Recently, I reread one of my favorite football analytics articles of all time from 538. The TL/DR is that conventional run first reasoning is wrong, and it's better to pass on first down: https://fivethirtyeight.com/features/you-called-a-run-on-first-down-youre-already-screwed/

I wondered how this applied to MFN, so I set up an experiment to replicate the same basic methodology from the 538 article. I took data for NYJ, GBY, PIT, JAX, and MIN for the past 5 seasons since they are consistently the best offenses for that time. I then set up the data to enumerate specific games and drives for those teams within the game. Then found three play sequences in drives where first and second down did not result in a turnover, touchdown, or first down. Then I determined the success of third down based on whether it resulted in a first down or a touchdown.

Here's what the aggregated data looks like:



I'm still working on this, so I'll drop a full analysis here later.

This is from the article which shows how pass-pass-pass was very ineffective in the actual NFL at least in that game. Your data shows it's pretty effective here.
sequence epa success frequency
Pass-rush-rush +0.56 88.9% 5%
Pass-pass-rush +0.50 71.4 4
Rush-rush-rush +0.31 52.0 13
Pass-rush-pass +0.34 50.0 12
Rush-rush-pass +0.17 41.2 26
Rush-pass-rush -0.15 38.5 7
Rush-pass-pass -0.08 34.0 25
Pass-pass-pass -0.39 21.1 10

It's still the least effective series combination for the teams I pulled data for.

Also, I'll drop this in without much explanation of how we got to this. But using the same dataset, this is the optimal Markov matrix for choosing plays based on sequence success:

"Content not viewable in your region".

I curse you North Americans to heck!