With 58 seconds left in regulation at Heinz Field, Miami backup quarterback Jarren Williams found KJ Osborn along the left hashmark. He slipped a tackle and burst upfield, scoring from 32 yards out to give the Hurricanes a 16-12 "upset" win over Pat Narduzzi's Pittsburgh Panthers.
Why the quotation marks? Because if you were looking at the right numbers, it might not have been an upset at all. Those same numbers suggest SMU could be in some trouble this coming weekend, as well.
As with many tools in sports analytics, the idea of my second-order-wins concept originates with Bill James. In his early, self-published "Baseball Abstracts," James espoused the ideas that would become a standard of the analytics toolbox: Pythagorean wins.
The idea is pretty simple, really: By looking at runs scored and runs allowed, you can devise a pretty good look at what a team's win percentage should be. Compare it to what it actually is, and you end up with a list of teams that might be a bit lucky or unlucky over a given period of time. The idea can be used for any sport once you figure out the right coefficient to use.
Baseball Prospectus took this concept a bit further, developing what they called Second Order wins, a Pythagorean figure based on how many runs you should have scored and allowed (according to typical production factors such as hits, hits allowed, etc.). Like football itself, analytics are a copycat sport, so I long ago began exploring my own second-order wins concept.
You can obviously win with a low postgame win expectancy or lose with a high one, but over time, it tends to regress toward the mean. (Some coaches prove exceptions to this rule, but it takes a large sample to figure that out.)
If you add up a team's postgame win expectancies, you get what amounts to an average expected win total, or my version of second-order wins. Over the course of a full season, this paints a pretty clear picture of good or bad fortune. And fortune tends to right itself.
Last year, Army and Northwestern had second-order win totals of just 8.1 and 6.2, respectively, but went 11-2 and 9-5; this year, they've gone a combined 4-11. Kentucky, Georgia Southern, Texas, Notre Dame and Syracuse also overachieved their second-order win total by quite a bit last season; they are not overachieving this year.
So what can second-order wins tell us about teams we might be under- or overestimating heading into 2019's stretch run?
Biggest underachievers (wins vs. second-order wins)
1. Iowa State (5-3 record, 6.8 second-order wins)
2. Miami (4-4, 5.7)
3. New Mexico State (0-8, 1.4)
4. Washington State (4-4, 5.2)
5. Purdue (2-6, 3.2)
6. Kansas (3-5, 4.1)
7. Washington (5-3, 6.1)
8. Fresno State (3-4, 4.1)
9. Tulsa (2-6, 3.0)
10. Rice (0-8, 0.9)
What a year for Iowa State to be snake-bitten. ("We're always snake-bitten," ISU fans are now responding in their heads. I understand, but still.) The Cyclones have now lost three games to SP+ top-30 teams by a total of 10 points. They suffered one of the least likely losses of the season to Iowa in Week 3 -- they had a 97% postgame win expectancy due to a plus-105-yardage advantage, a success-rate advantage and a large big-play advantage but lost because of unsustainably bad red zone execution and some bad luck with turnovers.
Their other two losses (23-21 to Baylor with a 38% postgame win expectancy and 34-27 to OSU at 44%) were statistical tossups, while all their wins were comprehensive, i.e., had a 99-100% postgame win expectancy.
Iowa State is a top-20 team per SP+; if the Cyclones finish there, it would be only the third time since World War II. Though their record is decidedly not top-20 caliber, they could give Oklahoma major trouble (again) in Norman in two weeks, and SP+ still gives them a 49% chance of getting to 8-4 or better.
Really, the theme among all of these teams is bad close-game outcomes. These 10 teams are a combined 8-27 in games decided by one possession. As Miami proved on Saturday, that theme can change at any point. And despite Saturday's good fortune, the Canes are still owed a bit more. Keep that in mind for their Saturday trip to Florida State -- the teams share the same record, but FSU's second-order-win total is 4.4, and SP+ again projects Miami as a slight favorite.
The 10 teams on the list below, by the way: 28-2 in one-score games. How sustainable do you figure that is?
Other games featuring these 10 teams: Nebraska at Purdue (+2.5), Utah at Washington (+3.5), Fresno State (+2.5) at Hawaii, Tulsa (+10.5) at Tulane, and Marshall at Rice (+10.5).
Biggest overachievers (wins vs. second-order wins)
1. SMU (8-0 record, 5.7 second-order wins)
2. Wake Forest (6-1, 4.2)
3. Toledo (5-3, 3.3)
4. Louisiana Tech (7-1, 5.4)
5. San Diego State (7-1, 5.4)
6. Miami (Ohio) (4-4, 2.6)
7. Nevada (4-4, 2.6)
8. Stanford (4-4, 2.6)
9. Minnesota (8-0, 6.7)
10. Appalachian State (7-0, 5.7)
It's always difficult trying to balance enjoying an incredible run and understanding what is probably to come. SMU's second-year surge under Sonny Dykes has been incredibly fun, as is the thought of the Mustangs winning out and representing the Group of 5 in the hometown Cotton Bowl. This is a great story.
It's also one that has already nearly ended on a few occasions. SMU won a toss-up at Arkansas State in Week 1 (postgame-win expectancy: 65%), then did so again in a Week 4 upset at TCU (55%). Ranked for the first time in decades, the Mustangs underwhelmed at home against Tulsa and had to rally for an unlikely overtime win (21%). And in a rainstorm against a tanking Houston last Thursday night, they were outgained by 125 yards, with a -4% success rate margin, and averaged fewer points per scoring opportunity. It took a couple of fumble recoveries and a late-game stop to survive 34-31.
Through eight games, SMU's second-order-win total is 5.7. This week's opponent, Memphis? 7.5. The Mustangs slid to 40th in SP+ following Week 9; that ranks just sixth in the AAC. Memphis is 19th, second in the AAC.
There's a reason, in other words, the early lines have had the Tigers as a three- to four-point favorite despite SMU's ranking 15th in the AP poll. SP+ says the margin could end up in the double digits.
Other games featuring these 10 teams: NC State at Wake Forest (-7.5), New Mexico at Nevada (-4), and Georgia Southern at Appalachian State (-16.5).
One final note: Just as your fortunes can change, so can your quality. Minnesota was crazy-lucky to finish September unbeaten -- the Gophers beat South Dakota State, unlucky Fresno State, and Georgia Southern by a combined 13 points and ranked just 50th in SP+ four weeks into the season. Since then, however, they've been legit.
Over the past four weeks, they've beaten Illinois, Nebraska, Rutgers and Maryland by an average score of 42-10, and they've surged to 12th in SP+. They might be lucky to be unbeaten, but they are certainly playing like an unbeaten team now.