Mid-Season Advanced Stats Review (Part I)

Is this real life?
It’s hard to believe the College Football regular season is already half over for the Washington Huskies. It is even harder to believe that the Huskies are sitting at 6-0 and have, to this point, lived up to every bit of pre-season hype that was heaped upon them by the national media. But here we are.

I thought this mid-season Bye week would be a good time to take stock of what has happened so far and the outlook for the second half of the season, from an advanced stats perspective. In the first part, we’ll look at some high level metrics on the Huskies and their past and future opponents. In a second installment, I plan to delve more deeply into what the statistics are telling us about the Washington offense and defense and try to put those numbers into context.

But there is one other matter to attend to before we get started...

“Do you know what ‘nemesis’ means?”
(Link Contains NSFW language)

The complete evisceration of the Oregon Ducks last Saturday doesn’t make up for a 12 game losing streak, not by a long shot, but it was a very good start.

And while it was certainly gratifying to see the Huskies score (and score, and score,...and score), the best part, for me, was watching the physical play on both sides of the ball absolutely crush the Webfoots’ Ducks’ will to compete.

Back to the matter at hand...

Framing What the Huskies Have Done So Far
If you are on this site, reading this article, you are doubtless aware that the Huskies are undefeated, with only one game having been remotely competitive. The 6-0 start is remarkable itself, as it hasn’t happened since the Bush Administration. The George H.W. Bush Administration. As in 1992.

You may have heard it mentioned on a telecast that the Huskies are on a nine game winning streak, the second longest in the country among FBS teams behind Alabama.

You might be aware that statistically the Huskies have been quite dominant this season. How dominant?

• #5 in FBS in Scoring
• #6 in FBS in Scoring Defense
• #1 in FBS in Turnover Margin
• #1 in FBS in Offensive Passing Efficiency

These are all great things. And they tell some of the story, but they don’t tell all of the story.

Slow Your Roll (at least a bit)
The raw numbers this season and dating back to the last three games of last season are impressive indeed: nine wins; zero losses; and an average of almost 49 points per game for the Huskies, versus just under 15 points per game for their opponents.

While it is no secret that Washington’s Out of Conference schedule was quite weak this season, it is also an uncomfortable truth that the schedule over the entirety of the aforementioned win streak has been pillow soft.

In terms of Jeff Sagarin’s Predictor (or Pure Points) measure, the average rank of the opponents during the UW’s win streak is #87. That average is certainly dragged down by the OOC schedule early this season and the victories over most of the better teams (relatively speaking) have been gratifyingly decisive, but the fact remains that the schedule has been extremely accomodating. Another thing to consider is trajectory of the teams: at this point I would take the under on the current Sagarin Predictor for all of the UW’s FBS opponents this season, except for maybe Arizona.

If you ever wondered what it would look like if you dropped a well-coached, Power 5 team with upper shelf talent into the MAC or MWC, now you know.

I was told there would be no math
Fortunately there are ways to determine how much of the UW’s recent good fortune has to do with the ineptitude of its opponents versus its own good play. Without getting into too much minutiae, there are a number of well-developed systems for rating college football teams on the basis of their on-the-field results that adjust for the quality of their opponents. For lack of a better term, these systems all fall under the broad umbrella of “advanced stats”.

Jeff Sagarin has been producing computer ratings of college basketball and football teams since the mid-80s. His ratings are based on game results and account for quality of opponent. While Sagarin produces several different types of ratings his Predictor rating (aka Pure Points) takes into account margins of victory as well as strength of opponent. Although utilizing scoring margins isn’t politically correct, those margins contain useful information. He does apply a diminishing returns function such that beating another team by, for instance, 50 points isn’t twice as significant as beating a team by 25. Although results always vary from year-to-year Sagarin’s Predictor has historically slightly outperformed the point spread. The long history of the Sagarin ratings and the fact he actually ranks all FBS and FCS teams makes his ratings particularly relevant.

Brian Fremeau developed the FEI ratings system which is published on the Football Outsiders website, as well as Fremeau’s own BCF Toys website. The FEI system rates FBS teams on the basis of offensive and defensive efficiency, with the drive (or possession) as the basic unit of measure. He provides data both adjusted for opponent quality and unadjusted. All of his data excludes drives that occur during “garbage time” of uncompetitive games. The projections based on FEI have been particularly effective versus the spread in recent years.

Bill Connelly is the publisher of the excellent SB Nation Football Study Hall website. He developed the S&P+ rating system, which is quite similar to FEI, except that it examines per play efficiency rather than per possession efficiency. His data is also available at Football Outsiders. The game projections based on S&P+ have generally beaten the spread by a couple of percentage points over the last several years.

Sports media giant ESPN produces ratings called the Football Power Index (FPI). There isn’t too much information available on how these ratings are calculated, beyond the fact that they are based on measures of offensive, defensive and special teams efficiency. They are included here mainly because they have proven to be extremely accurate against the spread this season.

The 2016 Huskies: Cross Section of Advanced Stats Ratings

Despite the soft schedule, all of the various systems described above are in general agreement that the Huskies are indeed a very strong team. The only teams ranked ahead of the UW on any of the above systems are Alabama, Ohio State, Clemson, Michigan, Louisville, Texas A&M, and Oklahoma (?!?).

OU only appears above the Huskies on S&P+, presumably because pre-season projections still account for a portion of the S&P+ rating . By the end of this weekend, those projections will no longer be considered by the model. The S&P+ model was high on the Sooners pre-season, ranking them #4 overall. Although both of OU’s losses are “quality losses” (to Houston and Ohio State), it seems certain they will sink below the Huskies following this weekend after pre-season projections are phased out of the model.

Looking Ahead
Based on the various advanced stats measures, Washington’s next opponent, Oregon State, is just about in line with the average quality of opponent the Huskies have faced over their nine game winning streak. But after that the schedule gets considerably more challenging.

Washington’s Remaining Opponents - Advanced Stats Ratings

Clearly WSU and Southern Cal stand out as the most highly rated upcoming opponents, with Utah falling in somewhere behind them. The back-to-back road games against Utah and Cal are also notable, as the Huskies haven’t played two in a row on the road since last season, when they played ASU (frustrating loss) and Oregon State (blowout win) in consecutive weeks.

Of course there will continue to be movement in these ratings as more games are played, but as things stand right now, it looks as if Washington should be favored by a decent margin in every remaining game on the schedule.

“Games aren’t played on paper, Poindexter.”
The S&P+ system, which has the Huskies ranked lowest of the four described, currently suggests a regular season record of 11-1 is most likely, at 39% probability. It pegs the odds of an undefeated regular season at 24%, for a combined probability of going 11-1 or better at 63%. Not bad.

ESPN’s FPI system projects that Washington has a 30% of going undefeated in the regular season and a 67% chance of winning the conference.

Although FEI and Sagarin don’t provide simulated win probabilities, based on where the Huskies and their upcoming opponents rank on their respective systems, the probability of an undefeated season based on those systems is probably +/- 30% as well.

Of course probability projections are just that...projections. They are both an inexact science and also constantly evolving as more data (i.e. games) is input into the models.

And of course there are questions that no amount of data science and simulation can ever answer. Will this still young team stay focused and continue to “Prove It” each week? Will upcoming games be close and therefore subject to outlier events and the whims of Lady Luck? Can the team continue to avoid key injuries?

That’s why they play the games, as the saying goes. It is also why college football is a lot more fun than a graduate level statistics course.