# Two-Point Study

By Noah Riley

In the 2017 NFL season, there were 96 two-point conversion attempts, many of those coming in key situations which decided the outcome of the game. There were also several 3rd/4th and goal plays that had a critical impacts on the outcome of games. In the past two super bowls alone, there were four 2-point conversions attempted. Without converting on both of their attempts in Super Bowl LI, the Patriots don’t beat the Falcons. In that same game, the game winning touchdown was on the 3rd 2-point play the Patriots had in their game plan. Also, the famous “Philly special” that helped the Eagles win last year’s Super Bowl was run on 4th and goal (which is a similar situation to a 2 point play). Since many games are won and lost on 2-point conversion type plays, it is important for a coach to have a great plan to convert.

Using NFL Game Pass, I was able to view all 506 2-point conversion attempts from the past 7 seasons. I then labeled each play 3 different ways (run/pass, play-type, and exact play), and kept track of whether or not the play converted.

According to the data I collected, teams converted on 247/506 2-point conversion attempts (48.8%) which is consistent with other studies done on NFL conversion rates. I first ran the data on run/ pass plays (play calls, not results, so sacks/scrambles are pass plays)

I then put each play into one of ten play-type categories in order to see the success rate of each type of play call.

One thing to note here, is that if you take goalline/ roll-out passes out of the data set, the conversion rates would be raised from 247/506 (48.8%) up to 213/399 (53.4%). If NFL teams simply eliminated these two play calls from their two point conversion arsenal, their conversion rates would rise 4.5%.

Specific plays: Within each play call category, I kept track of which individual play calls were most efficient. (Each play name is hyperlinked to a video of the play)

Quick pass: 42/97 (44%)

 slant corner 15/28 snag 4/16 double slant 2/8 goalline slant 0/7 stick 1/6 4/5 slant flat 1/4 vert-out 4/6

The majority of defenses that are run on two point plays are either man, or banjo/ bracket combos of man coverage. In the passing game, it seems that plays designed to beat bracket/ banjo coverage do best, and the plays that did poorly tended to be ones that can be matched by bracket/ banjo coverages. When calling 2 point plays is important to understand the rules of the defensive coverage, and have concepts that exploit them.

Drop back pass: 29/58 (50%)

 follow 8/11 mesh 5/8 smash flat 4/6

The follow concept was one of the most efficient concepts in the study. Since it is mostly run from a compressed set, defenses usually check to banjo coverage which allows the inside receiver to work an inside breaking route against an outside leverage corner.

Roll out pass: 19/59 (32%):

Pure roll-out passes performed very poorly statistically. This makes sense to me because on the goalline, the offense already has condensed vertical space; by rolling out, they are also shrinking the horizontal space. The few plays that worked were when the QB’s first progression to a flat route popped open and the QB hit it early. However, roll-out plays would normally end up with the QB’s roll-out action bringing the majority of the defense to the concept side, taking away space for the concept to break open. The fact that roll-outs were statistically ineffective is particularly interesting because many coaches install roll-out passes specifically for 2 point conversions.

As has been shown in other statistical studies, fades aren’t good calls on the Goalline with teams only converting 31% of the time. You may think that they are good calls if you have the right matchup, but most teams that ran them, did so because they thought they had a good matchup, and they still preformed poorly.

pick: 27/50 (54%)

 2 man pick flat 7/11 2 man pick in 6/10 3 man pick in 5/11 3 man pick flat 3/6 2 man pick wheel 3/3

The combination of similar success rates of each play call, and the small sample size makes it hard to say any particular type of pick play is especially efficient. The 2 man pick wheel play has been successful every time, but it still is a small sample size. Pick plays worked great against pure man coverages, and sometimes slipped open when there were miscommunications in banjo coverages.

Boot: 20/38 (53%)

Bootlegs did fairly well statistically. Surprisingly, I could not find a type of bootleg that stood out as more effective than the others. Whether it was off outside or inside run action, or from under center or out of the gun had no statistical significance. Also the route concepts teams ran all had similar success rates. The majority of bootlegs were slight variations of Y Cross, the most common of which is adding some kind of late leak route, Late leak variations went 6/12 (50%) and other variations of y cross went 10/21 (48%) which doesn’t show and statistical difference in success rates of the plays. There were also 3 bootleg pick plays which went 1/3, and 2 dash plays (QB pumps one way, rolls out the other), and those both converted.

Inside run: 46/74 (62%)

 power 16/18 zone 26/46 zone read 3/4 qb run 2/2

Since inside runs did very well statistically, and had a large sample size, it is safe to say that they are a very efficient play call. Power performed exceptionally well. Power was almost exclusively ran out of the shotgun from 11 personnel with either the Y off or the Y on the ball. Another thing that stood out is that any time the QB’s got involved in the run game, teams did well. This makes sense, because involving the QB gives the offense an extra blocker in a situation where the defense can usually find a way to be +1 in the run game. This could also be slightly mis-leading since many runs appeared to be checked to or run as RPOs.

Draw: 11/15 (73%) (11/13 for QB draws)

To me, the quarterback draw stood out as statistically the most efficient play in the study, converting 11/13 times (85%). The other surprising thing I saw watching this is that some QB draws were run by QB’s who are not known for their running abilities. Joe Flacco & Matt Ryan converted on QB draws. There were also two running back draws that both failed to convert.

Outside run: 9/15 (60%)

 fk run pitch 3/4 fly 2/3 outside zone 2/4

While there wasn’t a large sample size, it seems as though outside runs are fairly efficient play calls.

Other:  29/47 (62%)

 pa pass 5/13 pa pass, pop pass screen 3/7 swing, bubble, tunnel dbl move 6/9 stick nod, fk screen, fake pick throwback 4/6 throwback sneak shield 5/5 slant, screen, pitt special trick play 3/4 QB pass, QB pitch, direct snap

Of the rest of the plays, I found the shield concept to be particularly intriguing. While the sample size is small, it did work every time. Against man, the man on the screen/slant defender had to bubble over the block which makes it an easy completion for the offense, and since the inside receiver is still going inside, and actually blocking the man assigned to him, banjo coverages can’t switch responsibilities as they do against normal pick plays.

Conclusion: I hope coaches can take something from this study that will help with play-calling in two point conversion type situations. This study not only can help coaches with play-calling on two point conversions, but it also gives coaches insight for lower red-zone/ goalline play-calling. When calling 2-point type plays it is important for coaches to know and understand the defensive scheme/ personnel they are facing and find ways to attack them, and also consider the data and what has worked for other teams in the past.

If anyone has any questions or comments on the article, you can contact me by email at noahbriley@lclark.edu or on twitter at @noahriley21. Also, you can my Book Breaking Down The Oklahoma Offense which is now available on Amazon.

## 8 thoughts on “Two-Point Study”

1. Sean Peterson says:

This is tremendous! Thanks for doing this!

On Sun, Jul 8, 2018 at 4:40 PM Riley-Kolste Football wrote:

> Noah Riley posted: “By Noah Riley In the 2017 NFL season, there were 96 > two-point conversion attempts, many of those coming in key situations which > decided the outcome of the game. There were also several 3rd/4th and goal > plays that had a critical impacts on the outcome of ” >

Like

2. Too small a sample size especially with so many degrees of freedom (types of plays) to draw any conclusions. Also, you are ignoring game theory. The more you use any particular play the less effective it will be because the opposing team will be able to anticipate it more often. Game theory says that if the mix of plays is chosen optimally for that team and that opponent under those conditions, all plays should have the same chance of success. With so many types of plays and such small samples, we expect the empirical numbers to be all over the board because of random chance. Just looking at the numbers it seems somewhat consistent with all plays having the same chance of success with random variance added “on top.”

Like

1. Elie says:

You’re right about the game theory. But going for runs only 20 percent of the time when it’s by far the better play to passing is crazy.

And the data set is large enough to be 100 conclusive this didn’t happen by chance. The breakdowns of the passing section may be too small to be fully confident on its results, but the passing vs running is conclusive. You can use this tool to check for youself:

https://neilpatel.com/ab-testing-calculator/

Like

3. gabrielcerda says: