MGoPredictions2015 Survey Results
The results of the survey are in!
[More information here: link]
There were 669 responses and I removed 33 as outliers*. The first 143 responses were before I added the question asking to predict total wins, so there are 526 responses for that question, of which 29 were removed as outliers. Here is how the MGoCommunity rates our chances in each game:
| Game | Win % |
|---|---|
| 1) @ Utah | 54% |
| 2) Oregon State | 83% |
| 3) UNLV | 95% |
| 4) BYU | 67% |
| 5) @Maryland | 71% |
| 6) Northwestern | 79% |
| 7) Michigan State | 36% |
| 8) @Minnesota | 62% |
| 9) Rutgers | 82% |
| 10) @Indiana | 85% |
| 11) @Penn State | 56% |
| 12) Ohio State | 29% |
Predict the number of regular season wins:
| # of wins | # of responses | % |
|---|---|---|
| 5 | 2 | 0.4% |
| 6 | 9 | 1.8% |
| 7 | 51 | 10.3% |
| 8 | 188 | 37.8% |
| 9 | 176 | 35.4% |
| 10 | 62 | 12.5% |
| 11 | 9 | 1.8% |
I ran the Monte Carlo simulation and plotted the results alongside the results of question 2:
Note that the Monte Carlo simulation and the results both assign the highest probabilities to 8 wins followed by 9 wins. However, the spread is greater for the Monte Carlo than the whole-season predictions. I don't think that is very surprising.
I also plotted the cumulative probability function for each:
Interestingly, the expected value (P50) of the individual predictions are about a half game higher than the Monte Carlo simulation. Both predict we are more likely to win 8 or more games than 7 or less.
Utah Loss
If we factor in the Utah loss, the results of the simulation change slightly:
8 wins is still the most probable outcome, but 7 wins is now more likely than 9. 7 or less is more likely than 8 or more.
Comparison to other predictions
Dr. Ed Feng predicts 7.7 wins: Detroit News Link
| Game | MGo | Feng |
|---|---|---|
| 1) @ Utah | 54% | 40.4% |
| 2) Oregon State | 83% | 82.2% |
| 3) UNLV | 95% | 97.3% |
| 4) BYU | 67% | 67.2% |
| 5) @Maryland | 71% | 75.2% |
| 6) Northwestern | 79% | 77.7% |
| 7) Michigan State | 36% | 45.5% |
| 8) @Minnesota | 62% | 53.9% |
| 9) Rutgers | 82% | 83.3% |
| 10) @Indiana | 85% | 74.0% |
| 11) @Penn State | 56% | 46.1% |
| 12) Ohio State | 29% | 23.5% |
Interesting that so many match so closely! Games 2-6, 9, and 12 are all within 5% of each other. The outliers are Utah, MSU, Minnesota, Indiana, and Penn State. Feng surprisingly likes us more against MSU, but gives us a lower likelihood of beating the other four.
Angelique Chengelis predicts 7 wins with losses to Utah, Minnesota, Penn State, Michigan State, and Ohio State: Detroit News Link These are the lowest 5 ranked games in our results as well, with Minnesota @ 62% the highest win percentage.
I plan to add another simulation with Massey rankings here as well. If there are other predictions you think I should include in the comparison, please let me know in the comments.
*For those who are curious, my methodology for removing outliers was anyone assigning a win total of 12 or 0, anyone assigning 100% to Michigan State or Ohio State, and anyone assigning 0% to UNLV. If anyone else wants to play with the results, let me know and I'd be happy to send you the raw data.
September 8th, 2015 at 5:49 PM ^
September 8th, 2015 at 5:52 PM ^
September 8th, 2015 at 6:05 PM ^
Sent from MGoBlog HD for iPhone & iPad
September 8th, 2015 at 7:01 PM ^
If you just go by the first week, PSU looks easy, but NW looks like an auto-loss. There's lots of time and plenty of football to be played. I still worry about PSU because the Wolverines will be looking ahead to the first installment of Ten Year War v2.
September 8th, 2015 at 7:38 PM ^
Agree about the looking ahead on PSU. They look significantly easier with their no-pass-pro line not even making a UM 13-14 jump, but yeah, still worried about looking ahead.
NW does look a lot tougher, although I'm not ready to put them in the Top 25 after that win over the Cardinal. And since we just played a Top 25 team tough on the road in the opener, I'm not calling NW an auto-loss. But definitely a lot tougher. This may be that 1 out of 3/4 years that NW is a serious threat (no disrpesoect for Fitzgerald; I think that's a pretty good result for NW football).
September 8th, 2015 at 5:55 PM ^
Sent from MGoBlog HD for iPhone & iPad
September 8th, 2015 at 6:10 PM ^
September 8th, 2015 at 5:56 PM ^
September 9th, 2015 at 7:37 AM ^
Can't really say that because the "outliers" were removed.
September 8th, 2015 at 6:04 PM ^
Here's the estimated probabilities from Massey Ratings, just so we can compare this to one of the more established algorithms. It's actually pretty interesting to see as this algorithm doesn't appreciate some of the things that we might:
| Opponent | Result |
|
Utah
|
L
|
|
Oregon St
|
65%
|
|
UNLV
|
92%
|
|
BYU
|
52%
|
|
Maryland
|
44%
|
|
Northwestern
|
50%
|
|
Michigan St
|
19%
|
|
Minnesota
|
31%
|
|
Rutgers
|
59%
|
|
Indiana
|
62%
|
|
Penn St
|
50%
|
|
Ohio St
|
9%
|
September 8th, 2015 at 6:24 PM ^
Massey seems pretty down on us. Only 65% chance of beating Oregon State? 44% against Maryland?
September 8th, 2015 at 6:46 PM ^
I think the caveat here is that most of that is still based on recent historical performance which doesn't really appreciate the bump we may experience from Harbaugh and definitely - sadly - appreciates the long painful groan that was 2013 and 2014.
September 8th, 2015 at 6:53 PM ^
yea, he does a weighted average from the previous years ratings. Plus we dont know if coaching changes are involved in that,
September 8th, 2015 at 6:04 PM ^
I think Minnesota is overrated and will be exposed in the next 6 weeks.
September 8th, 2015 at 6:22 PM ^
September 8th, 2015 at 8:40 PM ^
Sent from MGoBlog HD for iPhone & iPad
September 8th, 2015 at 6:13 PM ^
September 8th, 2015 at 6:38 PM ^
I resent being dumped because of my 100% confidence in beating Ohio.
I did put 10% v MSC.
Every first year UM coach (except RR) defeated Ohio, and they all lost to MSC
September 8th, 2015 at 6:43 PM ^
Sent from MGoBlog HD for iPhone & iPad
September 8th, 2015 at 6:48 PM ^
see how the team improves week to week. I'm still hoping for a much improved team during the second half of the season which may make some upsets at home possible!
September 8th, 2015 at 6:49 PM ^
Outlier wise, I didnt understand the 0% with UNLV. I understand that theoretically speaking, but when you are forced to pick between 90% and 100%, and told to round to the closest, Im sure that its not as ridiculous. Besides, rounding up from lets say 96%, would only be an error of .04 wins, and when thats totaled with 600 other samples, wouldnt make a big difference.
However, I thought that the rest of the criteria was solid and it was interesting to see what we aspect from the season, so good work!
September 8th, 2015 at 8:29 PM ^
September 8th, 2015 at 8:50 PM ^
I took the "anyone assigning 0% to UNLV" to mean anyone giving UNLV a 0% chance but you're probably right
September 8th, 2015 at 7:47 PM ^
I think 8 wins would represent a really good coaching job.
Difficult schedule, new system, OL troubles seem to be partially personnel and not just prior coaching (prior coaching too, though, don't get me wrong).
September 8th, 2015 at 9:40 PM ^
I still see an 8 win team. Maybe 9, if everything falls right. I picked 8 wins at the start, and will stick with it, NW could be tough, but it's at home, and PSU should be easier, but that's on the road. I see 3-1 going into conference play.
September 8th, 2015 at 11:29 PM ^
Sent from MGoBlog HD for iPhone & iPad
September 8th, 2015 at 11:31 PM ^
Sent from MGoBlog HD for iPhone & iPad
September 9th, 2015 at 7:00 PM ^