Measuring the Tide with Advanced Stats

Bama's offense against Kent State: worse than we thought?

We're all familiar with traditional measures of offensive and defensive statistics: points scored/allowed, yards gained/allowed, time of possession, etc. These statistics, however, can be really misleading (in the case of points scored and yards gained) or downright irrelevant (in the case of time of possession).

Advanced statistics are a superior method of measuring the effectiveness of a team, and particularly its component parts: offense, defense and the various special teams units. At some point in the near future, I'll have a lengthy fan post up for reference explaining in more detail how these advanced statistics are derived, and include some appendixes with the data on which they are based.

However, in the meantime, here's the quick and dirty gist. Each offensive and defensive possession has an expected value based on where on the field the possession starts. For example, an offensive possession that begins 72 yards from the end zone (a team's own 28 yard line) has an expected value of 1.78 points. This would be the average number of points scored by all college football offenses over the last four seasons in all games played between FBS opponents on possessions that begin at that specific spot. Naturally, the expected value increases the closer the drive starts to the end zone, to the point where a drive that starts on the opponent's 1 yard line has an expected value of 6.53 points.

An offenses's scoring effectiveness, then, can better be measured based on the value added or lost based on the expected value of a given drive. An offense that marches down and scores on a drive that started on their own 28 would then add 5.19 points in value to the team (6.97* minus the expected value of 1.78), while an offense that punches it in from the 1 would only add value of 0.44 points (6.97* minus the expected value of 6.53).

*A touchdown is worth 6.97 points instead of just 6 because the score leads to the opportunity for an extra point, which has an expected value of 0.97 points. If the kicking unit then converts the extra point attempt, it adds 0.03 points in value, while missing the kick would lose 0.97 points in value.

Below the jump we'll use these advanced statistics to assess the performance of each Bama unit in the Kent State game. Sneak peak of the findings: the 48 points on the scoreboard masked what was, based on these advanced stats, a rather poor overall performance by the Alabama offense against the Golden Flashes.

Before we get to Bama's performance against Kent State, let's explain a little more about the stats we're looking at.

Value added by scoring is simple enough: the actual points scored by a unit minus the expected value of the opportunity they were given. This is a better measure of scoring success than simply raw points, which are misleading for two reasons. First, it doesn't account for field position on each drive. Should an offense get full credit for scoring 7 points when the defense and/or special teams sets you up inside the 20 to start the drive? Second, raw scoring doesn't account for how many possessions you get each game. A slower, methodical offense that scores 30 points on 10 possessions is more effective than a high-speed offense that scores 40 points on 15 possessions, because scoring more per possession limits the opponent's chances of scoring themselves.

However, while scoring is the primary way an offense can add value to a team, it is not the only way. The offense also influences the expected value of the opponent's following possession by determining the expected field position for the opposing offense.

For example, when a team starts a drive 72 yards from the end zone (their own 28), the average starting field position for the opponent's next possession is 70 yards from the end zone (the opponent's own 30). If the offense goes 3-and-out and has to punt from their own 28, they are giving the opponent better field position that what was expected. The opponent's field position is determined by adding the distance of the average net punt (approximately 33 yards) to the point where the drive stalled. In this example, that means the opponent's drive would start only 60 yards from the end zone instead of the expected 70. In this case, then, the offense not only loses the 1.78 points in value for failing to score, but they also increased the opponent's chances of scoring by 0.34 points by giving them better field position. The offense, then, would actually lose 2.12 points in value on that drive because they hurt their team in both offensive metrics.

As OTS has correctly pointed out to all who will listen, time of possession is, in and of itself, meaningless. Yes, by holding the ball you limit the other team's possessions.....and you also limit your own. All you are doing is limiting possessions by both teams, which does you no good unless you score more points per possession, which is why that is the metric we are interested in. Further, statistical evidence suggests there is no benefit to a defense being rested after a long drive by the offense. Basically, time of possession is meaningless--with the notable exception of when a team is trying to run out the clock late in the game. All that matters is how efficiently you score when you get opportunities, and to a lesser extent, the field position you set up for your defense on the following possession.

Similarly, kickoff coverage, kickoff return, punt coverage and punt return units can all add or lose value based on how they influence starting field position relative to the expected value. For example, as was noted earlier, the average net punt is about 33 yards, so if the punt coverage unit actually executes a 40-yard net punt, they have added the value of pushing the opponent's starting field position back 7 yards from the expected point.

Again, I'll do a more thorough explanation of all this in a separate post, but in the interest of getting to actual stats before this becomes a 20,000 word essay, let's take a look at the Bama defense's advanced statistics against Kent State.

Note: we are only assessing non-garbage-time possessions. Essentially this means excluding all drives where the starters weren't playing and all drives that begin too late in a half to have any real chance of scoring. This is important, because garbage-time stats can really water down statistics for both teams and individuals if included.

For the Kent State game, the last offensive drive in the first half, where Bama received the ball on its own 15 with less than a minute left in the half, and all drives that started in the 4th quarter (about the time Saban started pulling starters) will be counted as "garbage-time" possessions and excluded from all the data below.

Let's take a look at the Tide defense's performance, which, as we all know, was thoroughly dominating.

Note: field position is expressed in distance from end zone.

So what are we seeing here? Basically a statistical representation of what we saw on Saturday: the Tide defense absolutely stifled the Golden Flashes on nearly every single possession. Even the one possession that ended with a negative value--the one Kent State score--only ended with the defense losing less than a point in value. This is because the defense was put in such a bad spot thanks to the offense, that their expected value was already -5.99 before they even stepped on the field. This is also a prime example of why it's better to assess units based on starting field position rather than just raw scoring.

The value added of +28.97 doesn't sound terribly dominant in and of itself, but when you consider that the defense produced positive value on 14 of its 15 non-garbage-time possessions, you can see the consistency was clearly very high. The total value added would have been much higher except that the defense did not force a turnover while the starters were in, so the defense never really set the offense up with a prime scoring opportunity the way that we will hopefully see later on this season. Still, the defense generated 4.56 points in field position value by forcing so many 3-and-outs deep in Kent State territory, even without a turnover.

The stats for the offense aren't so fun to look at. In fact, they're a bit concerning.

As you can see, the offense ended the day with negative value added (at least before garbage time). Not good. Especially not against Kent State.

You may be wondering how we ended up with such ugly stats when when we put 38 points on the board before most of the starters left the game during garbage time. There are two reasons for this. First, the offense was consistently given great field position, thanks to the Bama defense and punt return units, which made each successful drive less valuable for the offense than the scoring might indicate. Second, the 4 turnovers (in only 13 non-garbage possessions) lead to huge swings in field position, for a total loss in value of 15.78 points.

Here's an illustration: on the drive that ended with the Philip Sims pick on the slant that led to a Kent State return to the Bama 3 yard line, the offense not only failed to score and therefore lost the expected value of 1.81 that they started the drive with, but they also swung field position so heavily in Kent State's favor (Kent started 3 yards from the end zone instead of the expected 70) that the turnover increased the Kent State offense's expected value by 4.51 points. Therefore the offense generated -5.96 points on that drive.

Think about that for a moment: those 4 interceptions lead to an expected loss of 15.78 points--a total that would have been even higher except 3 of the 4 picks had no meaningful return. Against a better opponent, particularly when playing on the road, a swing like that will likely be fatal. At any rate, it greatly reduced the offense's value for this game, and had it not been for the Bama defense's dominance, this likely would have shown more on the scoreboard.

The good news is that on the 9 non-garbage-time possessions where the ball wasn't turned over, the offense performed pretty well, scoring touchdowns on 5 of those possessions and setting up a field goal attempt on a 6th. Again, however, the offense often had a short field to work with, so the value added by the offense was somewhat limited.

In other good news, all five Alabama special teams units produced positive value overall. On the surface, the punt return unit was the strongest. However, the numbers are inflated because of the botched snap by the Kent State punting unit, which led to a 3.99-point swing in value. Without that play, the Bama punt return unit barely broke positive. That might seem counter-intuitive, since Marquis Maze broke a few fairly long returns, but these returns were largely offset by the solid punt distance throughout the day by the Kent State punt unit.

The expected field position is where you would expect the offensive drive to start based on the location of the punt. The actual field position is how far the punt return unit actually advanced the ball.

Cody Mandell and the punt coverage unit were about as effective as you could ask on the three times they were called upon.

Jeremy Shelley and the field goal unit also had a solid day, providing value of 1.44 points. Cade Foster's missed 53-yard field goal barely registered, as the expected value of a college kicker making from that distance is simply so low.

The kickoff coverage unit got plenty of work on the day, and ended up adding positive value to the team.

The kick return unit got the least work, with only two opportunities on the day, but a big return on the opening kickoff of the second half led to a positive result overall.

Overall, the special teams units combined added 7.33 points in value to the team pre-garbage time, a high amount to be sure, but all but 3.34 of those points came from the botched Kent State snap. Still, the special teams units clearly performed well--at least relative to the competition on the field.

And that brings me to my final point. All of these advanced statistics are only meaningful to the extent that they merely assess your performance against a given opponent relative to the expected performance of an average FBS team. What does it really mean that our defense produced 28 points in value against Kent State's offense? Maybe nothing, maybe a lot. It all depends on how Kent State's offense performs in advanced metrics against other defenses later this season.

Once enough games are played, we'll be able to adjust these statistics to control for opponent strength. At that point we can assess where each unit on Bama's team stacks up against other units in college football based on these advanced metrics. This is essentially what FEI does, which we'll be leaning on heavily later this season once we have enough data to adjust for opponent strength.

Trending Discussions

forgot?

As part of the new SB Nation launch, prior users will need to choose a permanent username, along with a new password.

Your username will be used to login to SB Nation going forward.

I already have a Vox Media account!

Verify Vox Media account

Please login to your Vox Media account. This account will be linked to your previously existing Eater account.

As part of the new SB Nation launch, prior MT authors will need to choose a new username and password.

Your username will be used to login to SB Nation going forward.

We'll email you a reset link.

If you signed up using a 3rd party account like Facebook or Twitter, please login with it instead.

Try another email?

Almost done,

By becoming a registered user, you are also agreeing to our Terms and confirming that you have read our Privacy Policy.

Join Roll 'Bama Roll

You must be a member of Roll 'Bama Roll to participate.

We have our own Community Guidelines at Roll 'Bama Roll. You should read them.

Join Roll 'Bama Roll

You must be a member of Roll 'Bama Roll to participate.

We have our own Community Guidelines at Roll 'Bama Roll. You should read them.