Redefining a "successful" season
What is a "successful" season?
There has been a lot of discussion recently about what defines a "successful" season. Regarding this year, many arguments follow something like, "I predicted the regular season record would be between 8-4 and 10-2. So, given the youth and injuries, this season could be seen as decent if Michigan plays well against Wisconsin and OSU but loses, and 'successful' if they win at least one out of the two last games."
At face value, this seems reasonable enough. We all knew coming into the season that it was probably going to be a rebuilding year. Michigan has been decimated by key injuries. And winning 8 or 9 games...well, you may feel uneasy scoffing at that considering Michigan only did it twice during the era of Henri, the Otter of Ennui.
--Harbaugh finally started Henri on Prozac.--
But the fact is that record alone is a pretty meaningless metric for determining whether or not a season is successful. Nobody's preseason projections had Florida as a tire fire or Air Force as bad. Very few people predicted a loss to MSU (granted, they are better than expected). And if you had polled MGoBloggers before the season and asked, "How many wins against opponents with winning records will Michigan have going into the Wisconsin game," users answering 0 would have been negged to oblivion Bolivia.
Is a win always Win?
As with everything in life, defining success was so much easier back in the day. The steps were simple:
- Toe meets leather at high noon.
- Michigan stomps on inferior teams.
- Michigan beats Ohio State regularly.
- Michigan owns Michigan State.
- Michigan wins the Big Ten.
Under Bo, Michigan followed that Winning For Dummies formula to the tune of an overall record of 194-48-5, with 13 Big Ten titles, and 11/21 seasons featuring at least 10 wins. He went 11-9-1 against OSU and 17-4 against MSU.
--TaiStreet's Law: As a discussion about success at Michigan grows longer, the probability of a comparison involving Bo approaches 1.--
Credit (or blame) Bo for the Michigan fan's fixation on 10 wins as a basic measuring stick for a successful season. When Harbaugh took over for the floundering Hoke, Michigan hit the 10-win threshold his first season after the impressive demolition of Florida in the Outback Bowl. Harbaugh's team followed it it up with another 10-win season the next year, in which they were only an imaginary inch short of real glory. There are many valid reasons why Harbaugh's first two seasons are viewed as successful, but measuring up to that 10-win mark is certainly a main one.
Unfortunately, this simple metric leaves a couple pink elephants in the room. For one, Michigan plays more games now than they did during the Bo era. In the 9 seasons with 10 wins under Bo, Michigan's winning percentage was 86%. Compare that to the 77% mark of a 10-3 season. In the modern era, Michigan has an extra built-in cupcake win every season.
The next issue to discuss is that Bo was playing in the era of the Big 2, Little 8. Outside of Michigan and OSU, there were no consistently good Big Ten teams. Under Bo, Michigan played an average of 5.57 games/season against teams with winning records. Compare that to Harbaugh's first two seasons, where the number was 6.50 games/season. Now, there is typically 1 additional [at least decent] opponent for Michigan to prove itself against each season than there was under Bo.
The last major obstacle to making this an apples-to-apples comparison is results against rivals. As outlined above, Bo beat OSU with regularity and owned MSU. While his tenure at Michigan is still in its infancy, Harbaugh has yet to replicate those results.
With these contextual factors added to the discussion, it's becoming clear that there are several key components to measuring a successful season. I thus set out to create a standardized season scoring metric to get a better sense of how Michigan seasons compare to one another.
The Season Score Metric
--All data comes from Sports Reference. Note, they only record the AP poll rankings--
In order to create a standardized season score metric, I went through each Michigan season since Bo's first year and recorded the following:
- Wins
- Win %
- Quality wins: Wins against opponents that were ranked (in the AP poll) when they played Michigan AND finished with a winning record, OR wins against opponents that finished the season ranked (In the AP poll)
- Bad losses: Losses against opponents that were not ranked (in the AP poll) when they played Michigan AND were not ranked (in the AP poll) at the end of the season
- Wins against opponents that finished the season with winning records
- Win % against opponents that finished the season with winning records
- Results against OSU, MSU, and bowl game
- B1G championships
The season score is thus calculated as follows:
(Wins x Win%) + Quality Wins - Bad Losses + (Wins vs. Opponents with Winning Records x Win% vs. Opponents with Winning Records) + 5 if Beat Ohio + 3 if Beat MSU + 3 if Win Bowl + 5 if B1G Champion
To see this in action, here is how it looks for Michigan's best season, the 1997 championship:
- 12 wins x 100% Win%: 12 points
- 6 Quality Wins: +6 points
- 0 Bad Losses: -0 points
- 7 Wins vs. Opponents with Winning Records x 100% Win% vs. OwWR: +7 points
- Beat OSU: +5 points
- Beat MSU: +3 points
- Won Bowl: +3 points
- B1G Champions: +5 points
- Total: 41 points
Some major caveats apply to making this a true apples-to-apples comparison:
- Quality wins are based only on AP poll
- A win vs. the #1 team is treated the same as a win vs. the #25 team
- The relative strength of OSU and MSU in a given year is ignored
- The B1G is much more competitive now than in previous years
- Shared vs. outright B1G Titles are ignored
- All advanced stats are ignored
However, this metric still provides a ton more context to the season than record alone. And, at the end of the day, the Winning For Dummies formula hasn't changed at all, it has simply gotten harder to follow. Edit: see comment section for some more justification of this metric.
So, how does it all stack up?
(Here is a link to the google sheet with all of the information for anyone interested.)
| Coach | Seasons | Season Score | Wins | Win% | Quality Wins | Bad Losses | Wins vs. OwWR | Win% vs. OwWR | OSU W | MSU W | ND W | Bowl W | B1G Champ |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Moeller | 5 | 20.45 | 8.80 | 73.4% | 2.2 | 0.60 | 3.60 | 53.7% | 60% | 60% | 40% | 80% | 60% |
| Carr | 13 | 20.21 | 9.38 | 75.4% | 3.0 | 0.54 | 4.15 | 59.2% | 46% | 77% | 56% | 46% | 39% |
| Bo | 21 | 19.55 | 9.24 | 78.8% | 1.6 | 0.38 | 3.29 | 59.9% | 52% | 81% | 40% | 29% | 62% |
| Harbaugh | 2 | 15.09 | 10 | 77% | 3.0 | 0.50 | 3.50 | 53.6% | 0% | 50% | NA | 50% | 0% |
| Hoke | 4 | 8.51 | 7.75 | 60.4% | 1.0 | 2.0 | 3.25 | 38.4% | 25% | 25% | 50% | 33% | 0% |
| Rich Rod | 3 | 0.24 | 5 | 40.3% | 0.3 | 2.67 | 1.33 | 15.3% | 0% | 0% | 67% | 0% | 0% |
When you factor in the importance of beating your rivals, winning the Big Ten, and finishing the year strong with a bowl win, it becomes clear that win totals alone aren't enough to measure the success of a season. This shows that Bo, Carr, and Moeller had nearly identical success at Michigan, and they all prduced at a tier above what Harbaugh has managed so far.
Here are the ten best seasons by this metric:
| Year | Coch | Season Score | Record |
AP Rank |
Quality Wins | Bad Losses | Wins vs. OwWR | Losses vs. OwwR | Beat OSU | Beat MSU | Beat ND | B1G Champ | Bowl Game |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1997 | Carr | 41.00 | 12-0 | 1 | 6 | 0 | 7 | 0 | Y | Y | Y | Y | Y |
| 1980 | Bo | 29.97 | 10-2 | 4 | 3 | 0 | 4 | 2 | Y | Y | N | Y | Y |
| 1986 | Bo | 28.47 | 11-2 | 8 | 3 | 1 | 5 | 1 | Y | Y | Y | Y | N |
| 1988 | Bo | 27.88 | 9-2-1 | 4 | 2 | 0 | 5 | 2 | Y | Y | N | Y | Y |
| 2003 | Carr | 27.83 | 10-3 | 6 | 4 | 0 | 5 | 3 | Y | Y | Y | Y | N |
| 2000 | Carr | 27.55 | 9-3 | 11 | 3 | 0 | 3 | 2 | Y | Y | NA | Y | Y |
| 1989 | Bo | 26.97 | 10-2 | 7 | 3 | 0 | 4 | 2 | Y | Y | N | Y | N |
| 1991 | Moeller | 26.97 | 10-2 | 6 | 3 | 0 | 4 | 2 | Y | Y | Y | Y | N |
| 1990 | Moeller | 26.75 | 9-3 | 7 | 3 | 0 | 6 | 3 | Y | N | N | Y | Y |
| 1971 | Bo | 26.33 | 11-1 | 6 | 1 | 0 | 3 | 1 | Y | Y | NA | Y | N |
Harbaugh's 2015 team comes in at #28 w/ 15.99, and last year's team is at #34 w/ 14.20. For those of us who never watched Bo teams, this gives some better context to judge the Season Score metric. While last year's team was definitely a more talented team than the 2015 unit, it is not unreasonable based on end results for its overall success score to be worse.
For another comparison, Hoke's flukish 2011 squad is ranked #15 with a score of 23.75. Again, it's hard to believe that team was better than either the 2015 or 2016 Harbaugh squads, but there is something to be said for beating OSU (regardless of how bad they were) and winning a BCS bowl. What it does suggest, however, is that looking at the score metric by itself does not necessarily predict future seasons' success.
Projecting this season's score
Michigan is currently sitting at 8-2, with 0 quality wins, 0 bad losses, 0 wins against opponents with winning records in 2 games, and a loss to MSU. All that is good for a season score of 6.40.
But, there is theoretically a huge range of final success for this season. Just for fun, let's look at all of the possible end results:
| Rest of Season Results | Season Score* | Record | Rank (out of 49 seasons) |
|---|---|---|---|
| LLL | 4.92 | 8-5 | 43 |
| WLL | 7.43 | 9-4 | 42 |
| LLW | 9.43 | 9-4 | 42 |
| LWL | 12.41 | 9-4 | 36 |
| WLW | 13.49 | 10-3 | 36 |
| WWL | 15.49 | 10-3 | 29 |
| LWW | 18.49 | 10-3 | 24 |
| WWLL | 19.14 | 10-4 | 21 |
| WWLW | 21.93 | 11-3 | 18 |
| WWW | 22.11 | 11-2 | 18 |
| WWWL | 23.93 | 11-3 | 14 |
| WWWW | 30.86 | 12-2 | 2 |
*Note: I looked at the Massey game predictor to determine the likely outcomes for our previous opponents' final games. No previous opponent is expected to gain a winning record. However, in the highly unlikely scenario where Michigan wins the B1G East division, it would probably only be possible if one previous opponent does gain a winning record.
So now we have some real context for how relatively successful this season can finish. Obviously, the most likely outcome of the season will see Michigan finish (by some combination) between 8-5 and 10-3. Let's look at how those season scores compare to previous Michigan ones:
- A score of 4.92 puts the season score between those of the 8-4, 2001 Carr team and the 7-6, 2013 Hoke team.
- Scores of 7.43 or 9.43 puts the season score between those of the 8-5, 2012 Hoke team and that 2001 Carr team.
- Scores of 12.41 or 13.49 puts the season score between those of the 9-4, 1995 Carr team and the 8-2-2, 1975 Bo team.
- A score of 15.94 puts the season score between those of the 10-3, 2015 Michigan team and the 8-4, 1993 Moeller team.
- A score of 18.49 puts the season score between those of the 9-0-3, 1992 Moeller team and the 10-0-1, 1992 Bo team.
Tl;dr
With all of this information, I think it is reasonable to argue that this season can be called "successful" if Michigan does (at least) either of the following:
- Beat OSU
- Beat Wisconsin AND win the bowl game
November 15th, 2017 at 10:28 AM ^
That 1995 game was a pretty big upset. OSU went into the game undefeated and ranked #2. They had wins over six ranked opponents (BC, ND, Iowa, Penn State, Washington, Wisconsin), most of them by blowout scores. They had a ridiculous roster: Eddie George, Orlando Pace, Terry Glenn, Shawn Springs, Mike Vrabel... I realize Michigan had John Cooper's number but that Buckeye team was really freaking good.
Michigan had talent of its own but hadn't looked dominant all season, was starting a backup QB, and had just lost at Penn State to enter the game 8-3. I certainly don't think anybody had any inkling that Biakabatuka was about to go off for 300+ against OSU.
I still consider winning that game to have been one of Lloyd Carr's greatest achievements.
November 15th, 2017 at 3:54 AM ^
...for me success in a given season really just comes down to how you do relative to reasonable expectations, where:
Reasonable expetations = (what your program can expect historically in a generic year + what you actually have on your roster)/2
As a program, we expect to compete for conference championships. But our roster is made up almost entirely of first year starters--typicaly not a recipe for success. So a reasonable expectation might be to win 8 or 9 games. For me, a successful season is doing better than that--and the more we exceed that, the more successful it is.
I've been pessimistic about our chances this year since preseason, so I'll set 8 as my reasonable expectation. If we lose against Wisconsin and OSU, then the year is adequate as a rebuilding year but not more than that. Win either and I think it's a success.
November 15th, 2017 at 10:29 AM ^
November 15th, 2017 at 8:52 AM ^
Seeing Mo at the top of this list tears off an old scab. I don't believe we will ever fully realize what a tragic loss it was to M football to lose Coach Moeller. He was everything Bo was along with a riverboat gambler's mindset to make him to me the most complete coach we have had. Maybe 5 years wasn't enough to fully make that call. Again wei'll never know. But that one lousy incident, and the temporary loss of discretion (to my mind forgivable under the circumstances), screwed up what I believe would have been a historically great career. Meanwhile, other coaches go one to have long storied careers with far less character and integrity with respect to recruiting, player management, etc. Total BS and shame on the PC regime for not just giving him a big public reprimand and giving him a second chance.
November 15th, 2017 at 9:21 AM ^
nice selfie, too.
November 15th, 2017 at 12:38 PM ^
thanks Gary for chiming in here :)
More seriously, I too wonder what would have happened. However, it's hard to knock Lloyd's career - a ton of success and a national championship.
But I'll never forget Grbac to Desmond to clinch that ND game. What balls!
November 15th, 2017 at 1:13 PM ^
Lloyd did deliver the NC that eluded Bo.
November 15th, 2017 at 9:42 AM ^
your chart ends at WWWW. Should be WWWWW. Wins against Wisconsin, Ohio State, Big Championship, Playoff Semifinal, Playoff Championship.
538 says we have a 7% chance of winning out, and a 4% chance of making the playoff if we do.
November 15th, 2017 at 11:29 AM ^
I suppose it’s theoretically possible, and I already included a scenario that only has a 1% chance of happening, so why not add one with a, say 0.001% chance.
November 15th, 2017 at 1:35 PM ^
It still doesn’t quite sound accurate, but it isn’t as far off as I thought
November 15th, 2017 at 12:52 PM ^
Thanks. Way closer to a real comparison of seasons.
Tweak suggestions:
- split quality wins into two or three levels, Top 5, Top 10 and Top 25 with 3, 2, and 1 points respectively. For instance, let's say Florida had ended up in the top 5 this year and we beat them, I would consider that equal to winning a "bad" OSU team with two losses (see next item).
-apply a weighting mechanism to rivalry wins based on their record or national rank, i.e. beating a 2016 MSU team is worth more than beating an 80s MSU team. So any rivalry win could get a +1 (perhaps +2, depending) or corresponding negative points based on the quality of the win (or depravity of the loss). Not all rivalry wins are equal. A 1995 win over undefeated OSU could be worth 6 or 7 points vs. defeating a lowly 80s MSU that would be better measured as an extra 1 or 2 points.
-bowl wins should apply a similar weighting to rivalry wins
-play off wins should be +5 (someday Dog willing), a national championship should get +10
You also might think about doing something based on performance vs. expectations for a particular season. So much of how a season is perceived is based on the relative expectations at the outset. An M team that starts the season completely unranked and finishes in the top 10, or vice versa, dramatically impacts how fans think about the overall success of that season. As nebulous and useless as pre-season rankings are, they do typically represent seasonal expectations and adding some metric for more extreme swings, like unranked to top 10/20 or pre-season top 10/20 to unranked, gives some measure to whether a particular team exceed or failed at meeting those expectations.
November 15th, 2017 at 9:25 PM ^
He was undefeated in '92 and 3 very close plays could have made it a 12-0 season. Instead 3 ties that year 9-0-3.
November 15th, 2017 at 9:31 PM ^
I knew we would have a good D but the offense kind of looked average for a Michigan team. Plus, Carr had two consecutive 8-4 seasons. If we had this blog back then, I would have predicted the 97 season as 9-3 or 10-2 type season.
November 15th, 2017 at 10:21 PM ^
BEAT OHIO.
November 15th, 2017 at 10:49 PM ^
November 16th, 2017 at 11:34 AM ^
Bo never had a season as good as Dantonio's best two seasons ('13 and '15) and only the '97 year was better than either of them.
November 16th, 2017 at 11:45 AM ^
Anyone care to throw the best Ohio State seasons into this metric? How about MSU's 2013 season too?
November 17th, 2017 at 12:22 PM ^
As I mentioned in the opening post of the Fan Satisfaction Index series, I started on a project like this - didn't finish it - and then decided to survey people instead. I will be very interested to compare the season scores this model generates with what the satisfaction index reveals after the season is over.
One thing I have already found is that fans don't appear to use all the data from a season to determine their satisfaction/feelings in the present moment. Instead, they either use the last thing that happened or they twig on the best/worst thing that has happened.
This explains, I think, why so many of the comments here take issue with calling any season a success unless it involves beating OSU or winning the Big Ten.
Comments