Filed under:

# On running out of gas and burning out in reverse

The Seahawks have been awash in costly holding penalties this year, and they've flopped on 3rd-and-short. Here's a closer look at how they stack up to the rest of the league.

Once again, it's time to look at drives!

I thought after last week's game that it would be interesting to see if Seattle is really seeing more long 1st down situations than the rest of the league, and whether it's doing better or worse than average with those situations. I'll get to that way down at the bottom of this article, so if you're impatient, feel free to jump to there. But if you're not impatient, there may be some rewards for you -- or just confusion at the statistical nonsense I came up with to do this in the first place.

The first problem is how to define "better or worse" in this context. Long down-and-distance combinations are hard to dig out of, so teams tend to play a very different style of football in those situations, resulting in success or failure that doesn't map well to binary metrics. Presumably, a 3rd-and-8 is less likely to be successful than a 3rd-and-3, but a 2nd-and-12 might be even less likely to be successful than a 3rd-and-8. Let's explore.

First, I'm going to modify the usual definition of "success" to include both fractional success and extra success. Obviously a 2-yard run on 1st-and-10 is relatively unsuccessful, but we want to sort out how unsuccessful it is relative to, say, a 4-yard run, or a 6-yard run. But any run that gets a first down is clearly fully successful, and I'm not about to give bonuses for racking up 10 yards on 2nd-and-8, or 30 yards on 3rd-and-2, even if your name is Thomas Rawls.

Here's my definition of "Success+":

1. For a 3rd down play, it is "successful" if it gains all the yardage necessary for a 1st down. There are no partial successes on 3rd down, so this is 0 or 1.

2. For a 2nd down play, the nominal success metric is 2/3 of the distance to go. (This varies by source, but I prefer the 1/3, 2/3, 3/3 measure, since it simplifies the math.) On 2nd down, I've allowed for partial success as well as "success-and-a-half" -- which is to say, a 1st down. This metric will go from 0 to 1.5, with a 1.5 being a first down.

3. For a 1st down play, I use 1/3 as the nominal success metric, but rather than having a max at "successful 3 times over" (i.e. a first down), I've chosen "doubly successful" as a 1st down. That means that yardage above the 1/3 mark is worth 1/2 that of yardage below that mark. The reason? I don't want an overly heavy weighting on 1st downs that result in a 1st down, but I do want to recognize that they're important. The metric on 1st down thus goes from 0 to 2, with 2 being a first down.

Success+ should give us a baseline for assessing down-and-distance: how often does the team succeed on a particular down-and-distance combination? It's similar to the Football Outsiders idea of play value, but in extremely stripped-down form.

In particular, it will give us a way to consider how likely it is that a particular down-and-distance gets us to another 1st down. To determine this, I'll take

Success+(total) = Success+( 3rd-and-2*N/3*( Success+( 2nd-and-N/3*( Success+( 1st-and-N ) ) ) )

This is, of course, not a perfect way to deal with the situation, but it's passable, assuming you like nested functions.

Here's what Success+ looks like by down-and distance across the league, where I've culled all the down-and-distance combinations that don't even produce 20 plays across the league and reduced the 1st down data to just 1st and {increments of 5} so that we don't count too many plays run near either team's goal line. Yes, this is a wee bit of cheating, but (a) offense near the opponent's goal line (say, on a 1st-and-3) is really different and (b) offense near your own goal line (say, on a 1st-and-12) is somewhat different. Plus, the number of plays in these regimes is extremely low compared to the increments of 5 surrounding them, so I don't want to make too much of a mess here.

Success+ without smoothing, by down and distance

I have no explanation for why 3rd-and-3N for N > 2 seem to be a little worse than 3rd-and-{3N +/- 1}, but that's the data. Since 2nd and 3rd down are pretty complete datasets, here's a windowed average plot using the distance +/- 1 to smooth it. (Again, I've avoided doing this with 1st down because 1st-and-5 really is 1st-and-5, whereas there's going to be some discretion on any "natural" 2nd-and-5.)

Success+ leaguewide, smoothed with a windowed average for 2nd and 3rd down

Let's put this to use by looking at a typical 1st-and-10 situation:

1st-and-10: Success+ ~ 0.86 -- Expected distance achieved = 10/3*0.86 ~ 2.9 yards
2nd-and-7: Success+ ~ 0.7 -- Expected distance achieved = 7*2/3*0.7 ~ 3.2 yards
3rd-and-4: Success+ ~ 0.42

The expected success of this "sample" drive is then 0.42, so 42% of the time we expect to see a 1st down from that original 1st down.

All that is a whole ton of introduction that should let us compare team-to-team results. Here are the Seahawks, with the rest of the league still strapped on the chart.

Success+ for the Seahawks, compared with league average, smoothed

Several things going on here that are worth taking a look at. In prelude, the dataset is pretty limited, so I've removed the extremely low-sample elements and limited this view to downs and distances where Seattle has run at least 5 plays. That's not to suggest the remaining data is unreliable: for short-distance situations, the sample sizes are plenty large enough to evaluate.

Right away, we can see that Seattle is slightly above average at converting on 1st down until they hit 1st-and-20: the team on average gets about half a yard less than the rest of the league on 1st-and-15 and 1st-and-20. More on that in a moment.

In other good news, Seattle is great at 2nd down conversions, finding ~20% more success over the league average on 2nd-and-10. What does that mean? That should translate to shorter 3rd downs, turning an expected 3rd-and-6 to an expected 3rd-and-5. That's huge!

Except when it isn't. That sinusoid on the bottom? The Seahawks' conversion rates on 3rd down. The team is somehow about 50/50 in the 3rd-and-{6,7,8} range but absolutely terrible on 3rd-and-{3,4,5}. The relative Success+ here is about the same as the relative annual rainfall in Los Angeles vs Seattle.

Before I go completely off the rails, I will note that this is a pretty small sample size. Seattle has run 13, 13, and 11 plays of 3rd-and-{3,4,5}, respectively; that's about 2/3 the number of plays run on 2nd-and-{6,7,8}.
That having been said, there's obviously something going on. The Seahawks are simply not converting 3rd-and-short, even though they're apparently capable of converting 3rd-and-medium. (Dropping in sample size by about 25% to get to those other distances, this is territory we should consider marginally valid data at best.) I have no ready explanation for this -- thoughts, anyone?

Right, back to the impetus for this post: penalties and long down-and-distance situations. Seattle's got a 1st-and-long problem -- a big, big, big 1st-and-long problem. And a 2nd-and-long problem. And even a 3rd-and-long problem. It turns out all that Success+ business above isn't particularly helpful, but it got me some useful subsidiary data.

The Seahawks have run 8 1st-and-20 plays. For those of you keeping score at home, that's 8 of the league's 158, or about 1/20. Not terrible, but that's 50% more than the average team. Just so you know, 4 (half!) of those were against the Cardinals, so they really did kill themselves with penalties that game. They've also run 2 of the league's 23 1st-and-25 plays, which gives them 10 of the 171 plays of the worst 1st-and-long distances -- essentially a guaranteed disaster for any of those drives. (Worst in this department is the Atlanta Falcons, who have 10 1st-and-20s and 1 1st-and-25.)

As noted above, the Seahawks are also worse at moving the ball on 1st-and-long than the league average, meaning they move back and don't move forward again. By down and distance, their 1st-and-really long performance is below league average, which means they can expect to preserve their undesirable rank.

Rather than try to separate out sacks and penalties for 2nd down, let's just lump them all in one bucket and look at the league as a whole. We'll assume for the sake of argument that anything worse than 2nd-and-14 is the result of one of these events. There have been 489 plays of 2nd-and-15 or longer, and Seattle holds 18 of them. That's quite the turnaround! The Seahawks move from producing 5.1% of the trashy 1st-and-long situations to working just 3.7% of the 2nd-and-long. But even this is above expectation -- 1/32 is 3.1%.

So do they improve on 3rd down? Sadly, no. Seattle is responsible for 14 of the 371 (~3.8%) 3rd-and-15 or longer on the year. That's below where they should be and well below where they need to be if they want to be a playoff contender.

Just looking at Success+ here, it's obvious that their 1st-and-10 mojo plummets when they're penalized. Their expected success after a 1st-and-20 drops from kind of far below league average (23.7%, based on their bad 3rd-and-4 performance; flattening the sinusoid, they're actually about league average) to something below 20% -- an expected 2nd-and-16 followed by a 3rd-and-12 or 3rd-and-13.

We can get even more specific by looking at the worst drives. Against Arizona, all of the drives that included a 1st-and-20 or 1st-and-25 went bad: 4 punts without a first down, and a safety. They've punted all three times they found themselves with 1st- or 2nd-and-20 or more against the 49ers this year. They punted after a 2nd-and-27 against the Packers, a 2nd-and-23 against the Lions, a 2nd-and-20 against the Bears, and a 1st-and-20 against the Panthers. They've successfully converted 2 of these situations into new first downs, which is what Success+ predicts.

Right, you say, but those are hard situations, and most teams can't recover from these long down-and-distance combinations. The problem is that the Seahawks are finding themselves in this hole more often than most teams this season, and they're drive killers.

But with an inexperienced offensive line, it's a tough sell to simply say "Don't get penalties!" More likely, Seattle is going to need to find a consistent way to gain 6 yards on 1st down so those squint-to-see-the-marker plays turn into something more manageable.

If there's a silver lining, it's that 2nd-down conversion rate, and to some extent the mid-range 3rd-down conversion rate. If the Seahawks can skip some of the major backwards hikes and tick their 3rd-and-short conversion rate up to league average, they should see immediate benefits.