I think the author makes a few assumptions that aren't true, but he's still correct RE predictable play calling, just not for the reasons he lists:
* That the on field/off field run% splits are indicative of predictable playcalling - this is sometimes accurate, but not always accurate, and often not accurate at all. Knowing whether a play is run or pass isn't always useful if you don't also know where the run is going or what route combination/concept the offense is using. Many plays also have built in tools to help confuse/obfuscate what they are doing/where they are doing it. Think counter runs, option runs, delayed release receivers, etc.
* Nagy decides the outcome of all plays - For the Bears RPO, Nagy doesn't actually decide who gets the ball, he just calls the formation and the play. Mitch has to decide to hand-off, keep, or pass based on his pre-snap and what are supposed to be post-snap reads (I don't think Mitch is doing the post snap reads yet, at least not doing them consistently correct). Theoretically, if one of these players (take Miller for example), saw 40% of his snaps be RPOs, but the QB chooses to run on 80% of their RPO plays, then Miller is going to have a highly skewed on field/off field run %. But hypothetically, this isn't being predictable, its reacting to what the defense is doing and choosing the most successful option. In reality though, I think our QB does have predictable habits on what he chooses for the RPO but that's a different issue.
* This is a data point that is relevant on player by player basis - I think this is the biggest one for me, the methodology the author uses (finding on field/off field run-pass % splits for a single player) would be much better off if he instead did it with 3-5 player personnel groups. Hypothetical example - Robinson as a single player does a lot of stuff on his own. He has run WR reverses; deep routes, intermediate routes, short routes across a plethora of different route combos/concepts; run blocks; etc. Just looking at only Robinsonon a play by play basis he could do anything. But if you instead look at the personnel group of "Robinson - Wimms - Braunecker - Montgomery - Holtz", you get a much more predictable picture of what this GROUP can do. IE, the listed group is the power run package - they are going to run inside power runs on 70% of their snaps, pass on 30% with 12% being PA pass, and 18% being Smash route concepts between Wimms and Holtz + Robinson and Braunecker." If the author was to find this data for 3-5 player personnel groups I think he would see even more polarized split %s, and I think a lot of the Bears predictability comes from these set personnel groups, not the individuals within them. The split that comes to my mind is Wimms - Braunecker - Holtz, when those 3 players were on the field at the same time I would be shocked if they adidn't run on over 80% of their snaps together.
* If defenses know run or pass, they can get a step ahead of the offense to reduce yards - This is often true, but just as often if the defense plays the %s and commits to run/pass, they are setting themselves to be burned by play-action/draw plays designed to punish defenses for doing just that. If a defense plays the %s and is able to reduce an offenses YPC by 1 yard per play on 9/10 plays, but then gets burned for an extra 20 yards and a TD by a play action pass on the 10th play, that's an overall net negative to the defense. Obviously at some point the #s work, IE if player X is 99% run, 1% pass then the above works, but the split almost has to be so extreme to make it work, that you're almost never going to see those split #'s on a single player. I would bet you start to get there though when you look at multiple player groups though, hence my above point.
TLDR: Bears play calling was predictable, but single player on field/off field run-pass splits aren't the reason why.
