10K Simulations (see if this cut and paste job works)
With Ben Roethlisberger’s future continuing to be up in the air, it is fair that many are looking at the other quarterbacks on the roster, those entering free agency, to trade situations, and even to the collegiate ranks in the chance that the Pittsburgh Steelers trade up to snag the quarterback of the future.
That said, as it currently stands, the Steelers will have Dwayne Haskins and Mason Rudolph on the quarterback depth chart if Roethlisberger were to announce his departure.
Steelers faithful generally knows what to expect with Rudolph under center. Haskins is a bigger question mark.
Regardless, entertaining the thought of how each would hypothetically do as a season-long starter for the Steelers could be an interesting exercise.
Thankfully, because of the incredible work of Lee Sharpe (@LeeSharpeNFL) and Sebastian Carl (@mrcaseb), we advanced analytics lovers now have access to nflseedR and the power to explore such hypothetical situations.
In short, nflseedR is a package based in the R-programming language that allows those interested to run thousands – or tens-of-thousands – of simulations across various NFL seasons. Moreover, it allows for the development of unique models for input into the simulation.
In this case, I wanted to explore how a 2020 Steelers’ season would end up with both Rudolph and Haskins under center instead of Roethlisberger.
To do so, you first have to determine how to calculate Roethlisberger’s impact on the team versus the other two. For this simulation, I opted to go with 538’s quarterback-adjusted ELO score. The metric is designed to “account for changes in performance – and personnel – at quarterback, the game’s most important position.”
However, inputting the season-beginning ELO metric for each quarterback is just one piece of the puzzle. In this case, we also want the data to update through each week of the simulation. Using a model pre-developed by the creators of nflseedR, we are able to pass the dynamically-changed ELO metric from week to week to allow for an even more definitive final result.
With that sorted, the process is as simple as swapping out Roethlisberger’s ELO metric for both Rudolph’s and Haskins’.
After that, nflseedR does all the heavy lifting (and heavy math) to determine game outcomes based on ELO, with a quarterback’s ELO evolving and changing throughout the simulation.
Before diving into hypothetical scenarios, let’s establish a baseline by simulating the 2020 NFL season 10,000 times with Roethlisberger as the quarterback.
Ben Roethlisberger: 10,000 Simulations
https://apis.mail.yahoo.com/ws/v3/ma...ils=true&pid=2
Over 10,000 simulations of the 2020 NFL season, the Steelers – with Roethlisberger as the starting quarterback – averaged roughly 9 wins per season. While that is three-games less than the 12 victories this past season, I believe it is a rather accurate depiction as there were certainly a few games that Pittsburgh was extremely lucky to win.
Moreover, with Roethlisberger as the quarterback, the Steelers made the playoffs in 61% of the simulations, won the AFC North in just under 30% of the simulations, and won the conference and Super Bowl 9% and 4% of the time, respectively.
The simulation results for Roethlisberger seem in line with what should be expected. Personally, I am a bit surprised at the low AFC North winner percentage but, once you take Lamar Jackson’s ELO score into consideration, it does make sense that the Baltimore Ravens were the beneficiary of the model’s inner workings.
Next, let’s move onto 10,000 simulations with Haskins as the quarterback.
Read more:
https://www.steelernation.com/is-rud...hat-i-learned/
What didn't show up when I pasted this into this board (this was 100% stolen) were the graphs produced by the simulators. The essentials of those graphs, in regard to the Steelers, were:
With Ben starting at QB the Steelers are likely to win 9.4 games in 2021 and make the playoffs.
With Haskins starting at QB the Steelers are likely to win 6.2 games in 2021.
With Rudolph starting at QB the Steelers are likely to win 5.2 games in 2021.
Just a simulator average, but interesting.
Re: 10K Simulations (see if this cut and paste job works)
Interesting for sure, but I think most people could come to that conclusion without doing 10,000 simulations. I really hope Haskins can turn it around and be the guy, but it's a longshot, IMO.
Re: 10K Simulations (see if this cut and paste job works)
I agree that the casual observer or 'fan' can come to a similar conclusion without the use of simulators. To me, the simulators validated what I've been thinking about the Steelers' situation. And, I assume, this sort of analysis might be helpful when an organization is trying to decide how many millions to spend on a player.
Re: 10K Simulations (see if this cut and paste job works)
Quote:
Originally Posted by
ThorndikeFFA
With Ben Roethlisberger’s future continuing to be up in the air, it is fair that many are looking at the other quarterbacks on the roster, those entering free agency, to trade situations, and even to the collegiate ranks in the chance that the Pittsburgh Steelers trade up to snag the quarterback of the future.
That said, as it currently stands, the Steelers will have Dwayne Haskins and Mason Rudolph on the quarterback depth chart if Roethlisberger were to announce his departure.
Steelers faithful generally knows what to expect with Rudolph under center. Haskins is a bigger question mark.
Regardless, entertaining the thought of how each would hypothetically do as a season-long starter for the Steelers could be an interesting exercise.
Thankfully, because of the incredible work of Lee Sharpe (@LeeSharpeNFL) and Sebastian Carl (@mrcaseb), we advanced analytics lovers now have access to nflseedR and the power to explore such hypothetical situations.
In short, nflseedR is a package based in the R-programming language that allows those interested to run thousands – or tens-of-thousands – of simulations across various NFL seasons. Moreover, it allows for the development of unique models for input into the simulation.
In this case, I wanted to explore how a 2020 Steelers’ season would end up with both Rudolph and Haskins under center instead of Roethlisberger.
To do so, you first have to determine how to calculate Roethlisberger’s impact on the team versus the other two. For this simulation, I opted to go with 538’s quarterback-adjusted ELO score. The metric is designed to “account for changes in performance – and personnel – at quarterback, the game’s most important position.”
However, inputting the season-beginning ELO metric for each quarterback is just one piece of the puzzle. In this case, we also want the data to update through each week of the simulation. Using a model pre-developed by the creators of nflseedR, we are able to pass the dynamically-changed ELO metric from week to week to allow for an even more definitive final result.
With that sorted, the process is as simple as swapping out Roethlisberger’s ELO metric for both Rudolph’s and Haskins’.
After that, nflseedR does all the heavy lifting (and heavy math) to determine game outcomes based on ELO, with a quarterback’s ELO evolving and changing throughout the simulation.
Before diving into hypothetical scenarios, let’s establish a baseline by simulating the 2020 NFL season 10,000 times with Roethlisberger as the quarterback.
Ben Roethlisberger: 10,000 Simulations
https://apis.mail.yahoo.com/ws/v3/ma...ils=true&pid=2
Over 10,000 simulations of the 2020 NFL season, the Steelers – with Roethlisberger as the starting quarterback – averaged roughly 9 wins per season. While that is three-games less than the 12 victories this past season, I believe it is a rather accurate depiction as there were certainly a few games that Pittsburgh was extremely lucky to win.
Moreover, with Roethlisberger as the quarterback, the Steelers made the playoffs in 61% of the simulations, won the AFC North in just under 30% of the simulations, and won the conference and Super Bowl 9% and 4% of the time, respectively.
The simulation results for Roethlisberger seem in line with what should be expected. Personally, I am a bit surprised at the low AFC North winner percentage but, once you take Lamar Jackson’s ELO score into consideration, it does make sense that the Baltimore Ravens were the beneficiary of the model’s inner workings.
Next, let’s move onto 10,000 simulations with Haskins as the quarterback.
Read more:
https://www.steelernation.com/is-rud...hat-i-learned/
What didn't show up when I pasted this into this board (this was 100% stolen) were the graphs produced by the simulators. The essentials of those graphs, in regard to the Steelers, were:
With Ben starting at QB the Steelers are likely to win 9.4 games in 2021 and make the playoffs.
With Haskins starting at QB the Steelers are likely to win 6.2 games in 2021.
With Rudolph starting at QB the Steelers are likely to win 5.2 games in 2021.
Just a simulator average, but interesting.
the 1 fatal flaw in all this is the sims are basing future performances on past results , now I am not in any way saying we are better off with Haskins at the helm as that has yet to be determined .. what I am saying is a simulation with haskins in the role as the starter here using his stats as a baseline when he was with a totally different team and scheme and skillset players could not even be considered by the most gullible as a reliable set of data or even a likely turn of events ...
now this same test with Mason Rudolph it would be far more believable outcome but even then with such a small sample size of playing time and being such a young player learning how to play at this level its likely very flawed baseline data ...
just my opinion
Re: 10K Simulations (see if this cut and paste job works)
I never meant to suggest that this was comprehensive or perfect in any way. I saw it, thought it to be interesting analysis and shared it. That's all.
Re: 10K Simulations (see if this cut and paste job works)
Quote:
Originally Posted by
ThorndikeFFA
I never meant to suggest that this was comprehensive or perfect in any way. I saw it, thought it to be interesting analysis and shared it. That's all.
I found it to be very interesting and thought provoking.
thanks for sharing! :drink:
Re: 10K Simulations (see if this cut and paste job works)
Interesting Out gap between QB starter and QB back up has to be one of the bigger one in the NFL. That's what Ben is back.
Re: 10K Simulations (see if this cut and paste job works)
Quote:
Originally Posted by
Six Rings
Interesting Out gap between QB starter and QB back up has to be one of the bigger one in the NFL. That's what Ben is back.
Probably not.
Re: 10K Simulations (see if this cut and paste job works)
Quote:
Originally Posted by
86WARD
Probably not.
Right? The NFL can barely cobble together competent starting QBs. Where are all these amazing back-ups and third stringers?
It was just a minute ago that the Steelers performed better than anyone thought they would with their back-up QB.
