clock menu more-arrow no yes mobile

Filed under:

Arkansas Fight’s Glossary of Advanced Football Stats

Meet all of our advanced stats

After years of scattering explanations for our advanced stats in articles that have to be linked back to, this Glossary is a long-time coming. Each statistic that we use for advanced stats is here, and some example box scores are posted at the bottom.

The EV System

All stats in this section are related to our use of EV, which is the center of all of our statistical analysis. All other stats support EV and help explain it.

  • Expected Value (EV), or Expected Points, is the “expected” number of points an offense will score on a drive given the down, distance, and yardline of a single play on that drive. For example, an offense facing 1st and 10 at its own 20 has an EV of 1.60, meaning that it can expect to score 1.6 points on the drive based on the situation of that play. A touchdown is worth 6.95 EV, because it gives six points plus the chance for an extra point, which is good about 95% of the time. If a team punts or turns the ball over, its EV falls to zero. If a team tries a field goal, its EV falls to a portion of the field goal’s value, based on how long the field goal attempt is (see “Placekicking EVA” below).
  • Expected Value Added (EVA) measures the change in EV brought about by each play. If the team facing 1st and 10 at its 20 (1.6 EV) completes a pass for a 7-yard gain, it sets up a 2nd and 3 from its own 27, a situation worth 1.73 EV. So that play was “worth” +0.13 EVA. It contributed that many points to the offenses’ overall effort. The sum of all EVA that an offense generates in a game, plus Starting Field Position EV, equals the actual number of points scored. The EV/EVA system allows us to find out where a team’s points actually came from. We can also find the exact value of plays: the Hogs’ 4th and 25 conversion against Ole Miss in 2015 was worth 4.24 points. Because EV falls to zero when a team punts, third-down EVA numbers are often worth huge swings. Getting stuffed on 3rd and 1 and having to punt is devastating.

There are different types of EV and EVA. Remember, these are calculated for both teams in a game, so we can judge a team’s defense based on how its opponent does:

  • Field Position EV is the expected number of points a team should score based on its starting field position for every drive. The team that wins this stat had better field position over the course of the game. There are two components to Field Position EV: Voluntary EV is the field position generated by the normal changes of possession in a football game. Voluntary EV assumes each drive starts 1st and 10 from your own 25 and assigns EV based on that. If your actual field position is better or worse, that’s counted as Forced EV, which is the difference between where the offense actually starts, and 1st and 10 at the 25. A positive Forced EV means the team forced turnovers or quick punts from the opponent and routinely started drives past its own 25. A negative Forced EV means the team was catching a lot of punts deep in its own territory. For this reason, Forced EV largely measures how quickly the defense generates stops.
  • Offense EVA consists of all EVA that an offense generates while it is on the field. This is normally broken down into Rushing EVA, Passing EVA, and Penalty EVA.
  • Special Teams EVA consists of all EVA generated by “offensive” special teams. Punt Return EVA is calculated as the difference between the EV for if the offense started where the punt was caught and the EV for where the offense actually started. Kick Return EVA is calculated as the EV difference between 1st and 10 at the 25 and 1st and 10 where the offense actually starts due to the return. Finally, Placekicking EVA includes the value assigned to the kicker for field goals and extra points. Extra points are worth 0.05 EVA (touchdowns are 6.95, the kicker gets the last 0.05 if he makes the extra point, and loses 0.95 EVA if he misses it), while the kicker-value of field goals depends on the length of the attempt. Field goals inside of 30 yards are worth only 0.3 EVA (the offense retains the other 2.7 EV), while attempts from 50+ yards grant the kicker 1.8 EVA and the offense retains just 1.2 EV.
  • Other EVA mostly includes defensive touchdowns, which are valued as the difference between 6.95 EV (a touchdown) and where the offense would start if the defender just fell down after collecting the turnover. Weird situations like returning a blocked extra point for two points are also included.

Here’s what a full EV/EVA game report looks like (this is from the 2019 Portland State game):

From this report, we can glean that Arkansas dominated field position but did not play well offensively.

There’s more we can do with EV:

  • Points Above Normal (PAN) is an opponent-adjusted version of EVA. Its calculation is rather complicated, but it involves comparing a team’s actual EVA performance against what its opponent usually gives up. For example, if Arkansas averages +0.05 EVA per rush on 40 rushes against Alabama, and Alabama’s run defense usually gives up +0.03 EVA per rush, then the Hogs had a Rushing PAN of +0.8 (+0.02 above average for 40 rushes), meaning that the run game generated 0.8 points more than Alabama normally gives up. PAN is typically not used early in the season because it takes a while to gather averages.
  • Adjusted Points is a predictive measure that tells us how a team would perform against the “average” FBS team, based on its all-season stats. It allows us to predict final scores and margins of victory or defeat. There are two numbers: Adjusted Points per Game and Adjusted Points Allowed per Game. Adjusted Points is calculated the same way that a team’s total points are calculated with EV/EVA: start with average Field Position EV (inclusive of returns) per game, add Offense PAN per game, add Penalty EVA per game, add Placekicking EVA per game, and add Defense EVA (defense scores) per game. While the result doesn’t necessarily back-fit over any single game, they can help you predict against the spread or get an idea of how well the team is playing so far.
  • Score predictions are based on deviation from the mean. If Team A’s adjusted points per game is 23.9, Team B’s adjusted points allowed per game is 25.5, and the FBS average adjusted points allowed per game is 25.2, then we predict Team A will score 24.2 points, which is their average (23.9) plus the the deviation of the defense (it allows 0.3 more adjusted points per game than the FBS average). We do the same for the other offense and defense to get a score prediction. For example, using 2018 data we can guess what would happen if the Hogs re-matched with Rutgers: Arkansas 30, Rutgers 21, if the game was in Fayetteville. Home teams get +3 points.

Efficiency & Explosiveness

The twin values of Efficiency and Explosiveness are important to monitor. They help explain the character of an offense (or defense!).

  • Success Rate is the basic efficiency measure. It is the percentage of a team’s offensive plays that are successful; basically, it is how good the offense is at staying on schedule. “Success” is defined as any play that generates positive EVA. It doesn’t matter how positive the play is: as long as it generates positive EVA, it is successful.
  • Marginal Efficiency is an opponent-adjusted version of success rate. It tells us how the offense performed compared to the quality of the opponent. For example, if Alabama normally gives up a 40% rushing success rate, and Arkansas has a rushing success rate of 44%, then Arkansas’ marginal efficiency is +10% (44% is 10% better than 40%).
  • Average Available Yards per Successful Play (isoAY) is our basic measure of explosiveness. It tells us the average percentage of all possible yards that were gained by successful plays. So if the ball is at the 50 and a successful play gains 20 yards, the isoAY of that play is 40%. This stat is only counted for successful plays were the number of available yards is at least 13.
  • Gini Coefficient of Success is another measure of explosiveness. We will typically use it for all-season data, not single games, since large amounts of data are necessary for it to be accurate. It measures the inequality (or variability) of a team’s successful offensive plays. Because only successful plays are used, it is independent of efficiency measures. A team that generates a lot of 40-, 60-, or 80-yard plays will have a high Gini coefficient. The concept comes from this article.

Run & Pass Stats

Because the EV/EVA system is so all-inclusive, we need more focused stats to break down very specific parts of an offense. These are those stats.

  • Power Success Rate is the success rate of run plays in short-yardage situations: third or fourth down and one or two yards to go. This measure is mostly about the offensive line’s ability to get a push.
  • Stuff Rate is the percentage of run plays stopped for 1 yard or less. Like success rate, this number is a yes/no question (“Was the play stopped for 1 yard or less?”) meaning that it measures efficiency more than explosiveness. High-efficiency running backs and good offensive lines have low stuff rates, while high-explosiveness running backs and bad offensive line have high ones.
  • Line Yards are the rushing yards credited to the offensive line. The offensive line is credited with 125% of lost yards, 100% of yards gained 0-3, and 50% of yards gained 4-6, so the maximum line yards on any play is 4.5 (a run of 6+ yards). This cap at six yards is in place because the offensive line is, in theory, only responsible for the first six yards. Yards after that skew the averages and are more about downfield blocking and the running back’s open-field skill. Good offensive lines have high line-yard averages. If this stat is significantly better than the next two stats, that’s a sign that your offensive line is better than your running backs in the run game.
  • Opportunity Rate is the percentage of run plays that gain at least six yards. All runs to hit six yards are potential home-run hitters. The more a back generates runs of 6+ yards, the more likely he is to break a long one. This stat is single most important run-related stat (besides Rush EVA itself, of course).
  • Bonus-Yards per Opportunity is the average yards gained beyond six for any opportunity run. Think of it this way: the first six yards of any run are “line-yards” and all yards after that are “bonus-yards”. So a 16-yard run is worth 6 line-yards and 10 bonus-yards. This statistic is only calculated if the run is considered an opportunity. Explosive backs always have high bonus-yards. If this stat is significantly better than line-yards, that’s a sign that your running backs are better than your offensive line in the run game.

Now for the passing stats:

  • Adjusted Net Yards per Attempt (ANY/A) is the average yards per pass attempt (sacks are pass attempts, too) with one caveat: touchdowns are worth 20 yards and interceptions are worth minus-45. These are not random numbers: they are the result of significant calculation into the ‘yardage value of a touchdown’.
  • Sack Rate is the percentage of dropbacks at end in a sack. This may be the most obvious stat on this list.
  • Adjusted Net Yards per Target (ANY/T) is just ANY/A but for receivers. Targets are used as the denominator.

Other stats

We’ll use a few other interesting stats in our analysis.

  • Marginal Third Down Conversion Percentage is third down conversion percentage adjusted based on the yards to go. Converting a 3rd and 1 is easier than converting a 3rd and 13, right? A logarithmic equation ( 0.7294 - 0.192 * ln(yards to go) ) gives you the chance. For example, a 3rd and 3 has a 48.1% chance of conversion. A positive marginal third down percentage means that the offense converted more third downs than expected, while a negative percentage means they converted fewer.
  • Misery Rate is the percentage of a team’s offense plays that include a lost fumble, an interception, or a sack. When calculated from the defense’s perspective, we call this havoc rate.
  • Leverage Rate is the percentage of offensive plays that the offense is “on-schedule”. This is done by divided all plays into two groups: standard downs and passing downs. Passing downs are situation where a pass seems obvious: 2nd down and more than 7, or 3rd/4th down and more than 3. Every other play is a standard down. One goal of an offense is to keep its leverage rate (percentage of standard downs) as high as possible, so the defense has to keep guessing.

We will also do the “Big 3” stats (EVA, Success Rate, Gini) for both standard and passing downs, to understand where the offense was better. Leverage rate helps contextualize that. You’ll also see run rate, which is the percentage of plays that are runs.

Drive-level stats

These stats will generally be represented graphically, not in a chart.

  • Available Yards is the percentage of yards gained by the offense out of the total possible. There is a maximum number of yards an offense can gain in a game based on the start of each drive. For example, if an offense starts at midfield (so, 50 available yards) and stalls out at the 30, then it picked up 40% of the available yards (20 out of 50). As you can see, this much better than “yards gained per drive” because it accounts for where the offense started.
  • Points per Drive is the basic measure of the offense’s ability to get points.
  • Scoring Chance % is the percentage of a team’s drives that include a first down inside the opponents’ 40-yard line. Once you get a first down inside the 40, you need to score.
  • Points per Scoring Chance measures how good the offense was at converting scoring chances into points. It’s important for measuring one of football’s key factors: finishing drives.

Putting it all together

Now that we’ve met all the stats, let’s see them in action. Remember the EV report from the Portland State game above? Here’s that same game with all these other stats:

There’s a lot here, but it’s all in this glossary. The top section tells us that both Arkansas and Portland State were bad on offense, but Arkansas was more efficient while Portland State was more explosive. The rushing section tells us that Arkansas was much better on the ground, thanks to a lower stuff rate and better line-yards per rush numbers. The passing section tells us that Portland State was abysmal through the air, posting a success rate of just 31% (anything below 40% is quite bad).

The “key stats” section tells us that Arkansas was slightly better on third down and in leverage rate.

Finally, the play type data tells us that Arkansas was fantastic on standard downs, but struggled in passing situations. Portland State was better in passing situations but struggled on early downs.

Have any questions about these stats? Drop a comment!