Writers of Pro Football Prospectus 2008

Most Recent FO Features


» Film Room: Josh Gordon

Does Josh Gordon deserve a chance at redemption? Can he still be a top-flight receiver after nearly three years in exile? Is he the most dangerous weapon on the Browns roster? Charles McDonald explains why the answers are yes, yes, and lord, yes.

24 Oct 2008

Varsity Numbers: The New Box Score

by Bill Connelly

Let's say you spent all last Saturday on a plane, on a boat without cell phone signal, or asleep after a particularly rough Friday night. You check the scores Sunday morning and see that Texas beat Missouri 56-31, while Penn State beat Michigan 46-17. Granted, Missouri's better than Michigan this year, but those still seem like pretty equivalent wins.

Or were they? Lucky for you, you have the new Varsity Numbers box score in front of you.

First, we will take on Texas-Mizzou. If you just saw the 56-31 final score, you'd probably think that Missouri's offense did what it normally does (score points and eat up yardage), but their defense couldn't stop what is becoming an offensive juggernaut in Austin, and eventually Texas pulled away. That really is not how the game unfolded.



Close % 29.1%
Field Position % 40.6% 59.4%
Leverage % 65.0% 81.1%
Plays 60 74
EqPts 25.99 44.53
Success Rate 53.3% 62.2%
Points Per Play (PPP) 0.43 0.60
S&P (Success + PPP) 0.967 1.223
Plays 12 27
EqPts 0.71 15.67
Success Rate 16.7% 66.7%
PPP 0.06 0.58
S&P 0.226 1.247
EqPts 5.31 19.61
Success Rate 47.1% 56.4%
PPP 0.31 0.50
S&P 0.783 1.067
Line Yards/carry
1.73 3.12
PASSING 20.68 24.93
Success Rate 55.8% 68.6%
PPP 0.48 0.71
S&P 1.039 1.398
Success Rate 59.0% 63.3%
PPP 0.46 0.47
S&P 1.051 1.107
Success Rate 42.9% 57.1%
PPP 0.38 1.15
S&P 0.810 1.722
S&P by Quarter
Q1 S&P 0.226 1.242
Q2 S&P 0.814 1.563
Q3 S&P 1.168 0.893
Q4 S&P 1.742 1.159
S&P by Down
1st Down S&P 1.150 1.190
2nd Down S&P 0.768 1.103
3rd Down S&P 0.891 1.577
Number 1 0
Points Lost 0.90 0.00
Points Given 4.25 0.00
Total T/O Pts 5.15 0.00
Turnover Pts Margin -5.15 +5.15

It's not the most succinct thing in the world (it's more like a one-page briefing than a box score), but it tells you pretty much all you need to know about the game that took place. Now let's take a look at what the Varsity Numbers briefing tells you.

Close % -- If a team led by 7 points all game but then scored two touchdowns late to win by 21, then people end up saying that "the game was much closer than the score indicates." Meanwhile, if a team led by 35 points when they took their starters out, and their scrubs gave up two late touchdowns to cut the final scoring margin to 21, they end up winning by the same scoring margin as above. "Close %" complements the final score by signifying the percentage of the game's plays that were run under "close game" circumstances, i.e. with a scoring margin of 16 points (two possessions) or less.

Field Position % -- A crude attempt at measuring the field position battle, "Field Position %" simply compares the number of plays one team ran in opposing field position to the number of plays the other team ran in opposing field position. The two numbers here add up to 100 percent.

Leverage % -- This is derived from the idea that leveraging a team into as many Passing Down situations as possible is a subtle, effective way of winning games. Leverage % = Total plays run on Non-Passing Downs / Total plays. The higher number, the better. Anything over about 75 percent means the offense stayed in comfortable situations most of the game.

EqPts -- One of the first concepts discussed on Varsity Numbers, EqPts are derived from determining a point value associated with each yard on the field.

Success Rate -- Also discussed in that first column, this is the Football Outsiders concept of Success Rates, adjusted for college data.

PPP -- EqPts Per Play.

S&P -- Success Rate + PPP. It is to football what OPS is to baseball, an attempt to measure both efficiency (success rate) and explosiveness (PPP). To date, PPP has a higher statistical correlation to wins and losses than do either Success Rates or S&P, but all three are useful measures.

Line Yards/carry -- This is a direct duplicate of the Line Yards concept long employed by Football Outsiders.

Points Lost and Points Given -- Discussed in the "Six Missing Numbers" piece, these ideas come from the assumption that not every turnover is created equal. "Points Lost" refers to the field position the offense lost when they turned the ball over, while "Points Given" refers to the resulting field position the defense secured with the takeaway.

With these concepts in mind, let's now take a look at the aforementioned Penn State-Michigan game.


Penn State

Close % 80.6%
Field Position % 35.4% 64.6%
Leverage % 72.1% 71.8%
Plays 68 71
EqPts 17.96 35.00
Success Rate 38.2% 47.9%
Points Per Play (PPP) 0.26 0.49
S&P (Success + PPP) 0.646 0.972
Plays 55 57
EqPts 15.66 25.49
Success Rate 38.2% 49.1%
PPP 0.28 0.45
S&P 0.667 0.938
EqPts 14.22 18.81
Success Rate 46.5% 50.0%
PPP 0.33 0.50
S&P 0.796 0.995
Line Yards/carry
3.01 3.11
PASSING 3.73 16.19
Success Rate 24.0% 45.5%
PPP 0.15 0.49
S&P 0.389 0.945
Success Rate 42.9% 47.1%
PPP 0.24 0.44
S&P 0.673 0.906
Success Rate 26.3% 50.0%
PPP 0.32 0.64
S&P 0.578 1.139
S&P by Quarter
Q1 S&P 1.007 1.261
Q2 S&P 0.446 0.796
Q3 S&P 0.326 0.956
Q4 S&P 0.518 1.081
S&P by Down
1st Down S&P 0.471 0.708
2nd Down S&P 0.868 1.159
3rd Down S&P 0.590 1.241
Number 1 1
Points Lost 1.27 1.94
Points Given 3.98 1.92
Total T/O Pts 5.25 3.86
Turnover Pts Margin -1.39 +1.39

What does this tell us about these two games?

  • What happens if you're Carl Edwards, and you have a mishap on the track in Lap 1 of a race? Your car is able to keep going, but you end up laps behind. You try to drive well the rest of the race, but it just doesn't matter because your day ended in the first minute, and the next three hours are just a formality. The MU-UT game is why the "close game" designation was created and why it is tracked throughout the season. While the full-game statistics are far from worthless, more emphasis is placed on the "close game" statistics. Overall, MU-UT was an offensive shootout, with Texas just having more success in the end. The "close game" stats show that this was never a game, not even for one second. Anybody who watched part of the game (or in my case, every painful second of it) can attest to that.
  • Michigan-PSU was a competitive game almost three times longer than Mizzou-UT.
  • Both Michigan and Penn State were mediocre on first downs, but Michigan's extreme troubles put them in awkward positions all game long, and they suffered with progressively worse per-quarter S&Ps because of it.
  • Defensive breakdowns on Passing Downs allowed Penn State and Texas both to build a strong cushion on the scoreboard. Texas was near-perfect on Passing Downs in the first half, while the Nittany Lions repeatedly converted on third-and-7 or worse in the second half.

Once the processing of play-by-plays is done in a quicker manner, and "+" numbers can be generated on an immediate basis, they can be added to the Varsity Numbers box score as well.

A Quick National Look

Just for some further background on what constitutes good, bad, and average Success Rate, PPP and S&P numbers, here is a quick national look.

National College Statistics (BCS games only, to date)
Scenario Success Rate PPP S&P
All plays 41.5% 0.34 0.750
Rushing only 42.7% 0.30 0.729
Passing only 40.3% 0.37 0.773
Q1 42.7% 0.33 0.756
Q2 41.8% 0.34 0.762
Q3 41.3% 0.34 0.754
Q4 40.4% 0.33 0.730
Overtime 38.8% 0.41 0.800
1st Down 43.0% 0.33 0.764
2nd Down 41.0% 0.33 0.745
3rd Down 38.4% 0.34 0.721
Non-Passing Downs 47.2% 0.34 0.810
Passing Downs 29.4% 0.33 0.621
In offense's field position 40.7% 0.23 0.638
In defense's field position 42.6% 0.47 0.892

Two quick notes:

1) The 5 percent difference in success rates between first down and third down means there might need to be a re-calibration there. The goal is to have the same success rate on every down. I will look at this again when the full data has been entered.

2) The difference in PPP between plays on each side of the 50 shows how the EqPts graph has a higher slope the closer you get to the goal line.


Finally, a couple of quick responses to comments from last week's column.

[P]layers are bigger and faster at all levels now, and there are probably more injuries. With scholarship limits, each team has fewer backups. With less "easy" games, those backups have fewer opportunities to play in blow-outs to get experience. So when they have to come in and play for an injured starter, we should expect a greater drop-off in play than in previous years.

Makes sense to me. I do think the last couple of years have seen a shift away from parity, but I still cannot tell if this will be a more long-term shift or not.

Penn State was "down" for 4 years -- 2000, 2001, 2003, and 2004. Prior to that, they finished ranked No. 11, No. 17, No. 16, No. 7 -- you get the picture, basically they were a perennial Top 25 team. In 2002 they finished ranked No. 16, 2005 they finished ranked No. 3, 2006 and 2007 they finished ranked No. 24 and No. 27(ish).

Right, but "perennial Top 25" is not the same thing as "perennial Top 5" or "perennial Top 10." There are quite a few perennial Top 25 teams who aren't ever threats to win the national title. My initial point was that there aren't as many "elite of the elite" programs (i.e. Miami from '83 to '92, USC from '02 to '05) right now, and simply being in the Top 25 every year doesn't qualify (though as a Mizzou fan, I sure don't mind the thought of being ranked every year).

I took every D-IA season since the modern era began in 1946 and calculated each team's pythagorean winning % using pythagenpat exponents. Then I ran the standard deviation of PW% for each season as a very broad measure of parity ... It's clear that the gap between the best and worst teams in the country is smaller than it's ever been before. People thought last year was an aberration, but in fact it's the new reality.

I do think last year was still a bit of an outlier. (I mean, twice as many "unranked team beats Top 5 team" upsets than any other year? That's pretty ridiculous, especially considering there have only been two so far this season.) Still, this is the kind of analysis I really enjoy, and I hope to prompt more of this in the future.

One thing that affected parity and competitive balance was the gradual desegragation of college football. As teams began recruiting African-american players, power shifted around. The dominant Oklahoma teams of the '50s had their first black player in '56, Alabama didn't until the early '70s. It is not surprising that teams showed more year-to-year consistency after this process completed, roughly in the mid-'70s.

One thing that becomes a certainty after looking at year-to-year results is that there is a never-ending cycle of years where an elite few teams have coalesced at the top (for instance, Miami and OU in the mid- to late '80s, with Nebraska and Florida State coming on strong), followed by a few years where the elite teams fall off, and there's a scrum to come out on top in the balance of power (for instance, the early '80s, which saw teams like Georgia, Clemson, Pittsburgh, Penn State, Alabama, and others all staking their claim from year to year). We appear to be in the middle of a scrum right now, though the youth and explosiveness that Texas displayed last Saturday night has me a little worried about their presence for the next couple of years.

Posted by: Bill Connelly on 24 Oct 2008

2 comments, Last at 25 Oct 2008, 3:58pm by Keegs


by Pat (filler) (not verified) :: Fri, 10/24/2008 - 11:20pm

Right, but "perennial Top 25" is not the same thing as "perennial Top 5" or "perennial Top 10." There are quite a few perennial Top 25 teams who aren't ever threats to win the national title.

Um. Who?

Taking the past 5 years, the only teams that were in the Top 25 every year were LSU, Georgia, USC, Ohio State, Oklahoma, Texas, Florida, and Auburn - all of whom either played in/won a national championship game in those years, or received #1 votes in the AP poll at the end of the year. So it's safe to say they're all threats.

A program that's always in the Top 25 - even for as short a stretch as 5 years - is an elite program. Obviously Penn State's not entirely back to being in the Top 25 every year, but over the past 4 years it definitely looks like it's close. Assuming they don't fall off a cliff next year (or this year), they'll only be very marginally off that mark. Note that it's not like Penn State was a perennial Top 25 team "forever" before 2000, either - it was only back to 1993.

Note that the number of perennial Top 25 teams (over a 5 year span) is only marginally different (8 vs 9) from 1990-1995. I'm just not really sure that the idea that there are fewer "elite" programs nowadays really holds water. At least, in the AP poll's view. In both 1991-1995, and 2003-2007, you have roughly the same number of perennial Top 25 teams, and the same number of perennial Top 5 teams (1: FSU/USC).

I just don't really notice a strong difference between the frequency of teams in the Top 25 recently. I think the 85-scholarship limit and the general improvement of facilities at universities around the country is pushing towards more parity - not a decline in "traditional powers." Traditional powers pretty much always wax and wane. Of the 8 perennial Top 25 teams I had listed earlier, a grand total of one was a perennial Top 25 team from 1991-1995: Florida.

Of the three teams you listed as being traditional powers who declined (Nebraska, Penn State, Florida State), I'm not sure any of them count. As an example, Florida State's on a bit of a downturn, but it's not exactly large - they're still borderline Top 25 most years. This isn't exactly uncommon - it's not much different than Miami in the latter 1990s. Nebraska is probably the team that's closest to counting, and they still were ranked in 2003 and 2005.

It just seems like the "traditional power X was on a downturn, but is surging this year" is way overplayed. Teams always have downturns, and they always rebound. In a few years we're going to hear about Miami's resurgence, and Michigan's resurgence. It just doesn't seem that different.

by Keegs (not verified) :: Sat, 10/25/2008 - 3:58pm

"Close %" complements the final score by signifying the percentage of the game's plays that were run under "close game" circumstances, i.e. with a scoring margin of 16 points (two possessions) or less.

I would suggest tweaking this. ALL first half plays should be "close," and any 2nd plays within 16, imo. Or maybe something like first half plays within 32 and 2nd half plays within 16.