If you’ve visited Our Daily Bears over the last three years during any football season, chances are that you’ve noticed that the ODB staff loves advanced statistics. Look at any post where stats are discussed and you’ll see advanced stats there 99% of the time. Mark and Prashanth get all of the credit for this bent, but I am absolutely a convert to this line of thinking and enjoy diving deeper into those stats.
In last week’s Big 12 Conference advanced stats post, I put the poll out there on whether folks would like a reintroduction to the advanced stats that we enjoy so much at ODB, and you overwhelmingly voted positively. Some suggested possibly linking to the original post on this topic, but I wasn’t able to find one from ODB. The information detailing each of the stats is also so spread out that it’s easier to just go ahead and compile it here with references. I will try to be as comprehensive as possible. If you have questions about any of this, please ask them in the comments. This stuff can be overwhelming at times, but hopefully we can help with that.
First things first: pretty much all of the advanced statistics that we use in our advanced stats posts and previews come from Football Outsiders (if you want a primer on what FO is all about, this is a good place to start). We look at three statistics and their constituent components: FEI, S&P+, and F/+. The F/+ is a combination of the two and serves as the "official" ranking of FO. FEI is created by Brian Fremeau, and S&P+ is created by SB Nation’s own Bill Connelly. These stats examine different sets of data and do different things, which we’ll discuss in more detail later. Both attempt to filter out meaningless drives (such as clock killers at the end of a half) and garbage time in an attempt to only look at drives considered "meaningful.".
The "margin of victory" debate has gone on for a very long time, and was heightened in the early days of the BCS rankings. Originally taken into consideration by the BCS polls, it was widely criticized because it supposedly encouraged teams to run up the score. It was eventually removed entirely, instead of attempting to find some middle ground where margin of victory could be considered without incentivizing teams running up the score. The FO stats attempt to do just that, by factoring out what they consider to be "garbage time" plays and drives. Here’s how garbage time is classified for S&P+*:
- 1st Quarter: 28 points
- 2nd Quarter: 24 points
- 3rd Quarter: 21 points
- 4th Quarter: 16 points
Source: FO’s S&P+ Page. The points listed above are the threshold for whether a game is in "garbage time." Once a team hits the margin listed in a given quarter, the FO stats do not consider those plays or drives. If a game drops below that threshold, the drives are counted again. EXAMPLE: Baylor vs. TCU 2014. Early in the 4th Quarter, TCU picks off a pass by Petty, returning it for a score, giving TCU a 21-point lead over Baylor. The FO stats do not take Baylor’s subsequent drive (5 run plays for a touchdown) into account because the game is in "garbage time" at that point. Once Baylor scores and the lead is cut to 14 points, the game is no longer in garbage time and stats are considered for the remainder of the game. It’s clearly not perfect, but it’s the best effort I’ve seen to filter out meaningless stat-padding.
*After posting this, I received a tweet from Brian Fremeau who informed me that FEI and S&P+ calculate Garbage Time differently. I did not realize this. Brian is going to email me with how he calculates garbage time, and once I have that, I will put it here.
[UPDATE] I received the following from Brian Fremeau regarding how he calculates Garbage Time. It's fascinating.
I calculate garbage time retroactively based on the scoring margin of the game and number of game possessions remaining. Specifically, garbage time begins when the score deficit is greater than 8 times the number of remaining possessions for the losing team plus one.
For example, there were 25 offensive possessions in the Baylor-Oklahoma game, 24 if we drop the end-of-first-half Baylor kneeldown. Baylor led 38-14 in the second half, a 24 point lead. Oklahoma would need three TDs and 2-point conversions to tie, so they would need the ball at least three more times in the game to accomplish such a comeback. The Sooners punted on their actual third-to-last possession. Baylor kicked a field goal on the ensuing drive to bring the score to 41-14, a 27 point lead. Oklahoma would need four more drives at this point in the game but would only possess the ball two more times. The Baylor field goal was the final non-garbage possession of the game.
I identify the non-garbage final scores in the game splits data on my site:http://www.bcftoys.com/2014-game-splits#baylor
In my method the game passes a garbage time threshold and never returns. Bill's method can shift back and forth from non-garbage to garbage, though that is rare.
We good so far? Let’s move to the advanced stats.
F/+: The Combined Stat
As I mentioned before, F/+ is the official ranking from Football Outsiders, born out of the collective efforts of Brian Fremeau’s FEI and Bill Connelly’s S&P+ (FEI + S&P+ = F/+). It’s calculated by first compiling Offensive, Defensive, and Special Teams F/+ Metrics and then combining those into the overall F/+ rating. I’m not completely sure what the formula is for this combination, so I’m going to do some hand waving and move on. /waves hands frantically
F/+ Special Teams
In the weekly Big 12 Conference Advanced Stats post, you’ll see included in the Primary Rankings table the stat F/+ Special Teams. This stat originated as part of FEI, but was split out in 2011 and became part of the overall F/+ combination. If this post is still accurate, it comprises 14% of the overall formula. For that reason, we include it the primary table in the Big 12 posts. In the advanced stats previews, special teams get their own table, which include the F/+ Special Teams rankings as well as the FEI Special Teams ratings. We’ll talk about those more in a moment.
Field Position Advantage (FPA)
Field Position Advantage is defined here as "the share of the value of total starting field position earned by each team against its opponents." While the stat has multiple components that factor into the total, we don’t spend too much time examining those stats. If you’re interested in the components, check out the linked article for more information.
Now it’s time for the nuts and bolts.
FEI: The Fremeau Efficiency Index
The name of this advanced statistic is a riddle wrapped in an enigma, so let me unpack it for you: this stat was created by Brian Fremeau as a measure of a team's efficiency. This from the Football Outsiders FEI combined page:
The Fremeau Efficiency Index (FEI) considers each of the nearly 20,000 possessions every season in major college football. All drives are filtered to eliminate first-half clock-kills and end-of-game garbage drives and scores. A scoring rate analysis of the remaining possessions then determines the baseline possession efficiency expectations against which each team is measured. A team is rewarded for playing well against good teams, win or lose, and is punished more severely for playing poorly against bad teams than it is rewarded for playing well against bad teams.
Confession time: I took a statistics course in college. I didn't do very well in it. When we start talking about baseline efficiency expectations and r² and the like, my eyes start to cross and I hear a loud buzzing noise. The key information in this explanation, for our purposes, is that FEI is a drive-based statistic that is less concerned with the type or results of individual plays. Garbage time and clock-killing drives are eliminated from the equation, so we're left with "meaningful" drive data. If you were to click on the link above and take a look, you'd see multiple components that we don't even look at in any of our statistical comparisons. You'll also see "OFEI" and "DFEI," the overall efficiency metric for each side of the ball. It's these metrics in particular (and the component statistics of each) that we focus on in our previews.
If you head to either the Offensive FEI or Defensive FEI page, you'll notice that the components of each statistic are the same. There are some metrics on each of the pages that we're not really concerned with for our purposes, so I'm going to leave those alone and focus on the component stats we look at in our statistical previews. I should also note that where available, if you mouseover any table where you see one of the abbreviations below, you'll see a tooltip that gives a brief explanation of the stat. For this purpose, I'm drawing the definitions from the Offensive FEI page. For defense, these stats measure how the opposing team's offense performs.
FD: First Down Rate
First Down rate, the percentage of offensive drives that result in at least one first down or touchdown.
Simple and straightforward. We're looking at drives that aren't three-and-outs or cut short by turnovers. The Bears currently have a First Down rate of .775, meaning that on 77.5% of their drives, the Bears either score a touchdown or get at least one first down. That's good for 7th in the nation for FD. So, when you see the table with FD, we're showing you their national rank in that category, not the raw stat. Helpful, but alone it doesn't tell us very much.
AY: Available Yards
Available Yards, yards earned by the offense divided by the total number of yards available based on starting field position.
Example: Baylor gets the ball on the 25 yard line. They march down the field to the opponent's 35, where on 4th-and-4, Petty throws an incomplete pass (Goodley is interfered with but it goes uncalled, naturally). The Bears gained 40 yards on the drive out of a possible 75. For the drive, their AY rate is .533 (repeating, of course). If they were to have scored a touchdown on the drive, their AY rate would have been 1.0. At present time, the Bears AY is .533, good for 5th best in the nation.
Still with me? Let's keep rolling.
Ex: Explosive Drives
Ex: Explosive Drives, the percentage of each offense's drives that average at least 10 yards per play.
Now we're cooking. This is a statistic that should be near and dear to Baylor fans' hearts. We're talking about big play drives. Petty hits Cannon for a 72-yard touchdown pass. 3 plays, 80 yards. Explosive drive! The Bears march 99 yards down the field in 9 plays to score a touchdown. Explosive drive! Any drive averaging 10 yards per play is explosive. It may come as somewhat of a shock to learn that in this season, Baylor currently has an Ex rating of .196, good for only 15th in the country.
Me: Methodical Drives
Methodical Drives, the percentage of each offense's drives that run 10 or more plays.
This advanced stat may be the easiest advanced stat to understand in the history of advanced stats. Did your team run a drive that's longer than 10 plays? Methodical! This is the "three yards and a cloud of dust" metric. It's also the one that keeps Iowa State's offense from absolutely nosediving in the FEI rankings, because they're #2. That's right, Iowa State runs more methodical drives in the country than anyone, save one team (San Jose State).
The interesting part about both the Explosive and Methodical drive components of FEI is that they don't consider success as a factor, at least as far as I can tell. A two play, 35 yard drive that ends in a recovered fumble is considered explosive. An eleven play, 35 yard drive that travels from a team's own 10 to the 45-yard line but ends in a punt is a methodical drive. That's why you want to see a high FD rate and also a high Value Drive percentage, which is the final FEI component we look at.
Va: Value Drives
Value Drives, the percentage of each offense's drives beginning on its own side of the field that reach at least the opponent's 30-yard line.
(Thank goodness they decided to go with Va, the alternative would have been...unfortunate) A Value Drive begins on an offense's own side of the 50 and reaches the opponent's 30-yard line. What we're looking at here is whether a team is putting itself in a position to score, be it a field goal or a touchdown. I wasn't clear about whether that counted big plays that went for touchdowns (e.g. a 78-yard bomb to Cannon), so I reached out to Brian Fremeau and he confirmed that he does, in fact, count those as Value Drives. A't any rate, Iowa State is 61st in Value Drives, with a 38.6% rating. That gives you an idea why their offense isn't better given their high Me percentage: they stall out and lack the finishing power to get down into scoring position.
There are your individual components to FEI to which we pay attention. Combine those numbers and you'll get the raw Offensive Efficiency metric that is listed on the page I linked above. Take that, adjust it for Strength of Schedule* (also listed on the page above), and you've got your Offensive/Defensive FEI number that we discuss in our previews.
*The description I listed at the beginning of this section mentioned that FEI rewards teams for playing well against good teams and punishes teams more severely for playing poorly against poor teams, regardless of wins or losses. I'm assuming that this is factored in by strength of schedule, but how exactly I'm not sure.
Eyes crossing yet? Shall we move on?
S&P+: The Brainchild of Bill C.
While FEI looks at drive data exclusively, S&P+ looks at individual plays to determine whether that play was "successful," and how explosive was that play. Created by Bill Connelly, he originally envisioned the stat to be something similar to OPS in baseball, where Success Rate and Points Per Play (first PPP and now IsoPPP, more on that in a moment) are added together to create S&P, then opponent adjustments are applied to give you the "+". Ever the tinkerer, the statistic evolved over the years, incorporating drive efficiency into the calculation as well as refinements to the Points Per Play metric which we'll discuss in a moment. As with FEI, S&P+ measures data for offenses and defenses, then combines them for a unified statistic.
Though we only look at the components that go into S&P and a couple of other metrics, Bill C. calculates a lot of other stats. If you look at either the Offensive or Defensive S&P+ pages on Football Outsiders, You'll see multiple stats that we don't discuss. Take a look at his work on Football Study Hall and you'll find even more stats, details and breakdowns. I highly recommend his advanced box scores. Keep his Advanced Stats Glossary handy, though. It's been a huge help for me in figuring out what all his stuff means.
What makes a play successful? Well, this is how Bill C. defines success on a given down:
- 1st Down: 50% of the necessary yardage
- 2nd Down: 70% of the necessary yardage
- 3rd & 4th Downs: 100% of the necessary yardage
So, if you have a 1st & 10 and get 7 yards, that play was successful. If you have 2nd & 10 and gain 7 yards, that's success. You get the picture. So basically it's a binary: unsuccessful 0, successful 1. Take the total number of successfuls, divide it by the total number of plays, and presto! You have your success rate. Easy, right? Just wait.
This is a little bit more complicated, so we'll start with Equivalent Points Per Play (PPP) which was the other half of S&P in seasons previous. Here's how it works: Take every yard line and assign a point value to it based on the number of points that an offense could expect to score from that yard line. Then, for a given play, subtract the ending point value (based on whatever yard line on which the play ends) from the starting point value and presto! You have the PPP value for that play. Then add individual plays' PPP values together to give you your final PPP for a game. In years past, you take that number, add it to the Success rate, and you have S&P.
For us mere mortals, adding Success Rate and PPP and then adjusting for opponents may have been enough, but not for Bill Connelly. Oh no. Bill C. decided that since Success Rate measures efficiency, he wanted to try and strip out explosiveness out of that. In other words, if you're being successful, just how successful are you? Enter IsoPPP, wherein he isolates the PPP for successful plays only. In other words, it's a measure of how explosive you are with your successful plays.
S&P+ now consists of a weighted combination of Success Rate and IsoPPP: 80% Success Rate, 20% IsoPPP. Throw in opponent adjustments and drive efficiency, and you get S&P+.
Whew, we're done, right? NOPE. Bill C. is a wizard and can't let a simple opponent-adjusted metric go that easily. Silly mortal.
These two are simple: the formula above, just for Rushing or for Passing plays only. Huzzah!
Passing Downs S&P+
Passing downs is always listed second on our charts, so why talk about it first? You'll see. Passing downs are defined as:
- 2nd down & 8 or more yards to go
- 3rd/4th down & 5 or more yards to go
What we're looking at is how successful (efficient and explosive) you are on downs that are traditionally considered "passing downs." How well do you perform on downs where you should be taking to the air? That's what this metric is looking at. So, Passing Down S&P+ looks at the opponent-adjusted efficiency of your offense (or defense) only on those specified downs.
Anything not considered a Passing Down is a Standard Down for the purposes of S&P+ is a Standard Down. So, we have:
- 1st & anything
- 2nd & 7 or less
- 3rd/4th & 4 or less
Those are your Standard Downs. Are you successful on Standard Downs, and if you are successful, just how successful are you? We're finding out, plus throwing some opponent adjustments and drive efficiency into the mix just for clarity's sake.
If you're at all interested in any of this, click the links that I put in the post or the Must Read sidebars in the article. Football Study Hall has a wealth of information about advanced stats; Bill C., Ian Boyd, and all the rest do a stellar job over there. There are more stats than just those to look at, including Offensive Line and Defensive Line-specific stats, havoc rates for defenses, and further drill downs on Standard/Passing Down run percentages, as well as Pace Adjustments. Check it out, it's some fascinating stuff.
That's it, I'm Done.
There are special teams metrics that we look at too, but we don't spend much time on them, and as this article is already over 3,000 words, I think I'll save those for a later date.