Reset: Revisiting the Five Factors

Back in January and February, before I got myself waylaid by other projects, I introduced the idea of the Five Factors for college football.

[O]ver time, I've come to realize that the sport comes down to five basic things, four of which you can mostly control. You make more big plays than your opponent, you stay on schedule, you tilt the field, you finish drives, and you fall on the ball. Explosiveness, efficiency, field position, finishing drives, and turnovers are the five factors to winning football games.

• If you win the explosiveness battle (using PPP), you win 86 percent of the time.
• If you win the efficency battle (using Success Rate), you win 83 percent of the time.
• If you win the drive-finishing battle (using points per trip inside the 40), you win 75 percent of the time.
• If you win the field position battle (using average starting field position), you win 72 percent of the time.
• If you win the turnover battle (using turnover margin), you win 73 percent of the time.

This is from 2013 college football game data. It's very, very similar from year to year.

These are good odds. And they speak to the fundamentals of football itself. You want to be efficient when you've got the ball, because if you fall behind schedule and into passing downs, you're far less likely to make a good play. You want to eat up chunks of yardage with big plays, because big plays mean both points and fewer opportunities to make mistakes. When you get the opportunity to score, you want to score. And when you give the ball back to your opponent, you want to give them to have to go as far as possible.

And you want that damned, pointy ball to bounce in a favorable way. Again, you control four of the five.

This idea seemed to resonate. It's easy to understand and intuitive, and if taken no further, it could be used to craft a more useful, descriptive, and simpler box score (like the one I attempted to craft in Chapter 7 of my book).

Among other things, you'll see some work I've done with this idea in the upcoming Athlon college football preview magazines -- regional and national -- when they're released in a couple of months. And I'm including Five Factors boxes with both regular and opponent-adjusted stats in my 2014 preview series. This idea is both logical and informative.

Really, though, there are three layers to this concept, and now that I have some time, I'm finally revisiting them.

1. Identify the five factors.

2. Strip these factors down into unrelated components.

3. Identify how these components -- some based on luck and randomness -- can most effectively be used from an evaluative or predictive standpoint.

Dean Oliver's Four Factors for basketball, to which I made liberal reference in the initial Five Factors post, are unrelated to each other. From a statistical standpoint, your ability to shoot efficiently has nothing to do with your ability to rebound, and your turnover rates aren't tied to your ability to draw or avoid fouls. These football factors as originally structured, however, are tangled into knots. Success Rate, the primary efficiency measure, has a strong correlation with Points Per Play, my original explosiveness measure. Field position is dictated in large part by the quality of the offense and defense (i.e. efficiency and explosiveness). While finishing drives is its own skill to an extent, it still has a strong correlation to pure quality as well. And even turnovers are tied in part to quality among the randomness.

That's why the second layer listed above, stripping these factors apart, is so important. And that's where we currently stand. My current goal is to work through each of these five factors and figure out the specific components that go into each one. The first two were pretty easy, and then it got more complicated.

Efficiency

This one required no work. The efficiency measure I will continue to use is Success Rate. This is a pretty simple concept: Every play is deemed successful or unsuccessful based on the following definition: 50% of needed yards on first down, 70% of needed yards on second down, or 100% of needed yards on third or fourth down.

I've gotten some questions about this definition; coaches tend to use a 50% bar for both first and second downs. But I felt that for this measure to be as effective as possible, there needs to be a single number. What I mean by that is, I want the baseline to be about the same for first, second, and third down. Generally speaking, success rates on third down hover between 40 and 45 percent. That means first- and second-down success rates need to do the same. Set the second-down bar at 50 percent, and everyone's second-down success rate trends far over 45 percent. Set it at 70 percent, and it lands within that range.

How to derive: Success Rate 100%

Explosiveness

I worked on isolating explosiveness from efficiency in this post. Instead of looking at teams' Equivalent Points Per Play (PPP) averages, I tinkered with what might happen if I looked at teams' PPP average for only successful plays. The results were pretty exciting.

Using full-season game data from 2012 and 2013 (with FCS games removed), I crafted a new version of S&P using Success Rate and this Isolated PPP idea (PPP on successful plays only). The most effective weights: 86% Success Rate, 14% IsoPPP. With that weighting, I was able to almost exactly recreate the strong correlations between S&P and both points scored and percentage of points scored. [...]

This new version of S&P passed another test as well. One of the better ways of determining the reliability and consistency of a measure is to compare its performance in one half of the season to its performance in the other. If the first-half and second-half numbers are drastically different, you might not be measuring a very consistent variable. Not a problem. [...]

This 'new' version of S&P is just as consistent and closely correlated with quality as the old one; plus, I was able to strip apart the concepts of efficiency and explosiveness to a strong degree.

Correlation between Success Rate and PPP: 0.666
Correlation between Success Rate and IsoPPP: 0.145

I tinkered a bit with opponent adjustments, and this IsoPPP idea remains sound so far. So for now...

How to derive: IsoPPP 100%

Field Position

Now it begins to get messy. The last three factors involved are results of all sorts of influences, including efficiency and explosiveness. In the third post in the series, I attempted to break field position down into digestible pieces.

Field Position might have more influences than any of the Five Factors. To win the field position battle is to move, kick, punt, and return the ball better than your opponent. Or at least three of the four. And you probably want to win the turnover battle, too. Field Position is a mix of a ton of other factors. How much of each? [...]

[U]sing Offensive and Defensive Success Rates (37%), Kick Margin (22%), Turnover Margin (21%), and Punt Margin (20%), I was able to create a pretty damn strong Projected FP Margin.

Anything dealing with special teams risks continuity issues. Special teams as a whole is such a small-sample thing that correlations from the first to the second half of the season are minimal and from one season to another are almost non-existent. It absolutely makes a difference, but it's not great from a prediction standpoint. But we'll worry about that when we get to the third layer.

How to derive: Success Rates (offense and defense) 37%, kickoffs (offense and defense) 22%, turnovers 21%, punts (offense and defense) 20%.

So that leaves the final two aspects -- finishing drives and turnovers. I'll be attacking the former tomorrow and hopefully the latter by the end of the week.