January 12th, 2019 at 11:47 AM ^

The chart is an interesting summary in a backwards looking sense. It has little predictive value though because there are all sorts of factors that go into the plays that were called, and the chart can't account for any of those. 

At the very least, I wish we could get a breakdown for our own offense. Mixing all teams into one analysis makes little sense, for obvious reasons. Leach passing is very different than army or ga tech passing. AhlSho they all start calling plays based on this chart? Of course not, because it's predictive value is zero. 


January 12th, 2019 at 8:53 AM ^

I think this is more nuanced bases on down and distance for each play in the sequence, and what the opponent is giving you.

Then again, I was one of the ones in the stands in the mid 80's booing everytime Bo called 3 yards and a cloud of dust on 1st down vs competition who were much more pass heavy. (The Jimmy came along FWIW.)

On the flip, the Bama fans at the Mizzou game I went to were grumbling they weren't running enough on first down ("basketball on grass hurr hurr")


January 12th, 2019 at 5:03 AM ^

Pass run run is the most successful but the 3rd play run is indicative of play number 1 or 2 being a chunk play most of the time in that 3 play scenario.

When the pass was incomplete and the 2 run was a minimal gain, you're going to pass the ball, and that's typically unsuccessful (lowest success rate on the chart).

So, the takeaway from this is that if you pass on first down, you're more likely to get a first down. The success of that is tied directly to if that first down pass was successful, so crossing routes, screens, etc would be your best bet. Going deep is a statistically less likely play to be completed and then your stuck in the worst end of the spectrum. 


January 12th, 2019 at 8:28 AM ^

Whaaa???  Did you read the fine print on the bottom?  It says that the chart charts 1st down plays that gained 5 yds or more, 2nd down gains of 4 yards or more, and converting 3rd into first.  That is all this chart says.

It says nothing about long bombs, other than that they were successful plays that make the chart.  All it does say is that getting a good gain on first and/or second, generally leaves you with a makeable 3rd conversion...  And I think we all knew that!


January 12th, 2019 at 9:19 AM ^

That works great against most teams. But most DC's of elite teams stop the run 1st. Good example is how msu 2015 @ Cbus, clemson 2016, UM 2016&17 and oklahoma 2017 shut down OSU with JT Barret. They were all-in stopping the could probably add 2017 Iowa as well. They were willing to give up the occasional pass plays too shut down the run


January 12th, 2019 at 5:11 AM ^

This suggests that the overall best strategy is pass-pass-run.

It's true pass-run-run is the best *when you must have three plays*. The most successful one sequence play is pass. IF you fail, on second down, the best option is pass. Then on third down you run.

The best overall sequence is pass-pass-run.

Arb lover

January 12th, 2019 at 8:30 AM ^

Not sure if Nate silver just grabbed the numbers or what. Run on 3rd down has a high success rate, regardless of 1+2  but that seems like it's with an average of 3rd and 2 with success of the third down being an EPA of over 1 vs 5 on 1 and 4 on 2.

Obviously NFL coaches are trying to call what's going to work. Pass on third probably has low success because youd rather run if you are close (especially in 4 down territory). So my guess is passes on 3rd down average significantly farther back than runs on 3rd.

If this is how he ran the numbers, really the most telling stat is just 1st down, followed by a distant 2nd as success of any individual series is highly dependent on what happened that first play.


January 12th, 2019 at 9:23 AM ^

Your last point is spot on. Throwing on first down is more successful at the pro level than running it. According to the chart running it on 3rd down is more successful than throwing it regardless of what happened before. As you said, that is very dependent on what happened previously.

I read earlier this year that statistically one of the biggest differences with what offenses were doing to increase scoring was simply passing more on 1st down. 


January 12th, 2019 at 1:35 PM ^

I think the success definition of the ex-post expected points for the drive minus the ex-ante expected points for the drive largely accounts for that.

Keep in mind that the decisions to run or pass bias the data.  The decision to run on 3rd is highly dependent on yards to go.  The breakout doesn't consider success/failure of preceding downs, only the run/pass decision.

One quick interpretation of the higher run success rate for all four conditions is that it is better to convert on third down (and keep your options open for the next set of downs), while it is better to use your options on first down.