NFL Coaches Are Getting Two-Point Conversions Completely Wrong

Historical two-point data can blur designed QB runs and scrambles. Excluding all QB carries, non-QB runs still converted 56.0% versus 45.4% for explicit passing plays, including sacks.

Many have asked about how the offense/defense strengths/weaknesses play into this. Non-QB runs converted 56% vs. 45% for passes. Adjusting for team strengths didn’t erase the gap. Note: Some sample sizes are very small

Defenses expect pass which opens the run. If run becomes more common there’s a good chance these results flip.

100%

There’s obviously a balance. The reason runs are more successful is they are less expect with these frequency numbers. And are being called/checked to in the most advantageous defensive alignments. I suspect the optimal balance is slightly more runs but not an extreme shift.

It’s 75/25 now, which is way out of whack if you look at conversion rate. I think it should start trending towards 50/50 over time (if actors are rational)

This is interesting but I think the primary reason runs have been more successful is because teams that choose to run the ball are a good running team. Everyone would choose to run if they have Derrick Henry or Bijan Robinson.

I checked for that. Better rushing teams have a modest increase in rush usage (true mean is somewhere between 0.2 and 5.4% more runs). That is a trivial difference considering the 75/25 chasm between the plays.

@floppingdimes20 Did you treat teams as being “better rushing teams” due to overall season success or game/drive rushing success? I think the opponent’s weakness against the run may be a big factor in the differential conversion success rates

@floppingdimes20 Overall season. No matter how you slice the data (by team, by scheme, by era, by personnel, by opponent, etc) the trend still strongly skews towards way more passing attempts at a lower conversion rate

Good data. For the 4th and 2 anywhere on the field pass results, you should exclude data where pass is caught > 12 yards. That must account for some meaningful % of the successful times.

I hear you! That was kept in on purpose as I wanted to show why the conversion rates were way closer, and the condensed field is most of the reason why. I may take a peak at looking at only plays under 12 air yards and see what it looks like

pretty crazy that this video will inevitably affect play calling in the NFL

Does this assume coaches have no insight into what they think will work best for their team against each opponent? For example, coaches who decide to run the ball might do so because they feel they have a physical advantage?

This should add some context!

Man, this is a great video! Love the animation too, made it fun.

Love the data, love the retro bowl style presentation. Will definitely use this info next time I am in this situation while coaching an NFL team.

@JohnCUrschel did a whole mathematical write up on this subject.

@JohnCUrschel Cool thanks for linking, although this paper is exclusively about two vs kick, not run vs pass on 2 point tries

@JohnCUrschel Yes, your analysis is great. Just adding that writeup. And he included variance, which would be important to consider. As coaches are certainly risk adverse.

@JohnCUrschel Yes, absolutely! My main point is more that WHEN they go for 2, they should run it more often, not necessarily that they should go for 2 more. Appreciate the added context

In RPO circumstances, I presume you count run/pass outcome respectively (end up running, it’s a run play, pass it’s a pass play). Is there an advantage to running an RPO vs. straight run/pass?

Yes. And also, yes. RPOs seem to be one of the better options.

Thank you kindly for your response.

love this, great work eric. would love to see how the data would be affected if we control or separate by formation - since that seems to heavily affect defensive personnel and play calling.

Probably the best way to run it in is from 11 personnel and spread out the field. I could look into this

There’s an element that you are not accounting for. Runs become more successful if you have shown the defender that you may pass. In other words, it feels like your conclusion is just to always run, but if you did that defenses would lean on the run…they must coexist to be successful…but I don’t disagree that the ratio should flip to put the odds back in your favor.

I see how it could come across that way. Hopefully you don’t think I’m that dumb to suggest they should run everytime just because the success rate is higher 😂 My real contention is they should run more until the success rates start to equalize

Would love to see the passes broken down by play action vs not.

For 2 point tries? I can do that. Also recommend checking out my play action video if you haven't seen it:

I’ve always thought runs are under utilized on 2s but can you parse out designed runs vs QBs scrambles bc I assume the scrambles are skewing this?

Scrambles are passes. Imagine if it was 75% passes if you included scrambles as runs!

@Clevta Does the defensive formation impact the run/pass play call decision? May out bias in the data

If that existed, I still don’t believe it would bias the data. If defenses can actually force offenses to audible from a run to a pass, then stop the pass, that would be a legitimate phenomenon and would still be an indictment of the offense. There’s no evidence that this is happening at a significant rate. I watched 150 two point conversions to help prevent any obvious methodological flaws.

Love this video

This would be different if the tush push was banned

Both have to remain viable options for either to be effective, so whatever the success the successful enjoys, it is quickly countered. What's not properly considered is the athleticism of 350 lb behemoths on crossing routes...or 7 ft. tight-ends....anything other than another god-forsaken back corner fade.

Amazing video

If there was a major shift in coaching philosophy towards runs on 2pt conversions, and defenses started to account/prepare for that, could that have an impact on this analysis?

Yes, it would converge towards equal effectiveness

@CoachHebert9

using historical data of 2 pt and 1 pt tries (only since line was moved back), should any/all teams always go for 2?

Fascinating question. One issue is that teams don’t have an endless Rolodex of plays for that distance. Since the PAT moved back in 2015, two point tries from the 2 averaged 0.963 net points vs. 0.944 for kick. Thats just 2 extra points per 100 tries. Break-even is roughly 47.7%. I would say if you think your particularly good at twos and defending them, theres some +EV there

Thanks kindly for the analysis and insight. Do you think it’s only a matter of time before a coach/team opts for a 2 point try after every touchdown?

Does a pass play that results in a QB scramble get counted as a run in this data? I would suspect that would account for a meaningful amount of these, feel like we often see teams move the pocket on pass plays near the goal line

Check the replies!

this is amazing. Well done and shown.

This is interesting because theoretically a 2-pt conversion rush after every TD is better than kicking the extra point. Although it doesn’t account for the other team knowing exactly what you are going to do.

Negligible, and probably not enough sample size, but defense may also be more likely to return an interception for points than a fumble.

What about passes that break down and turn into runs? What are those counted as?

Hello @Raiders

You failed to account for the influence the pass plays have on the success of the runs. Defenses anticipate pass bc 76% of attempts are passes. Runs can catch them off guard.

No, they should run more for precisely that reason.

Yes but you do understand that, if everyone took your advice, that would be to runs being less effective

The point is they are passing at an irrationally high rate. They should pass less often until teams start to recruit more resources to stopping the run

But how many of the “runs” were passes where the QB had an option to pass but scrambled in?

Less than 80 out of over 2000 plays. Some of the older data it’s not possible to parse out QB designed run vs scramble, but no matter the split it’s not enough to materially change the trend

Not disputing your amazing analysis. That’s all I could thing of since it’s a common goaline strategy to roll the QB out and if defender bites he passes and if he stays with his man he lobs it.

I think you have to distinguish between designed runs and scrambles.

See the follow up post!

Elite research

One could argue that there is a larger emotional tax to getting stopped on a run as pass. We talk about high impact plays. Id guess a team that fails a 2pt on a pass has a higher likelihood of winning the game than a team that fails a 2pt on a run.

Looked into it. Teams won 43% after failed runs vs. 35% after failed passes.

@TravisDHolmes Not privy to the data, but I’d imagine teams that choose to run, know they can get it, whereas teams choosing to pass likely have a poorer run game, specially in short yardage situations. As an example, Ravens and Bills are probably gonna run, a team like the chargers can’t