"X elite QB has only won 1 SB in the last X years" , "X teams who won a SB in the last X years had a top 5 {whatever}"
If you find yourself using those types of arguments regularly, this post is for you. Its a boring one but I think it might save you a shitload amount of time barking up the wrong tree. While there's no doubt SB winners will almost always be strong overall teams, reality is they are not a formula to try and replicate on their own.
Whether this year's winner is the Running game + Defense juggernaut or the strong passing game, the winner of this game shouldnt mean anything to your strategy for building an NFL team.
There's simply to much noise as to who wins a Super Bowl in any given year that limiting your data to SB winners will just lead you to an inevitable rudderless conclusion. Not only will you gain no real insights into what goes into building a great football team, you will do so on such a tiny amount of data that it'll be meaningless.
Lets ignore the game related luck for a second and just use common sense. 32 teams fight for a Championship, in a perfect parity environment, each team would have a tad above 3% shot at winning a SB every year. Of course, this isnt realistic so I just took the pre season future odds to win the super bowl to get somewhat realistic numbers for the sake of this argument.
Dont go all nuts here, those numbers only serve as an example and a couple ticks up or down wont change the argument Im trying to make.
The experiment goes like this, assume you've got this great team who's odds to win the SB year in and year out are around 16%. No down season, no injuries, you've just flat out figured out the league and you're always at the top for 10 years straight. How many SBs do you expect to win during that window?
Lets flip the coin a 1000 times and see what we get...
You need more data
Lets say instead you decide to go after points for and points against as the core of your analysis. If you build a great team thats a 7 point favorite on average(which pretty much amounts to a team thats got 16% odds to win the SB), you'll be more likely to get results that matter. 7 points favorites win around 75% of their games. So again running with common sense probability, if we run a 16 games season with those settings we get this:
The real takeaway here is. The amount of data that goes into your analysis matters. SB winners give you 1 data point per year and most of the time that data point just got lucky. Notice how 60% of your points differential analysis yields 11 wins or more and close to 90% go to the playoffs. And instead of having 1 data point per year you have over 500 data points to play with.
Reality is this is just the tip, play by play is where you find the real gems but you can still get alot from game stats.
If you find yourself using those types of arguments regularly, this post is for you. Its a boring one but I think it might save you a shitload amount of time barking up the wrong tree. While there's no doubt SB winners will almost always be strong overall teams, reality is they are not a formula to try and replicate on their own.
Whether this year's winner is the Running game + Defense juggernaut or the strong passing game, the winner of this game shouldnt mean anything to your strategy for building an NFL team.
There's simply to much noise as to who wins a Super Bowl in any given year that limiting your data to SB winners will just lead you to an inevitable rudderless conclusion. Not only will you gain no real insights into what goes into building a great football team, you will do so on such a tiny amount of data that it'll be meaningless.
Lets ignore the game related luck for a second and just use common sense. 32 teams fight for a Championship, in a perfect parity environment, each team would have a tad above 3% shot at winning a SB every year. Of course, this isnt realistic so I just took the pre season future odds to win the super bowl to get somewhat realistic numbers for the sake of this argument.
Dont go all nuts here, those numbers only serve as an example and a couple ticks up or down wont change the argument Im trying to make.
The experiment goes like this, assume you've got this great team who's odds to win the SB year in and year out are around 16%. No down season, no injuries, you've just flat out figured out the league and you're always at the top for 10 years straight. How many SBs do you expect to win during that window?
Lets flip the coin a 1000 times and see what we get...
- 33% of the time you win 1 SB during that 10 year window
- 27.9% you win 2
- 17.7% you win 0... *zero*
- 14% you win 3
- 5.8% you win 4
- 1% you win 5
- 0.1% you win 6
You need more data
Lets say instead you decide to go after points for and points against as the core of your analysis. If you build a great team thats a 7 point favorite on average(which pretty much amounts to a team thats got 16% odds to win the SB), you'll be more likely to get results that matter. 7 points favorites win around 75% of their games. So again running with common sense probability, if we run a 16 games season with those settings we get this:
- 21% of the time you win 12 games
- 21% 13 games
- 17% 11 games
- 15% 14 games
- 12% 10 games
The real takeaway here is. The amount of data that goes into your analysis matters. SB winners give you 1 data point per year and most of the time that data point just got lucky. Notice how 60% of your points differential analysis yields 11 wins or more and close to 90% go to the playoffs. And instead of having 1 data point per year you have over 500 data points to play with.
Reality is this is just the tip, play by play is where you find the real gems but you can still get alot from game stats.
Last edited: