Stat Analysis
Advanced analytics on player and team performance

Evaluating Goal-to-Go Strategy with Expected Points

Guest column by Spencer Harrison, Sports Info Solutions

Those coaches who are really good at knowing what to call in a given situation contend for Super Bowl titles (so long as they have the talent that can execute their wishes). Those who aren't good in that area may frequently be riding the coaching carousel. With that in mind, it is possible to ascertain what the most successful play calls for a given situation are using Expected Points Added (EPA) as the tool for evaluation.

As an example, here are looks at the most effective routes and run schemes in goal-to-go situations.

Route Types

How often have you seen a quarterback throwing a fade into the end zone with his team close to the goal line, sailing the ball over or off the receiver's outstretched arms as he jumps to try to make the play?

Dating to 2016, the fade (combining both traditional and back-shoulder varieties) is the second-most-targeted route type in goal-to-go situations. However, EPA suggests that this is one of the most inefficient throws a quarterback can make in these situations.

Over the past three seasons, there have been 14 route types (as classified by Sports Info Solutions Video Scouts) to garner at least 50 targets in goal-to-go situations. Of those route types, six average a positive EPA per target. The routes that do well tend to be shorter and more lateral with higher completion percentages. Drags work best in this situation, followed by throws into the flat.

Goal-To-Go Target Types, 2016-2018
Best Performing Target Types Worst Performing Target Types
Type Plays EPA Positive
Play%
Type Plays EPA Positive
Play%
Drag 145 0.35 52% Fade 275 -0.26 29%
Flat 290 0.14 48% Seam 59 -0.21 36%
Out 241 0.10 47% Whip 63 -0.14 44%
Slant 267 0.09 50% Corner 145 -0.04 41%
Dig 122 0.03 40% Curl 166 -0.02 43%

Curl, corner, whip, seam, and fade routes all result in a negative EPA per target, with fades typically costing a team the most. Fade routes in goal-to-go situations are completed just 29 percent of the time, so they very frequently result in a loss of down or worse.

SIS's Total Points model, which takes granular charting data and determines how much each player on the field contributed to the play, can also be applied and analyzed across these situations to show who the best and worst performers are.

For example, the most effective goal-to-go quarterbacks in Total Points on pass attempts (including sacks) in 2018 were Patrick Mahomes (11) and Drew Brees (10). That's expected, but the least effective ones may surprise -- Mitchell Trubisky (-14 Total Points), Tom Brady (-12), and Deshaun Watson (-12).

Run Schemes

There is a similar disconnect in goal-to-go situations between the most frequently used run schemes and the highest average EPA run schemes.

Among schemes with at least 30 plays, inside and outside zone runs have been called most frequently since 2016, with inside zone runs accounting for more than 40 percent of all runs in goal-to-go situations.

The most successful runs tend to be those by quarterbacks, including scrambles, though those plays (which aren't called by an offensive coordinator) put the most important player on the field at greater risk. As a result, those plays have been removed from the table below.

When the ball is put into the running back's hands, the most effective run schemes in goal-to-go situations tend to involve pulls, with power, trap, pitch, and sweep all performing well.

Goal-To-Go Rush Types, 2016-2018
Best Performing Rush Types Worst Performing Rush Types
Type Plays EPA Positive
Play%
Type Plays EPA Positive
Play%
Power 239 0.10 50% FB Dive 40 -0.16 50%
Trap 40 0.09 50% Stretch 66 -0.08 42%
Zone Read Handoff 98 0.08 47% Counter 56 -0.06 41%
Pitch 77 0.06 49% Lead 208 0.01 45%
Sweep 32 0.05 41% Inside Zone 1043 0.01 50%

One fear with pull scheme runs might be the possibility of a defender shooting the gap and blowing up the play for a loss. However, the rate at which those runs cost a team at least one expected point is in line with most other runs, while outside runs like stretch and lead plays tend to get blown up slightly more frequently than others.

Inside zone is essentially neutral from an EPA perspective, but actually produces a positive outcome more often than some of the schemes that fall above it in average EPA per play.

Our next article on the subject will look at another situational strategy -- what works and doesn't work on third-and-long.

Sports Info Solutions, based just outside Allentown, Pennsylvania, sells football analytics to teams and media. Sign up for a free week-long trial of their NFL or College Football data hub today at www.SISDataHub.com.

Comments

5 comments, Last at 29 Jan 2019, 5:37pm

4 Re: Evaluating Goal-to-Go Strategy with Expected Points

Question for the author:
If the fullback dive was a successful play 50% of the time (20 out of 40), the negative plays must be really pulling down the EPA. Were there some spectacular failures that pulled the average down? What are the median and mode of these numbers?
Logically, the EPA and success rate are positively correlated, but as the sample size gets smaller, there can be more variation. A little bit more explanation of this particular play would be helpful.
Also--it would have been nice to see the #'s for QB sneak, QB scramble, and QB draw. QB draw would be a called play, scramble would not be, and a sneak may or may not be--probably an option/audible, depending on the QB and the defenses alignment.

5 Re: Evaluating Goal-to-Go Strategy with Expected Points

Love the info. Would love to see a breakdown of run versus pass playcalls in various goal to go situations in a future post.

If I were an OC I think I'd run the ball three or four straight times (the number of times depending on score, time, strength of my defense, etc.) almost every time with goal-to-go from the 1 or 2 yard line. Obviously if I have one of the great QBs at the helm or a really bad rush offense this changes, but aside from the rare fumble the outcomes are touchdown, field goal, or the opposing team has to drive 99 yards for a TD. Even if they know you're calling a run play, they can't ignore receivers completely, and they still don't know what run play you're calling. And even if they know you're running, it's still hard to stop a team from gaining a yard on four consecutive plays (obviously a tackle for loss changes this).

Conversely, with goal-to-go from 9-10 yards out, I think I'd mostly pass three straight times. I never understood the point of running on first down that far out. You're almost never going to score on that run, or even get to the 1 or 2 yard line where you can run it again. Most likely you'll pick up 3-4 yards. And when the defense has 19-20 yards of field to cover, if you pick up 4 yards they now have 20% less field to cover (15-16 yards), so it's not like at midfield where picking up 4 yards on first down leaves the defense with essentially the same area to cover. So to me, I'd rather have one more pass opportunity than 25% less field on which to score.

This is of course entirely based on watching football for years and is just my hunch. I could be totally wrong. Would love to see the numbers one way or the other.