Another great read! A couple of questions come up in my mind, reading this article and comparing it to some of the rest of your catalog.
What about Dak’s performances lead you to concluding that keeping him in instead of switching to Romo is the right move? I know your writing well enough that it’s clearly not the winning streak, per se.
You compared it briefly here to the McCown-Cutler situation—but of course, that comes with the big caveat that the respective established QB in DAL was much better than the established QB in CHI. Given that rookie quarterbacks are predictably bad (cf. Falcons-Cousins article) and Prescott was a rookie, how does his performance to that point overcome your prior that we should expect a regression towards poorer performance more in line with rookie quarterbacks? His pre-draft evaluation probably plays a role in that, along with some healthy skepticism about how well Romo would play coming back from a major injury, but I’d imagine there’s something in Dak’s play that leads you to conclude he’ll continue playing well (even if not necessarily forecasting “best rookie QB ever” well).
Relatedly, what kind of a sample size gives you confidence in that kind of prediction? How much shorter would Romo’s injury timeline have to have been to make it a legitimate question, at least from a performance standpoint?
I like several of your questions David, and I'm going to answer them anyways, but I believe your argument is based on a false premise here. I don't recall ever stating that Dak Prescott over Tony Romo was the right move, because I'm not 100 percent sure that it was. I'm fairly sure, but not 100 percent sure.
The reason I'm not 100 percent sure is what you said. Rookie QBs tend to regress as the season goes along, although Dak Prescott is a little bit of a special case. Most great rookie seasons come as a result of a player dramatically overperforming their xEPA/Play. Take CJ Stroud in 2023 as a recent example, where his 0.058 xEPA/Play translated into 0.124 actual EPA/Play, primarily due to having such an elite offence around him. Ben Roethlisberger in 2004 (0.140 translates to 0.215). Matt Ryan in 2008 (0.133 translates to 0.185), and etcetera.
There is no difference with Dak Prescott. He falls into line here too, as his 0.202 xEPA/Play had translated into 0.324 actual EPA/Play through the end of the Steelers game, due to the wicked offensive environment he was in. What makes Dak a special case is that 0.202 xEPA/Play is still a top five QB in the game in 2016. I will admit 2016 is a weak QB season, but even with regression coming around the bend (which it did, Dak generated 0.145 EPA/Play after the Pittsburgh game), it's hard to replace a top five xEPA/Play guy.
That would be my rationale behind the decision. It was not the actual rationale. What happened in real life was that Dak Prescott just played so well that he made the optics of replacing him so bad that he left the Cowboys with no choice. Much like McCown and Cutler in 2013, if Tony came in and struggled even a little bit, there would've been anarchy in that locker room. Football players are a fickle bunch, and will turn on each other in a heartbeat if they think they see somebody better.
Tony Romo had better than a 0.202 xEPA/Play guy a bunch of times, but the simple question is whether you think he could've done that at age 36, coming into a season midstream, off some back trouble. The results in this Cowboy offence would've been fine no matter what, but I'm not concerned with that. I'm concerned as to whether he could've played as well as Dak was playing. There is a possibility that he could've, which is why I'm less than 100 percent sure that Dak was the right move, but that chance is small, which is why I characterise myself as better than 90 percent sure.
I wish Tony would've kept playing, so we could've found a more definitive answer to this question, but alas, TV money is a big draw.
As for sample size concern, the general rule of thumb answer is 11 games. 11 games is the point where NFL statistics truly begin to crystallise and stop varying so much, so I like 11 games as a good sample to begin to tell whether stats are real or not, but it can vary depending on statistic.
Dak Prescott did not get to 11 games before Tony came back, but he did get to nine, which is very close to 11, but in this case, I feel the decision was more about the narrative than anything else. I feel like if Tony could've gotten back at or prior to the Green Bay game in week six, he would've been inserted back into the lineup regardless of how well Dak played, and I would've agreed with that decision, because stats through five weeks are still quite meaningless, unless there's players providing whole season's worth of value in five games (i.e. 1997 Boomer Esiason, 2013 Josh McCown), but once that bye week happened, and the Dak Prescott hype began to swirl in a more and more uncontrolled manner, it was always going to be really difficult to put Tony back in.
I take your point about the impreciseness of my premise. I guess I took my premise from the overall tenor of the article, even if it wasn't stated directly. You favorably compared Dak to Josh McCown (we all know how you feel about that decision); described the fact that there was a question about putting Romo back in as ludicrous (I suppose that describes more the situation that the Cowboys have such a glut of talent at the QB position than the idea that Romo might be better than Prescott); and talked about keeping Dak as the common sense and obvious answer (sure common sense isn't always right, but it often is and is usually applied in a positive manner).
I omitted the disclaimer I had thought about adding to my original comment, so I'll do it now: To be clear, you are not nearly as critical of the idea of Romo coming back as the starter as you are Cutler in the Chicago situation or starting Penix over Cousins. The most negative you get about it is only when it is deservedly worthy of criticism: when Aikman brings it up in the middle of the playoff game. I suppose it could be understandable if Dak had thrown multiple interceptions by that point or was noticeably affected by an injury, but pretty ridiculous in the actual situation. It also makes sense that in a story celebrating Dak Prescott, you talk about how it's a good thing he stayed the starter (it's a good thing from his perspective, regardless of whether it was the right decision for the 2016 Cowboys). I think overall it reflects your now stated position about the correctness of the decision to keep starting Dak--one of strong likelihood but not certainty. Thank you for your expanded discussion on that topic in response to my questions.
Another great read! A couple of questions come up in my mind, reading this article and comparing it to some of the rest of your catalog.
What about Dak’s performances lead you to concluding that keeping him in instead of switching to Romo is the right move? I know your writing well enough that it’s clearly not the winning streak, per se.
You compared it briefly here to the McCown-Cutler situation—but of course, that comes with the big caveat that the respective established QB in DAL was much better than the established QB in CHI. Given that rookie quarterbacks are predictably bad (cf. Falcons-Cousins article) and Prescott was a rookie, how does his performance to that point overcome your prior that we should expect a regression towards poorer performance more in line with rookie quarterbacks? His pre-draft evaluation probably plays a role in that, along with some healthy skepticism about how well Romo would play coming back from a major injury, but I’d imagine there’s something in Dak’s play that leads you to conclude he’ll continue playing well (even if not necessarily forecasting “best rookie QB ever” well).
Relatedly, what kind of a sample size gives you confidence in that kind of prediction? How much shorter would Romo’s injury timeline have to have been to make it a legitimate question, at least from a performance standpoint?
I like several of your questions David, and I'm going to answer them anyways, but I believe your argument is based on a false premise here. I don't recall ever stating that Dak Prescott over Tony Romo was the right move, because I'm not 100 percent sure that it was. I'm fairly sure, but not 100 percent sure.
The reason I'm not 100 percent sure is what you said. Rookie QBs tend to regress as the season goes along, although Dak Prescott is a little bit of a special case. Most great rookie seasons come as a result of a player dramatically overperforming their xEPA/Play. Take CJ Stroud in 2023 as a recent example, where his 0.058 xEPA/Play translated into 0.124 actual EPA/Play, primarily due to having such an elite offence around him. Ben Roethlisberger in 2004 (0.140 translates to 0.215). Matt Ryan in 2008 (0.133 translates to 0.185), and etcetera.
There is no difference with Dak Prescott. He falls into line here too, as his 0.202 xEPA/Play had translated into 0.324 actual EPA/Play through the end of the Steelers game, due to the wicked offensive environment he was in. What makes Dak a special case is that 0.202 xEPA/Play is still a top five QB in the game in 2016. I will admit 2016 is a weak QB season, but even with regression coming around the bend (which it did, Dak generated 0.145 EPA/Play after the Pittsburgh game), it's hard to replace a top five xEPA/Play guy.
That would be my rationale behind the decision. It was not the actual rationale. What happened in real life was that Dak Prescott just played so well that he made the optics of replacing him so bad that he left the Cowboys with no choice. Much like McCown and Cutler in 2013, if Tony came in and struggled even a little bit, there would've been anarchy in that locker room. Football players are a fickle bunch, and will turn on each other in a heartbeat if they think they see somebody better.
Tony Romo had better than a 0.202 xEPA/Play guy a bunch of times, but the simple question is whether you think he could've done that at age 36, coming into a season midstream, off some back trouble. The results in this Cowboy offence would've been fine no matter what, but I'm not concerned with that. I'm concerned as to whether he could've played as well as Dak was playing. There is a possibility that he could've, which is why I'm less than 100 percent sure that Dak was the right move, but that chance is small, which is why I characterise myself as better than 90 percent sure.
I wish Tony would've kept playing, so we could've found a more definitive answer to this question, but alas, TV money is a big draw.
As for sample size concern, the general rule of thumb answer is 11 games. 11 games is the point where NFL statistics truly begin to crystallise and stop varying so much, so I like 11 games as a good sample to begin to tell whether stats are real or not, but it can vary depending on statistic.
Dak Prescott did not get to 11 games before Tony came back, but he did get to nine, which is very close to 11, but in this case, I feel the decision was more about the narrative than anything else. I feel like if Tony could've gotten back at or prior to the Green Bay game in week six, he would've been inserted back into the lineup regardless of how well Dak played, and I would've agreed with that decision, because stats through five weeks are still quite meaningless, unless there's players providing whole season's worth of value in five games (i.e. 1997 Boomer Esiason, 2013 Josh McCown), but once that bye week happened, and the Dak Prescott hype began to swirl in a more and more uncontrolled manner, it was always going to be really difficult to put Tony back in.
I take your point about the impreciseness of my premise. I guess I took my premise from the overall tenor of the article, even if it wasn't stated directly. You favorably compared Dak to Josh McCown (we all know how you feel about that decision); described the fact that there was a question about putting Romo back in as ludicrous (I suppose that describes more the situation that the Cowboys have such a glut of talent at the QB position than the idea that Romo might be better than Prescott); and talked about keeping Dak as the common sense and obvious answer (sure common sense isn't always right, but it often is and is usually applied in a positive manner).
I omitted the disclaimer I had thought about adding to my original comment, so I'll do it now: To be clear, you are not nearly as critical of the idea of Romo coming back as the starter as you are Cutler in the Chicago situation or starting Penix over Cousins. The most negative you get about it is only when it is deservedly worthy of criticism: when Aikman brings it up in the middle of the playoff game. I suppose it could be understandable if Dak had thrown multiple interceptions by that point or was noticeably affected by an injury, but pretty ridiculous in the actual situation. It also makes sense that in a story celebrating Dak Prescott, you talk about how it's a good thing he stayed the starter (it's a good thing from his perspective, regardless of whether it was the right decision for the 2016 Cowboys). I think overall it reflects your now stated position about the correctness of the decision to keep starting Dak--one of strong likelihood but not certainty. Thank you for your expanded discussion on that topic in response to my questions.