played against UMs defense. Well, just how much did UM effect those rankings?
ND would move from #24 to #48 without the UM game. UM held UMass under their season average. UMass would move from #17 to #15. UM held UConn to 50 less yards than their season average. They'd move from #57 to #41. So, UMs average offensive opponent taking out their UM game = #34. So, taking UM out of the equation pushed the rank of their opponents' offense from #33 to #34. Not much change. OSU was a different story.
Looking at OSU...
Marshall moves from #95 to #68
Miami has only played two games, but w/o OSU game move from #65 to #53
Ohio moves from #119 to #114.
OSU average opponents' offensive rank = #78. A move from #93.
So, when taking out the immediate matchup, UM was facing the #34 offense while OSU was facing the #78 offense. For comparison's sake, the #78 offense is Texas Tech at 345.67 yds/game. The #34 offense is Wake Forest at 430.67 yds/game.
How does this pertain to the rest of the schedule?
Based on the average of #34, UM will only face 3 more offenses better ranked than the average offense they've already played. (MSU, OSU, and Wisconsin at #28, #20, and #30, respectively)
Pushing the analysis further, UM is holding their opponents to 100.68% of their average offense.
UConn - 343/417 = 82.25%
ND - 535/409.5 = 130.64%
UMass - 439/481.5 = 91.17%
Total - 439/436 = 100.68%
OSU is holding their opponents offense to 68% of their average offense.
Marshall - 199/371 = 53.63%
Miami (YTM) - 352/405 = 86.91%
Ohio - 158/257 = 61.47%
Totals - 236.33/344.33 = 68.635%
If we take those numbers and look at the UM vs. OSU matchup at the end of the season, we get this in terms of expected offensive output:
UM: 350 yds
OSU: 463 yds
Of course, this same argument can be implemented to UMs offense vs. their opponents' defense. Without taking UM stats away from their opponents average, UM is still putting up 135.28% more yards than their opponent's defense normally gives up.
So, let's take away UMs impact on their opponents' defensive stats.
UConn - gave up 525 yds in their two other games. UM put up 473 yards on them. 473/262.5 = 180.19%
ND - gave up 799 yards to their two other opponents. UM put up 532 yards. 532/399.5 = 133.16%
UMass - gave up 544 yards to their other two opponents. UM put up 525 yards. 525/272 = 193.01%
So, to this point, UMs offense is putting up 163.81% more yards than their opponent typically gives up in a game.
In terms, of what this means vs. OSU... Ohio States defense is only giving up 236 yds/game. Based on UMs offensive output, they should put up 387 yards against OSU.
EDIT: So, I took all individual games and plugged them into Excel and came up with a pretty un-scientific predictor.
It is interesting to note that based on my predictor, UM will only be outgained by one team, OSU. I think these stats will really start to reign in the right picture once UM is two games deep in Big10 play.
We had over 180 participants in the "Prediction Survey". The results were a bit cheery, at 9-3. Perhaps everyone was a bit geeked up with the "Countdown to Kickoff" starting that day and various other prediction threads and hype videos. Without futher ado, here's the data...
At one point within the predictions thread, another user thought it'd be a good idea to provide a probability or likelyhood for a Michigan victory in each game. The online survey maker I used wouldn't tally the results in any valuable way, so I had to transfer all the data into an excel spreadsheet. I've uploaded the individual game results onto a Google Docs page. The link to that document is here:
Here are the results of the probability survey
Interesting, yet not entirely unexpected, the results came out to an 8-4 finish. What was surprising was the one game that flipped was the Penn State game. I was more expecting Notre Dame, or even perhaps the Michigan State game to be closer. Anyway, thanks for everyone's involvement. We had over 130 participants in the probability survey.