2019 Season Preview: Advanced Stats for Dummies
Hey y’all, Evan here. The season is still a few weeks away, so we’re continuing our preseason push of 2019 Louisiana Tech #content with a look at something that’s nerdy and terrifying all at the same time: MATH.
In football, like most sports, nerdy dweebs have decided that metrics like “Who looks the best to me” and “who has the most HEART and COURAGE and GRIT” and “Who smells the nicest” and “who makes me feel the happiest” are not actually good ways to judge players or a team’s ability to win games. Despite my best efforts to forget about all the math I learned in Algebra for Social Science Majors my freshman year at Tech, Nathan keeps forcing me to pretend to understand what he’s talking about.
Since I don’t actually understand, I thought it might be a good idea for him to introduce some of the advanced stats metrics we’ll be referring to all year in our pre and post-game recaps. To do this, we’ll be looking at some of the metrics’ predictions for Tech’s 2019 season, too!
So without further ado, I’ll turn it over to gtpdd’s resident
nerd Stats Expert Nathan for more:
If you followed gtpdd last year, you’ve probably heard of Massey Ratings. In the past, we’ve used them to show the odds that Tech would win a particular game. And we probably included a chart that looks like this:
As far as computer polls go, however, Massey is pretty simple (at least when it comes to inputs). The system only factors in three things from each game: the score, the venue, and the date of the game.
So when Massey judges an offense, the rankings don’t care about how good a team’s rushing attack is or how bad a pass defense is. The ratings only care about how many points a team scored, how many points that team gave up, when, where, and against whom.
I just used whom in a sentence, take that 10th grade English teacher!
Because of the simple inputs, the data that Massey gives you isn’t very detailed. In fact, there are only nine categories:
- Rating: How good the team has been
- Power: How good the team could be (more or less)
- Offense: How good the team is at scoring points
- Defense: How good the team is at not giving up points
- Home Field Advantage (HFA): How big of a impact does the team’s home stadium have?
- Strength of Schedule P (SOS): How difficult was the team’s past schedule?
- Strength of Schedule F (SSF): How difficult the team’s schedule is, both past and future
- Estimated Wins
- Estimated Losses
Let’s see how Tech’s 2019 team looks in those nine categories:
Here’s where it’s worth mentioning that because Massey only uses inputs from past games, these preseason ratings are almost entirely based 2018’s production.
As the season goes on, we can use Massey Ratings to see a clearer picture of how good a team is. But for now, we can pretend that the poll puts Tech in 7th place in C-USA only because it just doesn’t have enough 2019 data yet.
Instead of blindly trusting Massey, let’s move to a system that ranks Tech 5th in the conference.
S&P+ is a completely different beast than Massey Ratings. The overall concept is based on Bill Connelly’s Five Factors of Football. They are:
- Explosiveness (IsoPPP)
- Efficiency (Success Rate)
- Finishing Drives (Points per Trip inside the opponent’s 40)
- Field Position (Average starting field position)
- Turnovers (Turnover margin)
At a very high level, these factors determine which team wins a football game. The more explosive team wins 86% of the time. The more efficient team wins 83% of the time. The team that finishes more drives wins 75% of the time. The team that wins the field position battle wins 72% of the time. The team that wins the turnover battle wins 73% of the time.
Two of those stats do need further explanation though: Success Rate and IsoPPP.
Success Rate measures the percentage of plays that are successful. That means picking up 50% of the yards to gain on first down, 70% on second down, or 100% on third/fourth down. So gaining 5 yards on 1st and 10 is deemed a successful play, but gaining 5 yards on 2nd and 10 is not.
IsoPPP, on the other hand, measures the average yardage gained on each successful play. (IsoPPP also takes into account that the closer a team gets to the opponent’s endzone, the more difficult it is to move the ball).
The goal is to figure out just how successful a team is on its successful plays. For example, a team that has multiple 30+ yard gains a game is much more explosive than a team that didn’t have any.
The S&P+ ratings are a combination of those five factors. Each factor receives a different weight to determine an S&P+ rating (or score). The teams are then ranked based on who has the higher score.
Like with Massey, the preseason S&P+ rankings take into account the past year’s success. But unlike Massey, S&P+ also takes into account the production lost to graduation/NFL/transferring and the recruiting classes coming in.
As the season goes on, we can see how Tech compares to the rest of college football when it comes to Success Rate and IsoPPP (and the other three factors I guess). And we can hope that it looks a bit better than last year (at least offensively):
ESPN also has an advanced metric. They’re the Worldwide Leader in Sports™, why wouldn’t they?
FPI stands for Football Power Index, and like Massey and S&P+, it uses past results as inputs to make predictions for the future. The FPI number itself represents how many more points a given team would be expected to score over an average opponent on a neutral field. Here’s Tech’s FPI profile for 2019:
So what does all that mean, exactly? Well, Tech would be expected to score 5.9 fewer points than an “average” opponent. But luckily, C-USA is anything but average. And I don’t mean that in a good way.
ESPN uses the FPI to run 10,000 simulations of the season to determine a projected win-loss record, a strength of schedule ranking (for remaining games), and the odds that the team will win the conference or win out.
And just like Massey Ratings, FPI predicts that Tech hover somewhere around a 7-5 record:
So FPI picks Tech to be the sixth best team in CUSA, and the third best in the West, behind USM and UNT.
FPI also provides a prediction for each game on the schedule. Then, while the game is being played, FPI recalculates the percentage chance each team has to win. That’s where you see charts like these:
Oh and uhh… you might see questions about these charts in the gtpdd Contest… wink wink.
Here’s Tech’s chance to win each game, according to FPI:
Advanced Stats and computer polls work by
magic witchcraft hamsters running in those little wheel thingys powering the mainframe having an amount of data and analyzing it to see how good a team is and how likely that team is to beat its future opponents.
Before the season starts, the only data we have is from the past year. So in the preseason, these polls are relatively useless (and I know that’s what you wanted to hear after reading all of this). But as the season goes on, we’ll be supplementing our Great Opinions by looking at stats like IsoPPP to tell a more complete picture of how good (or bad) Tech is.