Portfolio Planning Made Simple

On to the Playoff!

Posted by Bela Barner on 12/22/16 4:40 PM

Well, that settled it.

Two weeks ago, we at Decision Lens were intrigued by the prospect of two potential upsets in conference championship games that threatened to turn the final College Football Rankings into a quagmire. At the time, there existed the very real prospect of the committee awarding two playoff slots to a pair of two-loss teams from a grab bag that would have included Clemson, Michigan, Washington, Colorado, the winner of Penn State vs. Wisconsin, and Oklahoma. We at Decision Lens do not wish difficult decisions upon anybody, but we do get a big kick out of sorting through challenging and important decisions, such as who should be awarded a precious playoff slot when the data suggests there are no obvious choices. It’s what we do.

As it turned out, both Clemson and Washington took care of business in their championship games against two very good opponents, and cemented their positions in the top four. The committee subsequently was left with the relatively easy choice of admitting the four remaining teams from power conferences that had fewer than two losses—Alabama, Clemson, Ohio State, and Washington.

Our model results, pictured below, agree with the committee selections. Alabama merits the stop slot based on an outstanding season (an interesting piece from fivethirtyeight.com judged Alabama in 2016 to be the best team in college football history, as in ever). Clemson and Ohio State each had very good seasons, losing once to ranked teams. These schools are effectively tied in our rankings, but our model gives Clemson the slightest of edges over Ohio State based on an additional Quality Win (a win over a CFP ranked opponent). By virtue of defeating Colorado, Washington retained its #4 slot.

Decision Lens College Playoff Rankings – Post Conference Championships
Criteria Weighted by Voters
December 15, 2016

image-1-nov-30

image-1-criteria_updated

We are pleased that, once again, our model appears to perform well in this highly debatable decision. But models, by definition, oversimplify matters, and there are a few results we can nitpick. Note that Colorado and Michigan, like Ohio State and Clemson, are effectively tied in our rankings—separated by only .003 points. Our model recognizes Colorado for two quality wins against a very tough schedule and no “bad” losses. However, Michigan defeated Colorado early in the season, and as the committee factors in head to head results while our model does not, one can argue that Michigan should be ranked ahead of Colorado, despite Michigan’s “bad” loss to unranked Iowa.

How damaging was to Michigan was its “bad” loss to 8-4 Iowa? In our model, it cost Michigan a playoff slot. The rankings below show the results of a season where Michigan avoids its upset at Iowa and finishes with 11 wins and no bad losses.

Decision Lens College Playoff Rankings – If Michigan Beats Iowa
Criteria Weighted by Voters
December 15, 2016

In reality, a Michigan defeat of Iowa still would have kept it out of the Big Ten Championship, so the committee would have faced the uncomfortable decision of selecting one team from two-loss Big Ten champion Penn State, one-loss Pac Twelve champion Washington, and one-loss Michigan. In this scenario, our best guess is that Washington still would have been awarded the final slot based on its previous ranking at #4 while Penn State sat at #7.

This is all speculation, but it shows how the “stickiness” or precedent set by prior week rankings by the committee might help it avoid very difficult decisions where recency bias confronts a season-long body of work. Our voters show a collective belief in the importance of Quality Wins (wins over CFP ranked opponents) and Strength of Schedule in determining overall ranking. There is no “right” set of criteria or weights for making a selection decision, but our model shows how hard data and judgment can work together to make decisions where hiding from scrutiny is not an option.

Enjoy the playoff!

 

Topics: General


Do More with Your Data
A blog about strategy execution problems and solutions.

We are passionate about problem solving and bringing your strategy execution to life.

Improve your real-time response to portfolio problems with industry leading innovation and case studies.

Readers of this blog:

  • Better define portfolio outcomes

  • Easily tie business metrics to strategy

  • React with the latest data + analytics