Playoff Chase: Where Humans and Computers Agree in College Football


Clemson’s Shaq Lawson on Saturday. Last week, the computer rankings loved Clemson, and so did the playoff selection committee.

Richard Shiro/Associated Press

Cover your ears, Bowl Championship Series haters. Many computer rankings, which rely on statistics and algorithms to rate teams, tracked closely with how the committee’s 12 flesh-and-blood experts ended up listing the teams in the season’s first College Football Playoff ranking last Tuesday night.

The computer rankings were far closer to the official ranking than last week’s Associated Press poll (voted on by journalists) and the USA Today Coaches’ Poll. For example, a computer composite ranking and the official ranking each had Alabama at No. 4, while The A.P. and the coaches’ poll each had the Crimson Tide at No. 7 before Saturday’s games.

Along with expanding the playoff to four teams from two, the College Football Playoff was intended to improve on the B.C.S. by putting such decisions in the hands of human beings.

But as the College Playoff’s website states, “Selection committee members have flexibility to examine whatever data they believe is relevant to inform their decisions.”

The committee has also contracted with SportSource Analytics to provide extra data. An example of such data, the College Football Playoff’s executive director, Bill Hancock, suggested last year, might be rushing yards adjusted for opponent — exactly the sort of statistic that the computer algorithms are able to easily take into account.

On the other hand, several of the advanced rankings — Massey’s, for instance — include expected future wins and losses in their calculations. The committee has made it clear it does not do this, instead emphasizing to-date résumés and starting each week with a clean slate.

If the committee is making extensive use of these rankings, it is not letting on. The selection committee chairman, Jeff Long, in his appearances on ESPN and with reporters after rankings are released, has tended to cite more ordinary factors such as quality of opponent; whether a win was decisive (“game control” is his preferred jargon); and whether a victory came at home or on the road.

Still, the rankings’ close correlation to several advanced algorithms suggests the committee could be consulting them and perhaps reverse-engineering them — seeing that Alabama is so highly esteemed, then figuring out why (answer: three good non-home wins and a somewhat fluky loss).

And who was vindicated? Well, the Crimson Tide decisively beat L.S.U. at home Saturday, and in their wisdom the A.P. voters moved them from No. 7 to No. 3. (Don’t be surprised if the committee slots them at No. 2 Tuesday night.)

Perhaps it’s simply that 12 informed observers sitting in a conference room discussing all the teams for two days is producing results that more closely track computers than do votes by journalists on deadline and coaches with games to prepare for.

Either way, until they replace the players with robots (amateur robots, of course), all rankings will retain some human element.

Source link

Can't Get Enough SportsCoaster?

Sign up for SportsCoaster's newsletter and get the latest sports news straight to your inbox!