Five Factors

[Bill Connelly]

Another post! I wanted to share something I posted today at Study Hall: College football's five factors.

http://www.footballstudyhall.com/2014/1/24/5337968/college-football-five-factors

As you see, I'm boiling football down to explosiveness, efficiency, field position, finishing drives, and turnovers. In the post itself, I talk about how and why each aspect is important, but as referenced in comments, these factors really aren't traits so much as end results. Teams that win are explosive, efficient, do well in field position, etc. They are all quite tangled up and related to each other -- they're more like five outcomes instead of five factors at this stage -- and for this to be an idea like Four Factors are in basketball, work is needed to untangle them. I have ideas, but I'd like others, too.

If you've read my work before, you know my style is generally to throw out an idea as quickly as is feasible, then flesh things out from there after getting as many reactions as possible. I'm pretty correct in assuming that I don't ever have things completely right the first time. This is no exception. Among other things, I was hoping this would start a conversation about how to measure these things, and I think we've gotten somewhere in the comments section. I'd like to bring it up here, too.

The main question is this: If you buy into this concept (not saying you do or don't), how could we go about isolating each of these factors? Maybe field position basically comes down to net kicking and punting. Maybe we use an isolated explosiveness measure -- instead of "How many yards (or equivalent points) are you averaging per play," maybe it's more "How many yards/points do you average on efficient plays (as a way to separate from the efficiency discussion)?"

Anyway, wanted to throw this out there for responses.

[David Fobare]

Turnovers: Burke at Advanced NFL Stats has shown that TOs are a persistent (weakly) phenomenon at that level. I suspect the same is true at the college level. But it will take some work to tease it out. TOs are also correlated to overall team strength. Not such a big deal in the NFL where the band of team abilities is pretty tight. But in a league where Idaho might meet Alabama its important.

My own work has shown that you can make a good prediction at net TOs by dividing the closing line by 23. That is, if the line is Bama -46 over Idaho the expectation is for Bama to finish +2 in TOs.

23 is also the number that works best for the NFL.

--
You received this message because you are subscribed to the Google Groups "Football Study Hall: Data Sharing" group.
To unsubscribe from this group and stop receiving emails from it, send an email to study-hall-cfb-...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

[Ed Feng]

Bill,

Great stuff, just tweeted it out.

Now go write a book about it :)

Ed

--
The Power Rank.  Will your team win?

Website Twitter Facebook

[Bill Connelly]

It's somebody else's turn!

[Brian Fremeau]

One thought I had while reading about it is how much field position affects everything. As you pointed out in the article, some turnovers mean more than others because of field position. Success in finishing a drive relates to where the drive starts. Explosiveness relates to field position in the red zone.

Your plan to measure the relationships between factors is an important step.

[Brian Fremeau]

Is this guy part of the group?

http://www.rockmnation.com/2014/1/26/5347850/rethinking-the-5-factors-is-there-a-better-way-to-measure-efficiency

I think there is something really smart and insightful here with the "continuous efficiency" suggestion in particular.

[Bill Connelly]

Ha, he is not, but I can invite him.

--

[Bill Connelly]

By the way, I took a stab at a no-efficiency version of explosiveness, and it was quite a bit more successful than I expected...

http://www.footballstudyhall.com/2014/1/27/5349762/five-factors-college-football-efficiency-explosiveness-isoppp

[fbsdrivestats]

I accidentally replied directly to Bill's post with a message. (Bill, don't worry about replying to my message. It was a just foolish mistake on my part.)  I'm not going to rewrite the post in its entirety, but I'll give a brief summary of what I found out. The table below lists how net statistics correlated with winning percentage:

 

Negatively correlated statistics are highlighted. I just used the absolute value of the negative correlations to make it easier to identify the statistics with the strongest correlations. The negatively correlated statistics deal with field position, turnovers, and punting. That makes sense because teams want to have a turnover rate and punting percentage lower than their opponents. They also want to start closer to the end zone. FYI, effective turnover percentage includes turnovers on downs and safeties on offensive plays.

The four statistics I would focus on are touchdown percentage, yards per play, effective turnover percentage (or something similar), and starting field position. Touchdown percentage does a better job of measuring the ability to finish drives than yards per drive, percentage of available yards gained, or ending field position. (Field goals per drive was quite easily the scoring statistic with the weakest correlation with winning.) Effective turnover percentage correlates more strongly with winning percentage and makes the other turnover statistics unnecessary for inclusion in the five factors of success.

 

Scoring percentage is football's equivalent to floor percentage, and points per drive is football's equivalent to points per possession. Neither statistic is used in basketball's four factors of success. I suspect Dean Oliver's rationale behind this is that using net points per possession is tantamount to using scoring margin as a determinant of winning. (My statistics don't always show that the winning team averaged more points per drive than its opponents. The primary culprit for this occurrence is that I don't account for non-offensive touchdowns.)

 

The most important formula from Basketball on Paper was the formula for possessions. Football's formula is quite a bit simpler to calculate possessions because it simply requires adding up the outcomes of each drive. Each drive has nine separate outcomes:

• Possessions can end with a score. This can obviously be a touchdown or field goal. (2)
• Possessions can end with a turnover. (4)
• Possessions can end with a missed field goal. (1)
• Possessions can end with a punt. (1)
• Possessions can end because time expired. (1)

 Some of you are so intelligent that this formula may seem obvious. I just felt that it's worth sharing for those unfamiliar with this concept.

[Bill Connelly]

To keep the thread going...I did a little work in unpacking Field Position today. Still missing something...not sure what yet...

http://www.footballstudyhall.com/2014/1/30/5361594/college-football-five-factors-field-position

--

[Matthew Smith]

Now that there's good cfbstats data I'm going to try and dig into this a decent amount over the offseason.  Not sure what I'll find/guess but we'll see.

One quick note that I think is worth considering is that turnovers in college are materially less random than in the NFL.  There are plenty of programs who will persist in materially above or below zero net turnovers per game.  I'd hypothesize this is mainly due to talent advantages, especially when it comes to cupcake games.  For instance, if Alabama hosts Eastern Kentucky, the expected turnover margin is probably something like +2 or +3 given talent advantage on a per play basis as well as the fact that EKU will be way behind and throwing it around while Bama pounds the ball late (though Bama's scrubs playing late may well zero out that impact on expected turnover margin in the second half).  Play two of those plus a game against a mediocre mid-major plus an SEC bottom feeder and they're already well into positive expected turnover margin territory.  There's still huge fluctuation on a game to game and year to year basis, but some programs are very rarely in negative turnover margin territory over the course of an entire season, and others are very rarely positive (including one wierd example, Ole Miss).

To unsubscribe from this group and stop receiving emails from it, send an email to study-hall-cfb-data+unsub...@googlegroups.com.

[David Fobare]

Again, you can get a good proxy for expected turnover margin by diving the betting line by 23. If Bama is favored by 46 over Eastern Kentucky the Tide's expected turnover margin is +2.

To unsubscribe from this group and stop receiving emails from it, send an email to study-hall-cfb-...@googlegroups.com.