/cdn.vox-cdn.com/uploads/chorus_image/image/47741119/usa-today-8942702.0.jpg)
You've undoubtedly heard all about the Washington State Cougars by now.
Correction: the ranked Washington State Cougars.
You read that correctly. Mike Leach and his Washington State Cougars are sitting pretty right now with an 8-3 overall record and the #20 spot in the College Football Playoff rankings. Not too shabby for a team that opened the season with a somewhat uninspiring loss to the Vikings of Portland State. That dark day led to a firestorm among the fanbase with our sister site even going so far as to analyze Mike Leach's contract situation to determine the financial consequences of letting their coach go.
Yet, here we are with Cougar nation having gone from up and arms to completely mollified. I offer as evidence two different editorials from Coug Center authored by two of their most experienced writers. In the first - following the PSU loss - Brian Floyd argues that it is time to panic. In the other, from just this week, Jeff Nusser argues that even if Luke Falk misses the game, he's not worried.
The comeback story with the Coug fanbase might be even better than the one of the team itself.
Since that dreary day in Pullman to open the season, it has been a good run for the Cougs. They've won eight of their last ten, scored over 30 points in seven of them and, importantly, played enough defense to prevent more than 30 points in six of them. That's the Mike Leach formula. It seems to be working.
With all of this success, it is somewhat shocking that the Washington Huskies are showing up in Bill Connelly's S&P projections as not only a favorite to beat WSU in the Apple Cup, but a prohibitive favorite. Per Connelly, the Huskies - who have a negative offensive FEI (meaning that on a strength of opponent adjusted basis the offense creates negative points per possession on offense) - have a projected win margin of over 11 points AND more than a 74% win probability.
Before we get into a crazy rabbit hole conversation about advanced stats and projections, it is important to acknowledge that, YES, we all agree that the game needs to be played on the field and, YES, there are limitations to any statistical exercise whose purpose purports to predict the future. All that you can say is that this kind of analysis establishes context for one to form expectations on how the game is most likely to turn out. Connelly has had remarkable success with this method. For the season, his teams that show up in the 70-79% win probability range have actually won 84% of the time.
So, that's good.
Beyond the analysis, however, lies a more interesting question. Why is WSU showing up as such a decided underdog? Keep in mind that this isn't Vegas we are talking about. This analysis isn't affected by subjective assessments or personal opinions. The crunching of objective data is turning out a decided result despite the fact that the Huskies have had a terrible time on offense all season, are 5-6 overall and are sporting true freshman in key positions all over the roster.
This is a question that merits some consideration as fans gear up for the Apple Cup.
I won't bore you with a breakdown of the actual numbers. Rather, I've observed three trends that I think, when boiled down to their underlying statistics, skew this particular model towards Washington.
WSU's Convenient Schedule
The Cougars have benefited greatly from a watered down schedule which, in turn has put a damper on the baseline statistics making up the core of Connelly's S&P based analysis. Recall that everything that they do with this method is adjusted for the strength of the opponent. The Cougars have four wins against very bad teams - at least statistically - in Rutgers, Wyoming, Oregon State and Colorado. They also have two wins against teams - Arizona State and Arizona - that are good wins but don't move the needle too much in terms of overall strength of schedule. The wins against Oregon and UCLA are the two bright spots on the schedule and serve to put upward pressure on that strength of schedule rating. Interestingly, the Cougs benefit from the fact that Vernon Adams contributions to the Ducks schedule overall is positive, but that they didn't have to face him in their victory over Oregon.
Missing from all of this is a win over a top 20 type of team. The Cougs lost to Stanford and missed two very tough matchups in Utah and USC in this year's rotating schedule. Of course, this is not WSU's fault and isn't to say that WSU would not have won one or both of those games. However, the Cougs do not benefit from missing them in the context of this type of analysis.
Compare and contrast this to UW who has played a more difficult schedule overall. A non-conference game against Boise State and conference matchups against Utah and USC add up to a stiffer level of competition than what WSU has faced with Rutgers, Colorado and UCLA. Add in the fact that UW has had tight losses (ASU, Oregon) to teams that WSU had tight wins against and you can see why WSU hasn't blown UW away in strength of schedule analysis.
The Name of the Game
We often talk here about how the only stat that matters is the final score. Indeed, the point of playing is to win and the way to win is to score more points than your opponent. Thus, score differential is a critical "basic" stat that many people pay attention to. It goes without saying that the larger the score differential, the better job the team is doing in consistently beating their competition.
WSU has had a bunch of nail-biters. Rutgers was a three point win. Oregon, Arizona and UCLA were all single digit affairs. Even their "easiest wins" - Oregon State and Wyoming - were by margins that were less than the experts would have predicted. Outside of Arizona State - which they won by 14 - WSU has had games too close for comfort (both in victory and defeat) all season.
UW, on the other hand, has benefited from two significant blowout victories with Arizona and Oregon State. They also played the good teams as closely in defeat as WSU did in victory. One common opponent - Cal - resulted in close defeats for both programs. Again, this doesn't mean that UW played "better" than WSU against those opponents. But it changes the math in the formulas underlying this statistic.
The Zen of Football
You will hear much this week about how WSU has resurrected their defense. And they have. Don't doubt that for one second. However, statistically, this means that they went from very bad to something resembling just ok. Consider that the Defensive FEI for WSU still has them ranked 46th in the nation. This means that on a "points contributed per possession basis", WSU's defense is 46th best in the country. Not bad at all. Their Special Teams, however, has not followed suit. That unit ranks 102nd in the nation in the same category.
WSU has been and remains offensively focused. This shows up in the numbers, but not as much as you think. Their offensive ranking in the efficiency category leaves them 36th in the nation. I can see the shock in your eyes through this screen. The reason for this is because the Cougs are often playing behind the chains, even when their "star" QB Luke Falk is manning the cockpit of the Coug offense.
Washington is a more balanced affair. It is true that the Husky offense is far worse - statistically - than WSU. Their offensive FEI ranking is 100th in the nation. However, their defense and special teams more than make up for it ranking 4th overall and 8th overall in the nation respectively. Simply stated, UW has more overall balance, at least as far as the numbers are concerned, across their three phases.
I wrote this article specifically to address the question as to why UW is showing up as a favorite in the statistical analysis. As we've all learned this season, there is no reason whatsoever to take the results of these analyses as gospel, in particular when you are talking about a rivalry game between two bitter foes. But the added context helps to establish an understanding of what is behind the records of each team and, hopefully, adds a little more drama to the game experience.
Data Points
The "Screw Advanced Statistics" crowd will get a kick out of this little oddity.
In case you missed it in the link above, the Oregon State / Washington game actually "broke" Bill Connelly's S&P model with a statistical oddity that warrants some mention.
Recall that the Connelly method uses per play data to make up its analysis. In particular, Connelly differentiates between a "successful play" (basically, a play that results in a team getting ahead of the sticks in terms of down and distance" and an "unsuccessful play". It then goes on to assess "how successful" a successful play really is for a team.
Washington, as you all know, dominated Oregon State last weekend. What you may not realize is that Washington prevented Oregon State from having any "successful" plays on offense.
That is, except for one.
The 76 yard rush by RB Paul Lucas was the only positive play for Oregon State during the entire game (the Connelly analysis rejects "garbage time" plays). Because it was the only successful play, it skewed Oregon State's "yardage average per successful" average dramatically - it was 76 yards! That same stat showed up on the opposite side of the ledge on UW's defense.
Oregon State had just 1 successful play against UW before garbage time adjustments kicked in … and it went for 76 yards.
— Bill Connelly (@SBN_BillC) November 24, 2015
I never worried about putting an IsoPPP cap in place because that didn’t happen in anything that I sampled. Didn’t cross my mind.
— Bill Connelly (@SBN_BillC) November 24, 2015
Alright, F/+, S&P+, Off. S&P+, and Def. S&P+ pages have all been updated to account for the UW-OSU glitch.
— Bill Connelly (@SBN_BillC) November 24, 2015
It is extremely unusual for any team of any caliber to have such a dominating defensive showing against another team of any other caliber to the extent that they only allowed them to get ahead of the chains on a single play over the course of the game. Obviously, a 76 yard average per "successful play" is an outlier and one that Connelly has now had to make an adjustment for.
Further proof that every statistical method has its exposed flanks.