Out of the box, SB winners analysis is a waste of your time...

[fansinceGWilson]

This is my whole point, you dont even need to do any research to prove your point... GREAT OVERALL TEAMS dont even make it to the SB frequently.

I agree, this is why Im trying to steer people into analysing what wins football games instead of who wins seasons. Its simply too difficult to filter out the noise in the latter.

If you identify what wins football games, you gain access to much more data than simply a couple of teams per year over 20 years. You'll have 500+ data points per year to work with and be able to draw much more precise conclusions from it.

What correlates to winning%? What correlates to scoring points?, What correlates to stopping your opponents from scoring points? Whats the relationship between scoring alot of points and defensive play? Does scoring alot of points help your defense?

You can anwser all those questions with more than 500+ experiments per season, if you'd use those for your 20 seasons window, you'd have access to 10 000+ data points vs. the 40 you currently use... Which one do you think is more likely to be accurate/relevant/helpful?

I'm a little confused on "500+ data points." Other than that, we agree quite a bit. Even then, I think the intuitive data mirrors the exhaustive analysis . . . to a point. That is, there is no data that can or will correlate with scoring points or stopping opponents. NFL teams have millions of data points and have top stats guys and they haven't figured it out yet. IMO, the reason for that is, within reason, there are a number of ways to win consistently and fans know this without any statistical evidence. Top QB. Top D. Top HC. Top OL. Top Run game. All those can win. Even better, as we know intuitively, is a combination of factors. we've already discussed that.

A little OT, but one of my obsessions. One factor common in winning teams is the ABSENCE of bottom tier players among initial starters. Clearly, injuries affect that, but top 5 teams don't have a 63rd best OT. They have, on average, 2-4 All Pros, a few top 10 guys, and a LOT of average players. But, that's just one of too many factors.

Nonetheless, you have taken on a difficult task. I suspect there is little tight correlation in anything related to football that can be reduced to numbers since the numbers themselves are almost universally open to debate. To me, it's reasonable to say top 10 QBs are more likely to win than middle third QB, but digging to deep into the data will likely cloud that generalization. I'D be VERY interested in any data you can find on specifics of winning teams.

 

[NBP81]

I'm a little confused on "500+ data points

32 teams play 16 games in the regular season. 512 data points. You could absolutly only use 32 datapoints, as in each teams aggregated stats and then study those. ie. how does Teams average YPA relate to scoring points vs. how Y/C relate to scoring points. But this still includes alot of noise as it is not taking into account for a number of things like injuries, blowouts... etc.

You could also use all 512 games as your data points. Then you'd study points scored relative to YPA and do the same for Y/C in each single game and compare the results. This is much better than the previous analysis as outliers are much less likely to affect the results significantly since every data point is only a small part of the sample size.

If you'd like filter out even more noise, you could actuall remove those outliers. ie. Remove blowouts all together from your sample.

If you go to play-by-play data, you can filter for game situation to get even more precise results. The more data points you have, the easier it is to remove noise.

there is no data that can or will correlate with scoring points or stopping opponents. NFL teams have millions of data points and have top stats guys and they haven't figured it out yet.

This is not true at all, YPA is directly correlated to scoring points, more than any other factor in football. There are hundreds of study confirming this. Also, scoring points correlates to win% more than point allowed. Its all there.

One factor common in winning teams is the ABSENCE of bottom tier players among initial starters.

This is obvious, teams with better players tend to win more games on average.

If I go to you and tell you: "Hey friend, I'd really like to be financially independant someday, do you have any advice for me?" and your response is: "Well I did a little bit of research on this and turns out that the top 40 most wealthiest man in the world have alot of money". While true, how does that help me accomplish my goal?

I'd much rather you give me a list of the 500 most wealthiest people on the planet with a spreadsheet of all the individual financial moves they made throughout their career... I'd be much more likely to be able to ignore those who just inherited their wealth and focus on those who started with nothing, evaluate the effeciency of different types of moves on a much larger scale and so on.

When your giving me a list of 40 SB teams and tell me they all had great overall teams, how does that help me build a great team?

 

[NBP81]

NFL teams have millions of data points and have top stats guys and they haven't figured it out yet

What makes you say that? YPA correlates to winning more than any other stats in football and what do you know, QB and LT are by far the most valuable positions in pro Football... Coincidence?

 

[fansinceGWilson]

32 teams play 16 games in the regular season. 512 data points. You could absolutly only use 32 datapoints, as in each teams aggregated stats and then study those. ie. how does Teams average YPA relate to scoring points vs. how Y/C relate to scoring points. But this still includes alot of noise as it is not taking into account for a number of things like injuries, blowouts... etc.

You could also use all 512 games as your data points. Then you'd study points scored relative to YPA and do the same for Y/C in each single game and compare the results. This is much better than the previous analysis as outliers are much less likely to affect the results significantly since every data point is only a small part of the sample size.

If you'd like filter out even more noise, you could actuall remove those outliers. ie. Remove blowouts all together from your sample.

If you go to play-by-play data, you can filter for game situation to get even more precise results. The more data points you have, the easier it is to remove noise.



This is not true at all, YPA is directly correlated to scoring points, more than any other factor in football. There are hundreds of study confirming this. Also, scoring points correlates to win% more than point allowed. Its all there.

This is obvious, teams with better players tend to win more games on average.

If I go to you and tell you: "Hey friend, I'd really like to be financially independant someday, do you have any advice for me?" and your response is: "Well I did a little bit of research on this and turns out that the top 40 most wealthiest man in the world have alot of money". While true, how does that help me accomplish my goal?

I'd much rather you give me a list of the 500 most wealthiest people on the planet with a spreadsheet of all the individual financial moves they made throughout their career... I'd be much more likely to be able to ignore those who just inherited their wealth and focus on those who started with nothing, evaluate the effeciency of different types of moves on a much larger scale and so on.

When your giving me a list of 40 SB teams and tell me they all had great overall teams, how does that help me build a great team?

Yeah, I've seen some of those studies. You are correct - "YPA is highly correlated to scoring points, more than any other factor in football." Apologies for changing "direct" to "highly."

And I agree, the more data points, particularly over several scenarios, would be more valuable. I make no attempt to disagree. That's stats 101. All I can do is repeat 'the universe I was given is QBs in the SB.' Not the best 2 teams. Not the top 5. I've divided that into two parts. The easy one - how often does an individual 'elite' QB get to the SB. After Brady, PManning is the most recognizable elite QB and he made it 3 of 17 times, and one of those, with little argument, was behind Denver's D. A number of elite QBs made it once. The second part is a little more difficult - if elite QBs get to the SB infrequently, what other factors are just as common? Actually, quite a few. I've listed 3, solely as indications QBs aren't a unique factor. No effort to build a winning team. No effort to claim every SB contains the top 2 teams. No effort to say a top D or top OL is any more necessary than a top QB. No effort to say other factors don't apply. That's all there is - QBs in the SB.

I think our goals are causing some confusion here. I get it . . . you're trying to build a winning team. You're goal is much more difficult than mine. In this case, I care little about building a winning team. I have only one goal - disprove a QB, by himself, can get to SBs frequently.

It's one thing to have, as a goal, "focus on those who started with nothing, evaluate the effeciency of different types of moves on a much larger scale and so on." If I tell you I want a study doing nothing but provide evidence of the following - of the 20 wealthy people, how many inherited their money? That goal is MUCH easier. That's where I am. The universe is smaller and the number of data points needed is smaller.

Now, to your goal. By definition, you need all the data points you can get. All the variables and scenarios you can include adds validity to the conclusion. Combining that data with the human side - (quality of HC, OC, DC) is unimaginably difficult. My background is pattern analysis and some multivariate work, very little involving the messy human stuff. I don't envy your task.

 

[Kamelion4291]

What makes you say that? YPA correlates to winning more than any other stats in football and what do you know, QB and LT are by far the most valuable positions in pro Football... Coincidence?

I disagree on the LT point. Used to be that way, but defenses these days just attack your line everywhere and they're too good to hold the ball long enough and ask your LT to just hold up play after play.

 

[NBP81]

I disagree on the LT point. Used to be that way, but defenses these days just attack your line everywhere and they're too good to hold the ball long enough and ask your LT to just hold up play after play.

I should've specified, I was refering to salary.

 

[j-off-her-doll]

Excellent thread!