Regression to the mean (1 Viewer)

[kutta]

A short, open letter to the FBG community:

I'm really tired of hearing the phrase "regression to the mean." Just because someone has a great year doesn't mean the next year is going to be worse. There are tons of factors that go into projections, and just saying you think someone is going to do worse because of so called "regression" is weak. Can we try to avoid using this over-used phrase and start basing our projections on real stuff?

Thank you.

 

[ScottyDog]

I am guessing that the next thread you start will be better than this one due to regression to the mean.

 

[butcher boy]

I remember this was debated last year. My position is that while "regression to the mean" by itself can't be the only thing that determines a reduction in numbers, the fact is that a "law of averages" so to speak does exist. Injuries, weather, and all sorts of stuff that can't be predicted can affect this.

This year, I hear the term used a lot in describing Cam Newton. In his case i think it's valid, most likely because defenses will adjust to his play, and I doubt the coaches really want him running wild again that much. So, in a sense I do expect his numbers to regress to the mean a bit. I'll admit though that I don't really know what Cam's mean is since he's only had one year of stats.

 

[kutta]

I remember this was debated last year. My position is that while "regression to the mean" by itself can't be the only thing that determines a reduction in numbers, the fact is that a "law of averages" so to speak does exist. Injuries, weather, and all sorts of stuff that can't be predicted can affect this.This year, I hear the term used a lot in describing Cam Newton. In his case i think it's valid, most likely because defenses will adjust to his play, and I doubt the coaches really want him running wild again that much. So, in a sense I do expect his numbers to regress to the mean a bit. I'll admit though that I don't really know what Cam's mean is since he's only had one year of stats.

You hear it about Cam, Calvin, Gronk, Graham, and even Brees, Brady, and Rogers. You hear it about the passing numbers overall in the league. But you don't hear it much about the RB's. Instead of addressing why we are seeing an uptick in passing stats, we keeping about regression to the mean. Well, rules have changed and the game has changed. The mean is longer the mean.

 

[Kool-Aid Larry]

good luck with that, dude.

fantasy football, and the entire internet for that matter, is filled with a bunch of c average nerds who want to feel smarter than they are.

the funniest thing about that whole abuse is that 95% of the people using the phrase don't even know what a mean means.

 

[Run It Up]

I remember this was debated last year. My position is that while "regression to the mean" by itself can't be the only thing that determines a reduction in numbers, the fact is that a "law of averages" so to speak does exist. Injuries, weather, and all sorts of stuff that can't be predicted can affect this.This year, I hear the term used a lot in describing Cam Newton. In his case i think it's valid, most likely because defenses will adjust to his play, and I doubt the coaches really want him running wild again that much. So, in a sense I do expect his numbers to regress to the mean a bit. I'll admit though that I don't really know what Cam's mean is since he's only had one year of stats.

You hear it about Cam, Calvin, Gronk, Graham, and even Brees, Brady, and Rogers. You hear it about the passing numbers overall in the league. But you don't hear it much about the RB's. Instead of addressing why we are seeing an uptick in passing stats, we keeping about regression to the mean. Well, rules have changed and the game has changed. The mean is longer the mean.

Im not sure what your are worried about. As far as reasoning goes for projections, reflection on the law of averages may not be the smartest or most bold stance, but it will be the most correct more often then not.Its obviously a very general stance, but I dont see a problem with it at all. I have bold opinions and observations that I sometimes choose not to share because the discussion isnt warranted, but when people approach every player highlight or discussion by saying Player X did achievement Y last year the only reasonable response is that its unlikely to be repeated.Gronk for example, I think its incredibly unlikely he matches or outproduces last years production, but I also think hes incredibly likely to outproduce his rookie production, which puts him still in the top 3 TEs (and regressing to the mean).

 

[Kool-Aid Larry]

I remember this was debated last year. My position is that while "regression to the mean" by itself can't be the only thing that determines a reduction in numbers, the fact is that a "law of averages" so to speak does exist. Injuries, weather, and all sorts of stuff that can't be predicted can affect this.

This year, I hear the term used a lot in describing Cam Newton. In his case i think it's valid, most likely because defenses will adjust to his play, and I doubt the coaches really want him running wild again that much. So, in a sense I do expect his numbers to regress to the mean a bit. I'll admit though that I don't really know what Cam's mean is since he's only had one year of stats.

You hear it about Cam, Calvin, Gronk, Graham, and even Brees, Brady, and Rogers. You hear it about the passing numbers overall in the league. But you don't hear it much about the RB's. Instead of addressing why we are seeing an uptick in passing stats, we keeping about regression to the mean. Well, rules have changed and the game has changed. The mean is longer the mean.

Im not sure what your are worried about. As far as reasoning goes for projections, reflection on the law of averages may not be the smartest or most bold stance, but it will be the most correct more often then not.Its obviously a very general stance, but I dont see a problem with it at all. I have bold opinions and observations that I sometimes choose not to share because the discussion isnt warranted, but when people approach every player highlight or discussion by saying Player X did achievement Y last year the only reasonable response is that its unlikely to be repeated.

Gronk for example, I think its incredibly unlikely he matches or outproduces last years production, but I also think hes incredibly likely to outproduce his rookie production, which puts him still in the top 3 TEs (and regressing to the mean).

what's the mean?

 

[Chazzhawk]

It is a very valid statement because about 95 percent of all players do. There are exceptions but for every Brees, Rodgers, Brady. There are Derek Anderson, Javon Walker, Drew Bennett, Ron Dayne, Scott Mitchell, Daunte Culpepper, Kevin Kolb, and the list goes on and on. It is hard to argue when the numbers back it up in every sport. The talent is how you tell where they are going to end up.

 

[Chazzhawk]

I remember this was debated last year. My position is that while "regression to the mean" by itself can't be the only thing that determines a reduction in numbers, the fact is that a "law of averages" so to speak does exist. Injuries, weather, and all sorts of stuff that can't be predicted can affect this.

This year, I hear the term used a lot in describing Cam Newton. In his case i think it's valid, most likely because defenses will adjust to his play, and I doubt the coaches really want him running wild again that much. So, in a sense I do expect his numbers to regress to the mean a bit. I'll admit though that I don't really know what Cam's mean is since he's only had one year of stats.

You hear it about Cam, Calvin, Gronk, Graham, and even Brees, Brady, and Rogers. You hear it about the passing numbers overall in the league. But you don't hear it much about the RB's. Instead of addressing why we are seeing an uptick in passing stats, we keeping about regression to the mean. Well, rules have changed and the game has changed. The mean is longer the mean.

Im not sure what your are worried about. As far as reasoning goes for projections, reflection on the law of averages may not be the smartest or most bold stance, but it will be the most correct more often then not.Its obviously a very general stance, but I dont see a problem with it at all. I have bold opinions and observations that I sometimes choose not to share because the discussion isnt warranted, but when people approach every player highlight or discussion by saying Player X did achievement Y last year the only reasonable response is that its unlikely to be repeated.

Gronk for example, I think its incredibly unlikely he matches or outproduces last years production, but I also think hes incredibly likely to outproduce his rookie production, which puts him still in the top 3 TEs (and regressing to the mean).

what's the mean?

If you believe he is a top 3 tight end then the mean is about 70 to 90 catches for 1000 to 1200 yards and 8 to 12 tds.

 

[Kool-Aid Larry]

I remember this was debated last year. My position is that while "regression to the mean" by itself can't be the only thing that determines a reduction in numbers, the fact is that a "law of averages" so to speak does exist. Injuries, weather, and all sorts of stuff that can't be predicted can affect this.

This year, I hear the term used a lot in describing Cam Newton. In his case i think it's valid, most likely because defenses will adjust to his play, and I doubt the coaches really want him running wild again that much. So, in a sense I do expect his numbers to regress to the mean a bit. I'll admit though that I don't really know what Cam's mean is since he's only had one year of stats.

You hear it about Cam, Calvin, Gronk, Graham, and even Brees, Brady, and Rogers. You hear it about the passing numbers overall in the league. But you don't hear it much about the RB's. Instead of addressing why we are seeing an uptick in passing stats, we keeping about regression to the mean. Well, rules have changed and the game has changed. The mean is longer the mean.

Im not sure what your are worried about. As far as reasoning goes for projections, reflection on the law of averages may not be the smartest or most bold stance, but it will be the most correct more often then not.Its obviously a very general stance, but I dont see a problem with it at all. I have bold opinions and observations that I sometimes choose not to share because the discussion isnt warranted, but when people approach every player highlight or discussion by saying Player X did achievement Y last year the only reasonable response is that its unlikely to be repeated.

Gronk for example, I think its incredibly unlikely he matches or outproduces last years production, but I also think hes incredibly likely to outproduce his rookie production, which puts him still in the top 3 TEs (and regressing to the mean).

what's the mean?

If you believe he is a top 3 tight end then the mean is about 70 to 90 catches for 1000 to 1200 yards and 8 to 12 tds.

what if you believe he is the #1 te?but that would also mean he's a top ten te, so what's the mean on that one?

 

[Synesthesia]

It's a valid concept, but people need to be clear just whose mean we are referring to. Regression to the mean doesn't mean regression to the league average. Jimmy Graham and Tony Scheffler are not regressing to the same mean.

To illustrate using a baseball analogy: let's say you have a .300 hitter. Let's say over 10 at bats, he goes 8-for-10. How many hits should we expect over the next 10? Right around 3. What if he goes 0-for-10 over those 10, what should we expect over the 10 after that? Right around 3 hits. If he starts out the first two weeks of the season hitting .500, where should we expect him to finish the season? Right around .300. Why? Because the guy's mean is .300, and he'll tend to regress to it. A .200 hitter will regress to something completely different.

Take Calvin Johnson last season. Over the first 4 weeks, Calvin scored 2 TDs every week. Everyone started speculating about how many TDs Calvin would manage to put up, taking it as foregone conclusion that he was going to set the TD record. And then a funny thing happened on the way to the record books- over the last 12 weeks of the season, Calvin's TD-per-game rate was essentially identical to his TD-per-game rate from 2008-2010. His hot start meant nothing- he regressed to pretty much his exact career averages. Regression to the mean at work!

Players can improve, and their "true mean" can change, but the concept of regression to the mean really only means that any player can (and will) play over (or under) their head over a short enough timeline, but as you extend the timeline they'll always return to their "true" performance level. It's up to us to try to determine whether, for instance, Cam Newton's 14 rushing scores, or Aaron Rogers' 45 passing scores represent their true mean or an example of them playing over their heads over a short timeline.

 

[Xue]

See: Dwayne Bowe 2010, Peyton Manning 2004, Tom Brady 2007, Ben Roethlisberger 2007, Steve Smith 2005.

Just 4 examples off the top of my head. Usually "regression to the mean" is applied to TD rates.

 

[Raider Nation]

I think regression to the mean is far more unlikely for the high motor guys. You know... the ones who give 110%.

 

[bonesman]

I agree with Kutta. The FBG community should really be progressing towards the nice. All this regressing and meanness is unbecoming. Be Excellent to each other.

 

[Kool-Aid Larry]

It's a valid concept, but people need to be clear just whose mean we are referring to. Regression to the mean doesn't mean regression to the league average. Jimmy Graham and Tony Scheffler are not regressing to the same mean. To illustrate using a baseball analogy: let's say you have a .300 hitter. Let's say over 10 at bats, he goes 8-for-10. How many hits should we expect over the next 10? Right around 3. What if he goes 0-for-10 over those 10, what should we expect over the 10 after that? Right around 3 hits. If he starts out the first two weeks of the season hitting .500, where should we expect him to finish the season? Right around .300. Why? Because the guy's mean is .300, and he'll tend to regress to it. A .200 hitter will regress to something completely different. Take Calvin Johnson last season. Over the first 4 weeks, Calvin scored 2 TDs every week. Everyone started speculating about how many TDs Calvin would manage to put up, taking it as foregone conclusion that he was going to set the TD record. And then a funny thing happened on the way to the record books- over the last 12 weeks of the season, Calvin's TD-per-game rate was essentially identical to his TD-per-game rate from 2008-2010. His hot start meant nothing- he regressed to pretty much his exact career averages. Regression to the mean at work!Players can improve, and their "true mean" can change, but the concept of regression to the mean really only means that any player can (and will) play over (or under) their head over a short enough timeline, but as you extend the timeline they'll always return to their "true" performance level. It's up to us to try to determine whether, for instance, Cam Newton's 14 rushing scores, or Aaron Rogers' 45 passing scores represent their true mean or an example of them playing over their heads over a short timeline.

yeah, why is it nobody ever talks about 'progression to the mean' for some guy like jamarcus russell?

 

[Chase Stuart]

It's a valid concept, but people need to be clear just whose mean we are referring to. Regression to the mean doesn't mean regression to the league average. Jimmy Graham and Tony Scheffler are not regressing to the same mean. To illustrate using a baseball analogy: let's say you have a .300 hitter. Let's say over 10 at bats, he goes 8-for-10. How many hits should we expect over the next 10? Right around 3. What if he goes 0-for-10 over those 10, what should we expect over the 10 after that? Right around 3 hits. If he starts out the first two weeks of the season hitting .500, where should we expect him to finish the season? Right around .300. Why? Because the guy's mean is .300, and he'll tend to regress to it. A .200 hitter will regress to something completely different. Take Calvin Johnson last season. Over the first 4 weeks, Calvin scored 2 TDs every week. Everyone started speculating about how many TDs Calvin would manage to put up, taking it as foregone conclusion that he was going to set the TD record. And then a funny thing happened on the way to the record books- over the last 12 weeks of the season, Calvin's TD-per-game rate was essentially identical to his TD-per-game rate from 2008-2010. His hot start meant nothing- he regressed to pretty much his exact career averages. Regression to the mean at work!Players can improve, and their "true mean" can change, but the concept of regression to the mean really only means that any player can (and will) play over (or under) their head over a short enough timeline, but as you extend the timeline they'll always return to their "true" performance level. It's up to us to try to determine whether, for instance, Cam Newton's 14 rushing scores, or Aaron Rogers' 45 passing scores represent their true mean or an example of them playing over their heads over a short timeline.

 

[Chase Stuart]

It's a valid concept, but people need to be clear just whose mean we are referring to. Regression to the mean doesn't mean regression to the league average. Jimmy Graham and Tony Scheffler are not regressing to the same mean. To illustrate using a baseball analogy: let's say you have a .300 hitter. Let's say over 10 at bats, he goes 8-for-10. How many hits should we expect over the next 10? Right around 3. What if he goes 0-for-10 over those 10, what should we expect over the 10 after that? Right around 3 hits. If he starts out the first two weeks of the season hitting .500, where should we expect him to finish the season? Right around .300. Why? Because the guy's mean is .300, and he'll tend to regress to it. A .200 hitter will regress to something completely different. Take Calvin Johnson last season. Over the first 4 weeks, Calvin scored 2 TDs every week. Everyone started speculating about how many TDs Calvin would manage to put up, taking it as foregone conclusion that he was going to set the TD record. And then a funny thing happened on the way to the record books- over the last 12 weeks of the season, Calvin's TD-per-game rate was essentially identical to his TD-per-game rate from 2008-2010. His hot start meant nothing- he regressed to pretty much his exact career averages. Regression to the mean at work!Players can improve, and their "true mean" can change, but the concept of regression to the mean really only means that any player can (and will) play over (or under) their head over a short enough timeline, but as you extend the timeline they'll always return to their "true" performance level. It's up to us to try to determine whether, for instance, Cam Newton's 14 rushing scores, or Aaron Rogers' 45 passing scores represent their true mean or an example of them playing over their heads over a short timeline.

yeah, why is it nobody ever talks about 'progression to the mean' for some guy like jamarcus russell?

Do you think Chris Johnson will be better this year?

 

[Chazzhawk]

I remember this was debated last year. My position is that while "regression to the mean" by itself can't be the only thing that determines a reduction in numbers, the fact is that a "law of averages" so to speak does exist. Injuries, weather, and all sorts of stuff that can't be predicted can affect this.

This year, I hear the term used a lot in describing Cam Newton. In his case i think it's valid, most likely because defenses will adjust to his play, and I doubt the coaches really want him running wild again that much. So, in a sense I do expect his numbers to regress to the mean a bit. I'll admit though that I don't really know what Cam's mean is since he's only had one year of stats.

You hear it about Cam, Calvin, Gronk, Graham, and even Brees, Brady, and Rogers. You hear it about the passing numbers overall in the league. But you don't hear it much about the RB's. Instead of addressing why we are seeing an uptick in passing stats, we keeping about regression to the mean. Well, rules have changed and the game has changed. The mean is longer the mean.

Im not sure what your are worried about. As far as reasoning goes for projections, reflection on the law of averages may not be the smartest or most bold stance, but it will be the most correct more often then not.Its obviously a very general stance, but I dont see a problem with it at all. I have bold opinions and observations that I sometimes choose not to share because the discussion isnt warranted, but when people approach every player highlight or discussion by saying Player X did achievement Y last year the only reasonable response is that its unlikely to be repeated.

Gronk for example, I think its incredibly unlikely he matches or outproduces last years production, but I also think hes incredibly likely to outproduce his rookie production, which puts him still in the top 3 TEs (and regressing to the mean).

what's the mean?

If you believe he is a top 3 tight end then the mean is about 70 to 90 catches for 1000 to 1200 yards and 8 to 12 tds.

what if you believe he is the #1 te?but that would also mean he's a top ten te, so what's the mean on that one?

Now you are just arguing to argue

 

[Chazzhawk]

It's a valid concept, but people need to be clear just whose mean we are referring to. Regression to the mean doesn't mean regression to the league average. Jimmy Graham and Tony Scheffler are not regressing to the same mean. To illustrate using a baseball analogy: let's say you have a .300 hitter. Let's say over 10 at bats, he goes 8-for-10. How many hits should we expect over the next 10? Right around 3. What if he goes 0-for-10 over those 10, what should we expect over the 10 after that? Right around 3 hits. If he starts out the first two weeks of the season hitting .500, where should we expect him to finish the season? Right around .300. Why? Because the guy's mean is .300, and he'll tend to regress to it. A .200 hitter will regress to something completely different. Take Calvin Johnson last season. Over the first 4 weeks, Calvin scored 2 TDs every week. Everyone started speculating about how many TDs Calvin would manage to put up, taking it as foregone conclusion that he was going to set the TD record. And then a funny thing happened on the way to the record books- over the last 12 weeks of the season, Calvin's TD-per-game rate was essentially identical to his TD-per-game rate from 2008-2010. His hot start meant nothing- he regressed to pretty much his exact career averages. Regression to the mean at work!Players can improve, and their "true mean" can change, but the concept of regression to the mean really only means that any player can (and will) play over (or under) their head over a short enough timeline, but as you extend the timeline they'll always return to their "true" performance level. It's up to us to try to determine whether, for instance, Cam Newton's 14 rushing scores, or Aaron Rogers' 45 passing scores represent their true mean or an example of them playing over their heads over a short timeline.

 

[Chase Stuart]

Regression to the mean is not a debatable concept, but it's one that is often misunderstood and inappropriately applied. At its core, regression to the mean is a concept borne out of the fact that there are statistical outliers in small samples that are not predictive of future results. If a roulette wheel lands on black 10 times in a row, we don't expect it to land on black in 10 of the next 10 spins, we'd expect it to regress to the mean.

If we replayed the 2011 season 1,000,000 times, do you think Gronkowski would average 17 TDs a season? I don't think so. Some years, he would get hurt. Other years, Brady would get hurt. He might be asked to block more often in some seasons, and in some years he would lose his confidence and play poorly. He'd do better than 17 TDs in a season sometimes, too. But on average, I don't think we'd expect Gronk to be a 17 TD/year guy.

I wrote a pretty thorough explanation of the concept here: http://www.pro-football-reference.com/blog/?p=370

 

[Kool-Aid Larry]

Regression to the mean is not a debatable concept, but it's one that is often misunderstood and inappropriately applied.

yeah, I find that most often people use 'regression to the mean' to dress up an opinion that they arrived at by throwing a dart at a board, but want to pass it off as insightful.there's another one from physics that people abuse all the time to sound smarter.......ohhh I can't think of it now.

 

[Kool-Aid Larry]

See: Dwayne Bowe 2010, Peyton Manning 2004, Tom Brady 2007, Ben Roethlisberger 2007, Steve Smith 2005.Just 4 examples off the top of my head. Usually "regression to the mean" is applied to TD rates.

gronkowski 2010that's just one example off the top of my head of a player not regressing to the mean.

 

[kutta]

Here's my main issue with it. Regression the mean is usually used to account for variations due to randomness or luck (see Chase's Roulette wheel example). So sure, of course I agree that there is an element of luck involved with players' stats in football. And, all things being equal, if a player completely blows up one year there is an expectation that he will come back down the next.

But the problem I have specifically with this year is that people are not looking at the reasons why some of the stats were so out of whack. I don't hear people talking about Ray Rice regressing to the mean this year, or Arian Foster, or McCoy. But they do talk about Graham, Gronk, Calvin, and the QB's regressing only because nobody has done what they did before. But they can't all be outliers. The game is changing. Pass Interference rules greatly favor the WR's. The QB's are not allowed to be hit. Teams are seeing other teams have great success throwing the ball, so they are emulating them. There is no reason to believe that the passing stats are going to go down anytime soon.

Do you really think that three (almost four, Eli was about 70 yards short) QB's throwing for 5000 yards in the same season is an anomaly? If it was one, maybe. But four? In 2011 we had four of the top six most prolific passing yardage performances in the history of the NFL. That is staggering. There were three (almost four - Brady had 39) QB's with over 40 passing TD's. Before 2010 there had only been 5 QB's in the history of the NFL with over 40 TD's. The league-wide passer rating average was 84.3 in 2011, breaking the record set in, you guessed it, 2010 when it was 84.1. I could go on and on.

The point being, the passing stats last year were consistent across many players and many teams. This is not one guy going crazy and blowing the rest of the league out of the water. It is happening all over the league. With the rule changes and more prolific offenses, these passing offenses are going to continue to put up gaudy numbers, and players are going to continue put up stats that people will continue to say must "regress to the mean." But what they don't understand is that the mean is changing and it is moving upward.

 

[TheJewru]

Avoiding career years is key in fantasy. Avoiding players likely to under perform last year's stats is important. Generally I won't end up with players coming off career years in that another owner will like that player more than me. People " Guppies" draft based on last years stats. Whether it was inordinate number of plays a receiver or back scored when a defender slipped, a ball bounced the right way, or other scenarios. Regression is just mathematical probability.

Guys I won't be drafting. Victor Cruz, Jordy Nelson, Rob Gronkowski, Marshawn Lynch, Darren Sproles, Cam Newton, Matthew Stafford. I like all of the players listed but I can see all leaving an owner disappointed based on where you will have to draft them.

 

[kutta]

Avoiding career years is key in fantasy. Avoiding players likely to under perform last year's stats is important. Generally I won't end up with players coming off career years in that another owner will like that player more than me. People " Guppies" draft based on last years stats. Whether it was inordinate number of plays a receiver or back scored when a defender slipped, a ball bounced the right way, or other scenarios. Regression is just mathematical probability. Guys I won't be drafting. Victor Cruz, Jordy Nelson, Rob Gronkowski, Marshawn Lynch, Darren Sproles, Cam Newton, Matthew Stafford. I like all of the players listed but I can see all leaving an owner disappointed based on where you will have to draft them.

The problem with this reasoning is that not only will you not draft Stafford this year, you probably didn't draft him last year. If you thought he was already playing at his mean, why draft him last year? How do we know that last year was not really his mean?Would you draft McCoy this year? Did you avoid Foster after his breakout season?

 

[Chazzhawk]

Here's my main issue with it. Regression the mean is usually used to account for variations due to randomness or luck (see Chase's Roulette wheel example). So sure, of course I agree that there is an element of luck involved with players' stats in football. And, all things being equal, if a player completely blows up one year there is an expectation that he will come back down the next.

But the problem I have specifically with this year is that people are not looking at the reasons why some of the stats were so out of whack. I don't hear people talking about Ray Rice regressing to the mean this year, or Arian Foster, or McCoy. But they do talk about Graham, Gronk, Calvin, and the QB's regressing only because nobody has done what they did before. But they can't all be outliers. The game is changing. Pass Interference rules greatly favor the WR's. The QB's are not allowed to be hit. Teams are seeing other teams have great success throwing the ball, so they are emulating them. There is no reason to believe that the passing stats are going to go down anytime soon.

Do you really think that three (almost four, Eli was about 70 yards short) QB's throwing for 5000 yards in the same season is an anomaly? If it was one, maybe. But four? In 2011 we had four of the top six most prolific passing yardage performances in the history of the NFL. That is staggering. There were three (almost four - Brady had 39) QB's with over 40 passing TD's. Before 2010 there had only been 5 QB's in the history of the NFL with over 40 TD's. The league-wide passer rating average was 84.3 in 2011, breaking the record set in, you guessed it, 2010 when it was 84.1. I could go on and on.

The point being, the passing stats last year were consistent across many players and many teams. This is not one guy going crazy and blowing the rest of the league out of the water. It is happening all over the league. With the rule changes and more prolific offenses, these passing offenses are going to continue to put up gaudy numbers, and players are going to continue put up stats that people will continue to say must "regress to the mean." But what they don't understand is that the mean is changing and it is moving upward.

This is the perfect example of regress to the mean, as much as the rules favor the offense right now, do you think defenses are just standing around saying lets hope they fumble or throw a pick other wise we can't stop them? No, they are scheming ways around the rules, and sooner or later a defense or two will find a way around the rules and everyone will copy it and magically the spike in offensive numbers will come back down. All sports have cycles like that, in the 80's it seemed like half the players NHL were scoring 100 points in a season, now you are lucky if you get 10 guys a year that do it. Baseball had the steroid era when a guy hit 35 homeruns in a year you asked yourself "That seems low, how many games did he miss?, and now we are back to pitchers dominating.

 

[kutta]

Here's my main issue with it. Regression the mean is usually used to account for variations due to randomness or luck (see Chase's Roulette wheel example). So sure, of course I agree that there is an element of luck involved with players' stats in football. And, all things being equal, if a player completely blows up one year there is an expectation that he will come back down the next.

But the problem I have specifically with this year is that people are not looking at the reasons why some of the stats were so out of whack. I don't hear people talking about Ray Rice regressing to the mean this year, or Arian Foster, or McCoy. But they do talk about Graham, Gronk, Calvin, and the QB's regressing only because nobody has done what they did before. But they can't all be outliers. The game is changing. Pass Interference rules greatly favor the WR's. The QB's are not allowed to be hit. Teams are seeing other teams have great success throwing the ball, so they are emulating them. There is no reason to believe that the passing stats are going to go down anytime soon.

Do you really think that three (almost four, Eli was about 70 yards short) QB's throwing for 5000 yards in the same season is an anomaly? If it was one, maybe. But four? In 2011 we had four of the top six most prolific passing yardage performances in the history of the NFL. That is staggering. There were three (almost four - Brady had 39) QB's with over 40 passing TD's. Before 2010 there had only been 5 QB's in the history of the NFL with over 40 TD's. The league-wide passer rating average was 84.3 in 2011, breaking the record set in, you guessed it, 2010 when it was 84.1. I could go on and on.

The point being, the passing stats last year were consistent across many players and many teams. This is not one guy going crazy and blowing the rest of the league out of the water. It is happening all over the league. With the rule changes and more prolific offenses, these passing offenses are going to continue to put up gaudy numbers, and players are going to continue put up stats that people will continue to say must "regress to the mean." But what they don't understand is that the mean is changing and it is moving upward.

This is the perfect example of regress to the mean, as much as the rules favor the offense right now, do you think defenses are just standing around saying lets hope they fumble or throw a pick other wise we can't stop them? No, they are scheming ways around the rules, and sooner or later a defense or two will find a way around the rules and everyone will copy it and magically the spike in offensive numbers will come back down. All sports have cycles like that, in the 80's it seemed like half the players NHL were scoring 100 points in a season, now you are lucky if you get 10 guys a year that do it. Baseball had the steroid era when a guy hit 35 homeruns in a year you asked yourself "That seems low, how many games did he miss?, and now we are back to pitchers dominating.

I respectfully disagree.I don't really know much about the NHL, so I can't comment. But your baseball example is flawed. The reason the home runs went up is because players were cheating and gaining an advantage. Pitchers didn't compensate - the advantage batters got from taking steroids was removed. The advantage the passing game in the NFL has now is not going to be removed, especially with the emphasis on player safety. This isn't something where offenses suddenly figured out the defenses and the defenses will adjust. These are rule changes that will affect the game forever. It would be akin to baseball saying the pitchers had to slow pitch underhand. That would cause an offensive explosion that could not be remedied by better defense.

In the NFL now, defenses are not allowed to hit QB's, they are not allowed to deck a guy going over the middle, the are not allowed to touch a WR past five yards. And these things are actually being enforced. Sure, I agree that if the NFL went back to the old ways of enforcing these rules then the passing stats would drop. But they are not going to do that. If anything, they are going to get more strict which will only help the offenses more.

As Bob Dillon says, "The times they are a changin'." There will be no "regressing to the mean" with the passing games.

 

[Chazzhawk]

Avoiding career years is key in fantasy. Avoiding players likely to under perform last year's stats is important. Generally I won't end up with players coming off career years in that another owner will like that player more than me. People " Guppies" draft based on last years stats. Whether it was inordinate number of plays a receiver or back scored when a defender slipped, a ball bounced the right way, or other scenarios. Regression is just mathematical probability. Guys I won't be drafting. Victor Cruz, Jordy Nelson, Rob Gronkowski, Marshawn Lynch, Darren Sproles, Cam Newton, Matthew Stafford. I like all of the players listed but I can see all leaving an owner disappointed based on where you will have to draft them.

That is just flawed logic and being scared. You need to look at the talent of the player, as well as how many times he got the breaks. Stafford looked very good as a first and second year qb. The problem was he was never on the field. I think his potential is real. However Jordy Nelson, and Victor Cruz had some long touchdowns and could have surprised some teams that were focusing on other weapons, they both have talent but you could see both their numbers sliding back some. Gronkowski showed flashes of this during his rookie year and put it all together his second year. Does he get 17 tds again? Probably not, but he is the real deal, plus the Patriots don't pay just anyone. I think Lynch finally grew up, I think it took getting traded for a 4th round pick to do it, his whole life he has been told you are the man, when he was in high school, in college, when he was drafted, and even his first couple years in the league. Now a team basically says ya we spent a first round pick on you, but we are happy to get your out of town for a 4th. He showed flashes against late in his first year and for sure against the Saints in the playoff game . (Remember he was in a new system) I think this year he realized it is time to show up, (I am sure trying to get a new contract helped too) but it is hard to argue against the only guy to put up 100 yards rushing against the 49ers when they faced backs like Mccoy, Rice, Wells x 2, Jackson x 2, and Mendenhall. Sproles has always had the talent to be a great 3rd down back, I think his numbers come back some, but he was also under used in San Diego.

 

[CalBear]

I remember this was debated last year. My position is that while "regression to the mean" by itself can't be the only thing that determines a reduction in numbers, the fact is that a "law of averages" so to speak does exist. Injuries, weather, and all sorts of stuff that can't be predicted can affect this.

Guess what: Everyone has injuries, weather, and all sorts of stuff that can't be predicted. It is a statistical fact that most players who have a great year will follow it up with a year that's not as good. But it's also a statistical fact that most players who don't have a great year will follow it up with a year that's not as good. Most people who talk about regression to the mean don't understand the statistics.

 

[Chazzhawk]

Here's my main issue with it. Regression the mean is usually used to account for variations due to randomness or luck (see Chase's Roulette wheel example). So sure, of course I agree that there is an element of luck involved with players' stats in football. And, all things being equal, if a player completely blows up one year there is an expectation that he will come back down the next.

But the problem I have specifically with this year is that people are not looking at the reasons why some of the stats were so out of whack. I don't hear people talking about Ray Rice regressing to the mean this year, or Arian Foster, or McCoy. But they do talk about Graham, Gronk, Calvin, and the QB's regressing only because nobody has done what they did before. But they can't all be outliers. The game is changing. Pass Interference rules greatly favor the WR's. The QB's are not allowed to be hit. Teams are seeing other teams have great success throwing the ball, so they are emulating them. There is no reason to believe that the passing stats are going to go down anytime soon.

Do you really think that three (almost four, Eli was about 70 yards short) QB's throwing for 5000 yards in the same season is an anomaly? If it was one, maybe. But four? In 2011 we had four of the top six most prolific passing yardage performances in the history of the NFL. That is staggering. There were three (almost four - Brady had 39) QB's with over 40 passing TD's. Before 2010 there had only been 5 QB's in the history of the NFL with over 40 TD's. The league-wide passer rating average was 84.3 in 2011, breaking the record set in, you guessed it, 2010 when it was 84.1. I could go on and on.

The point being, the passing stats last year were consistent across many players and many teams. This is not one guy going crazy and blowing the rest of the league out of the water. It is happening all over the league. With the rule changes and more prolific offenses, these passing offenses are going to continue to put up gaudy numbers, and players are going to continue put up stats that people will continue to say must "regress to the mean." But what they don't understand is that the mean is changing and it is moving upward.

This is the perfect example of regress to the mean, as much as the rules favor the offense right now, do you think defenses are just standing around saying lets hope they fumble or throw a pick other wise we can't stop them? No, they are scheming ways around the rules, and sooner or later a defense or two will find a way around the rules and everyone will copy it and magically the spike in offensive numbers will come back down. All sports have cycles like that, in the 80's it seemed like half the players NHL were scoring 100 points in a season, now you are lucky if you get 10 guys a year that do it. Baseball had the steroid era when a guy hit 35 homeruns in a year you asked yourself "That seems low, how many games did he miss?, and now we are back to pitchers dominating.

I respectfully disagree.I don't really know much about the NHL, so I can't comment. But your baseball example is flawed. The reason the home runs went up is because players were cheating and gaining an advantage. Pitchers didn't compensate - the advantage batters got from taking steroids was removed. The advantage the passing game in the NFL has now is not going to be removed, especially with the emphasis on player safety. This isn't something where offenses suddenly figured out the defenses and the defenses will adjust. These are rule changes that will affect the game forever. It would be akin to baseball saying the pitchers had to slow pitch underhand. That would cause an offensive explosion that could not be remedied by better defense.

In the NFL now, defenses are not allowed to hit QB's, they are not allowed to deck a guy going over the middle, the are not allowed to touch a WR past five yards. And these things are actually being enforced. Sure, I agree that if the NFL went back to the old ways of enforcing these rules then the passing stats would drop. But they are not going to do that. If anything, they are going to get more strict which will only help the offenses more.

As Bob Dillon says, "The times they are a changin'." There will be no "regressing to the mean" with the passing games.

Someone will find a way around it, they always do. The steroid era was an example, but the pitchers were cheating too. Look in the 1940's and 1950's in baseball they were scoring alot of runs back then too, were they all using PED's too?

 

[Time Kibitzer]

Here's my main issue with it. Regression the mean is usually used to account for variations due to randomness or luck (see Chase's Roulette wheel example). So sure, of course I agree that there is an element of luck involved with players' stats in football. And, all things being equal, if a player completely blows up one year there is an expectation that he will come back down the next.

But the problem I have specifically with this year is that people are not looking at the reasons why some of the stats were so out of whack. I don't hear people talking about Ray Rice regressing to the mean this year, or Arian Foster, or McCoy. But they do talk about Graham, Gronk, Calvin, and the QB's regressing only because nobody has done what they did before. But they can't all be outliers. The game is changing. Pass Interference rules greatly favor the WR's. The QB's are not allowed to be hit. Teams are seeing other teams have great success throwing the ball, so they are emulating them. There is no reason to believe that the passing stats are going to go down anytime soon.

Do you really think that three (almost four, Eli was about 70 yards short) QB's throwing for 5000 yards in the same season is an anomaly? If it was one, maybe. But four? In 2011 we had four of the top six most prolific passing yardage performances in the history of the NFL. That is staggering. There were three (almost four - Brady had 39) QB's with over 40 passing TD's. Before 2010 there had only been 5 QB's in the history of the NFL with over 40 TD's. The league-wide passer rating average was 84.3 in 2011, breaking the record set in, you guessed it, 2010 when it was 84.1. I could go on and on.

The point being, the passing stats last year were consistent across many players and many teams. This is not one guy going crazy and blowing the rest of the league out of the water. It is happening all over the league. With the rule changes and more prolific offenses, these passing offenses are going to continue to put up gaudy numbers, and players are going to continue put up stats that people will continue to say must "regress to the mean." But what they don't understand is that the mean is changing and it is moving upward.

I get what you're saying, and for the most part I agree and think it's safe to say that the mean passing total by QBs is on increase, but to ask people to stop acknowledging the regression to mean concept is a silly response that doesn't properly deal with the issue you are trying to deal with. This post of yours here above explaining how people are using the concept incorrectly is much better than your initial post where you simply ask people to ignore the concept altogether. I'd also add that for these young players like Cam Newton and Rob Gronkowski, they really haven't played enough seasons for one to approximate what exactly their mean stats are imo. Will they both be hard pressed to repeat their TD numbers from last season? Probably. But I'd still feel more comfortable reserving the regression to the mean concept for veteran type players where you have more data to work with.

 

[TheJewru]

Avoiding career years is key in fantasy. Avoiding players likely to under perform last year's stats is important. Generally I won't end up with players coming off career years in that another owner will like that player more than me. People " Guppies" draft based on last years stats. Whether it was inordinate number of plays a receiver or back scored when a defender slipped, a ball bounced the right way, or other scenarios. Regression is just mathematical probability. Guys I won't be drafting. Victor Cruz, Jordy Nelson, Rob Gronkowski, Marshawn Lynch, Darren Sproles, Cam Newton, Matthew Stafford. I like all of the players listed but I can see all leaving an owner disappointed based on where you will have to draft them.

That is just flawed logic and being scared. You need to look at the talent of the player, as well as how many times he got the breaks. Stafford looked very good as a first and second year qb. The problem was he was never on the field. I think his potential is real. However Jordy Nelson, and Victor Cruz had some long touchdowns and could have surprised some teams that were focusing on other weapons, they both have talent but you could see both their numbers sliding back some. Gronkowski showed flashes of this during his rookie year and put it all together his second year. Does he get 17 tds again? Probably not, but he is the real deal, plus the Patriots don't pay just anyone. I think Lynch finally grew up, I think it took getting traded for a 4th round pick to do it, his whole life he has been told you are the man, when he was in high school, in college, when he was drafted, and even his first couple years in the league. Now a team basically says ya we spent a first round pick on you, but we are happy to get your out of town for a 4th. He showed flashes against late in his first year and for sure against the Saints in the playoff game . (Remember he was in a new system) I think this year he realized it is time to show up, (I am sure trying to get a new contract helped too) but it is hard to argue against the only guy to put up 100 yards rushing against the 49ers when they faced backs like Mccoy, Rice, Wells x 2, Jackson x 2, and Mendenhall. Sproles has always had the talent to be a great 3rd down back, I think his numbers come back some, but he was also under used in San Diego.

How is my logic flawed and yet you seem to completely agree with it? Nelson and Cruz "Long Touchdowns", Gronkowski " Does he get 17 TD's again, probably not" Sproles " I think his numbers come back some." All I am saying is based on ADP, there will be wiser picks to be made.

 

[PIK95]

good luck with that, dude.fantasy football, and the entire internet for that matter, is filled with a bunch of c average nerds who want to feel smarter than they are.the funniest thing about that whole abuse is that 95% of the people using the phrase don't even know what a mean means.

This.

 

[habsfan]

Avoiding career years is key in fantasy. Avoiding players likely to under perform last year's stats is important. Generally I won't end up with players coming off career years in that another owner will like that player more than me. People " Guppies" draft based on last years stats. Whether it was inordinate number of plays a receiver or back scored when a defender slipped, a ball bounced the right way, or other scenarios. Regression is just mathematical probability. Guys I won't be drafting. Victor Cruz, Jordy Nelson, Rob Gronkowski, Marshawn Lynch, Darren Sproles, Cam Newton, Matthew Stafford. I like all of the players listed but I can see all leaving an owner disappointed based on where you will have to draft them.

The flipside of drafting based on last year's stats is avoiding players based on last year's stats. I don't know which behaviour is more common though. I guess it depends on your league.

 

[Kool-Aid Larry]

I get what you're saying, and for the most part I agree and think it's safe to say that the mean passing total by QBs is on increase, but to ask people to stop acknowledging the regression to mean concept is a silly response that doesn't properly deal with the issue you are trying to deal with. This post of yours here above explaining how people are using the concept incorrectly is much better than your initial post where you simply ask people to ignore the concept altogether.

I don't think he wants people to ignore it, as much as shut up about it.

 

[David Yudkin]

Regression to the mean is not a debatable concept, but it's one that is often misunderstood and inappropriately applied. At its core, regression to the mean is a concept borne out of the fact that there are statistical outliers in small samples that are not predictive of future results. If a roulette wheel lands on black 10 times in a row, we don't expect it to land on black in 10 of the next 10 spins, we'd expect it to regress to the mean.

If we replayed the 2011 season 1,000,000 times, do you think Gronkowski would average 17 TDs a season? I don't think so. Some years, he would get hurt. Other years, Brady would get hurt. He might be asked to block more often in some seasons, and in some years he would lose his confidence and play poorly. He'd do better than 17 TDs in a season sometimes, too. But on average, I don't think we'd expect Gronk to be a 17 TD/year guy.

I wrote a pretty thorough explanation of the concept here: http://www.pro-football-reference.com/blog/?p=370

As far as Gronk goes, I agree and disagree. I, too, don't think he will score 17 TD this year . . . but for different reasons. Over his career, he's played in 36 games (regular and post-season), in which he's scored 31 TD. That would make his run rate at 14 TD a season for a 16 game season. Counting all games evenly, his baseline over 16 games would be 77-976-14.I see Gronk losing some production compared to last year because of his recovery from his injured ankle (which I suggest will limit him some early in the season and could be a nagging injury throughout the year). I also see the new faces at WR taking away some of his targets. Other people have suggested that Brady won't throw the ball 600+ times again (which is a valid concern although I am not fully on board with that one).

IMO, there are potentially reasons why Gronk won't score 17 TD (which was actually 18) again this year. Sure, maybe regression has an element there, but there are other evidentiary factors that could also be valid concerns why he won't have a repeat of last year in the scoring department.

 

[Instinctive]

See: Dwayne Bowe 2010, Peyton Manning 2004, Tom Brady 2007, Ben Roethlisberger 2007, Steve Smith 2005.Just 4 examples off the top of my head. Usually "regression to the mean" is applied to TD rates.

gronkowski 2010that's just one example off the top of my head of a player not regressing to the mean.

Gronkowski was a rookie in 2010...he couldn't regress to anything, he had no established "mean."When people talk about Gronkowski regressing to the mean, you should probably understand that the term doesn't refer to the mean of the top 10 TEs or a similar category...it refers to HIS mean. Based on his TD rate, utilization inside the red zone, average yards after the catch, average depth of target, reception %, etc...Now, I think you do know that, and you're kind of being a #### to try and prove the point that most people don't get it. Which is understandable, but also a little annoying when you respond in a similar way to a guy like Synthesia (who clearly seems to understand) and ask questions about JaMarcus Russell and "progression to the mean." Either you're incredibly dense, kind of a jerk, or you know that JaMarcus Russell wasn't (isn't, I suppose) very good at playing quarterback in the NFL...his baseline of production kind of sucked. You don't progress to the average of your position, you simply tend to move, be it up or down, from a hot streak or cold streak, back to your typical production. Now - we have to be wary of simply assuming that something is a hot streak or a cold streak...for instance when LT's ypc dropped off, it turned out not to be a cold streak, but rather a decline in abilities. One example of a similar thing I expect this year is Jimmy Graham. For all we know, last year's production is the true reflection of his abilities. It was his first true full season in the NFL...he hasn't truly established a baseline production to regress to.

 

[Kool-Aid Larry]

See: Dwayne Bowe 2010, Peyton Manning 2004, Tom Brady 2007, Ben Roethlisberger 2007, Steve Smith 2005.Just 4 examples off the top of my head. Usually "regression to the mean" is applied to TD rates.

gronkowski 2010that's just one example off the top of my head of a player not regressing to the mean.

Gronkowski was a rookie in 2010...he couldn't regress to anything, he had no established "mean."When people talk about Gronkowski regressing to the mean, you should probably understand that the term doesn't refer to the mean of the top 10 TEs or a similar category...it refers to HIS mean.

whooooaa, buddy.some guy earlier in the thread pegged it as the mean of the top 3 te.if you want to go your own way on that it's understandable --- which games count towards calculating gronk's mean?his rookie games count equally with sophomore games?psyou misunderstood my example

 

[kutta]

Regression to the mean is not a debatable concept, but it's one that is often misunderstood and inappropriately applied. At its core, regression to the mean is a concept borne out of the fact that there are statistical outliers in small samples that are not predictive of future results. If a roulette wheel lands on black 10 times in a row, we don't expect it to land on black in 10 of the next 10 spins, we'd expect it to regress to the mean.

If we replayed the 2011 season 1,000,000 times, do you think Gronkowski would average 17 TDs a season? I don't think so. Some years, he would get hurt. Other years, Brady would get hurt. He might be asked to block more often in some seasons, and in some years he would lose his confidence and play poorly. He'd do better than 17 TDs in a season sometimes, too. But on average, I don't think we'd expect Gronk to be a 17 TD/year guy.

I wrote a pretty thorough explanation of the concept here: http://www.pro-football-reference.com/blog/?p=370

As far as Gronk goes, I agree and disagree. I, too, don't think he will score 17 TD this year . . . but for different reasons. Over his career, he's played in 36 games (regular and post-season), in which he's scored 31 TD. That would make his run rate at 14 TD a season for a 16 game season. Counting all games evenly, his baseline over 16 games would be 77-976-14.I see Gronk losing some production compared to last year because of his recovery from his injured ankle (which I suggest will limit him some early in the season and could be a nagging injury throughout the year). I also see the new faces at WR taking away some of his targets. Other people have suggested that Brady won't throw the ball 600+ times again (which is a valid concern although I am not fully on board with that one).

IMO, there are potentially reasons why Gronk won't score 17 TD (which was actually 18) again this year. Sure, maybe regression has an element there, but there are other evidentiary factors that could also be valid concerns why he won't have a repeat of last year in the scoring department.

I agree 100%. Let's look more at the reasons someone's stats may decline instead of just saying they will decline due to regression.

 

[kutta]

Here's my main issue with it. Regression the mean is usually used to account for variations due to randomness or luck (see Chase's Roulette wheel example). So sure, of course I agree that there is an element of luck involved with players' stats in football. And, all things being equal, if a player completely blows up one year there is an expectation that he will come back down the next.

But the problem I have specifically with this year is that people are not looking at the reasons why some of the stats were so out of whack. I don't hear people talking about Ray Rice regressing to the mean this year, or Arian Foster, or McCoy. But they do talk about Graham, Gronk, Calvin, and the QB's regressing only because nobody has done what they did before. But they can't all be outliers. The game is changing. Pass Interference rules greatly favor the WR's. The QB's are not allowed to be hit. Teams are seeing other teams have great success throwing the ball, so they are emulating them. There is no reason to believe that the passing stats are going to go down anytime soon.

Do you really think that three (almost four, Eli was about 70 yards short) QB's throwing for 5000 yards in the same season is an anomaly? If it was one, maybe. But four? In 2011 we had four of the top six most prolific passing yardage performances in the history of the NFL. That is staggering. There were three (almost four - Brady had 39) QB's with over 40 passing TD's. Before 2010 there had only been 5 QB's in the history of the NFL with over 40 TD's. The league-wide passer rating average was 84.3 in 2011, breaking the record set in, you guessed it, 2010 when it was 84.1. I could go on and on.

The point being, the passing stats last year were consistent across many players and many teams. This is not one guy going crazy and blowing the rest of the league out of the water. It is happening all over the league. With the rule changes and more prolific offenses, these passing offenses are going to continue to put up gaudy numbers, and players are going to continue put up stats that people will continue to say must "regress to the mean." But what they don't understand is that the mean is changing and it is moving upward.

I get what you're saying, and for the most part I agree and think it's safe to say that the mean passing total by QBs is on increase, but to ask people to stop acknowledging the regression to mean concept is a silly response that doesn't properly deal with the issue you are trying to deal with. This post of yours here above explaining how people are using the concept incorrectly is much better than your initial post where you simply ask people to ignore the concept altogether. I'd also add that for these young players like Cam Newton and Rob Gronkowski, they really haven't played enough seasons for one to approximate what exactly their mean stats are imo. Will they both be hard pressed to repeat their TD numbers from last season? Probably. But I'd still feel more comfortable reserving the regression to the mean concept for veteran type players where you have more data to work with.

I am not asking anyone to stop acknowledging the concept. But I am asking that people not simply site "regression to the mean" as reason a player's stats will decline from one year to the next.

 

[kutta]

It's a valid concept, but people need to be clear just whose mean we are referring to. Regression to the mean doesn't mean regression to the league average. Jimmy Graham and Tony Scheffler are not regressing to the same mean. To illustrate using a baseball analogy: let's say you have a .300 hitter. Let's say over 10 at bats, he goes 8-for-10. How many hits should we expect over the next 10? Right around 3. What if he goes 0-for-10 over those 10, what should we expect over the 10 after that? Right around 3 hits. If he starts out the first two weeks of the season hitting .500, where should we expect him to finish the season? Right around .300. Why? Because the guy's mean is .300, and he'll tend to regress to it. A .200 hitter will regress to something completely different. Take Calvin Johnson last season. Over the first 4 weeks, Calvin scored 2 TDs every week. Everyone started speculating about how many TDs Calvin would manage to put up, taking it as foregone conclusion that he was going to set the TD record. And then a funny thing happened on the way to the record books- over the last 12 weeks of the season, Calvin's TD-per-game rate was essentially identical to his TD-per-game rate from 2008-2010. His hot start meant nothing- he regressed to pretty much his exact career averages. Regression to the mean at work!Players can improve, and their "true mean" can change, but the concept of regression to the mean really only means that any player can (and will) play over (or under) their head over a short enough timeline, but as you extend the timeline they'll always return to their "true" performance level. It's up to us to try to determine whether, for instance, Cam Newton's 14 rushing scores, or Aaron Rogers' 45 passing scores represent their true mean or an example of them playing over their heads over a short timeline.

yeah, why is it nobody ever talks about 'progression to the mean' for some guy like jamarcus russell?

Do you think Chris Johnson will be better this year?

Chris Johnson's drop off last year was due to specific reasons - he held out and didn't get the work in camp, and the offensive line was playing poorly. Yes, he will do better this year, but not because of this mystical regression/progression to the mean. He will do better because he will be in camp, he is working harder, and the offensive line will play better.

 

[az_prof]

A short, open letter to the FBG community:I'm really tired of hearing the phrase "regression to the mean." Just because someone has a great year doesn't mean the next year is going to be worse. There are tons of factors that go into projections, and just saying you think someone is going to do worse because of so called "regression" is weak. Can we try to avoid using this over-used phrase and start basing our projections on real stuff?Thank you.

If someone has a career year or a record setting year, I am sorry, but the odds are that they will in fact regress to the mean. You may not like it, but it is true. Most people tend to assume the opposite--that a good year will automatically become even better. Pointing out RTM is simply trying to correct that false tendency.

 

[CalBear]

If someone has a career year or a record setting year, I am sorry, but the odds are that they will in fact regress to the mean. You may not like it, but it is true.

But it's irrelevant. If one RB runs for 2000 yards and another one runs for 1200 yards, the smart money is on the 2000 yard rusher to out-gain the 1200 yards rusher in year N+1. He probably won't go for 2000 yards again but he's still a better pick than the guy who isn't as good as he is.

 

[rizzler]

TIP OF THE DAY:



Taking a first year stats course makes you as much a statistician as a film buff is a director.

As much as I love this website, and Shark Pool... The terms value and regression to the mean have lost all meaning to me. They are so often misappropriated by numbskulls who give their all to project themselves as fantasy guru's, that I couldn't care to argue the concepts anymore.

to me for being a gut instinct and NFL fanatic type drafter. Not getting caught up in alien concepts has been very kind to me.

 

[Fensalk]

A short, open letter to the FBG community:I'm really tired of hearing the phrase "regression to the mean." Just because someone has a great year doesn't mean the next year is going to be worse. There are tons of factors that go into projections, and just saying you think someone is going to do worse because of so called "regression" is weak. Can we try to avoid using this over-used phrase and start basing our projections on real stuff?Thank you.

I haven't been paying attention, but is whoever is "misusing" this phrase saying 1. Player will WILL CERTAINLY fall back closer to his average production.or2. Because statictics tend to regress to the mean, we should be wary of expecting player x to post those outlier statistics from last year, again this year.I say this because often I see people misinterpret someone's statement, and then go on a rant claiming they guaranteed it, when they did not.

 

[Time Kibitzer]

TIP OF THE DAY:



Taking a first year stats course makes you as much a statistician as a film buff is a director.

As much as I love this website, and Shark Pool... The terms value and regression to the mean have lost all meaning to me. They are so often misappropriated by numbskulls who give their all to project themselves as fantasy guru's, that I couldn't care to argue the concepts anymore.

to me for being a gut instinct and NFL fanatic type drafter. Not getting caught up in alien concepts has been very kind to me.

Regression to the mean is the opposite of an alien concept though.

 

[rizzler]

Regression to the mean is the opposite of an alien concept though.

Anything aside from what I see or my gut tells me, is alien. That was my point.

 

[Instinctive]

See: Dwayne Bowe 2010, Peyton Manning 2004, Tom Brady 2007, Ben Roethlisberger 2007, Steve Smith 2005.Just 4 examples off the top of my head. Usually "regression to the mean" is applied to TD rates.

gronkowski 2010that's just one example off the top of my head of a player not regressing to the mean.

Gronkowski was a rookie in 2010...he couldn't regress to anything, he had no established "mean."When people talk about Gronkowski regressing to the mean, you should probably understand that the term doesn't refer to the mean of the top 10 TEs or a similar category...it refers to HIS mean.

whooooaa, buddy.some guy earlier in the thread pegged it as the mean of the top 3 te.if you want to go your own way on that it's understandable --- which games count towards calculating gronk's mean?his rookie games count equally with sophomore games?psyou misunderstood my example

It's not about which way I want to go. Regression to the mean is an individual player phenomenon when applied to sports. The .300 hitter on a hot streak or the Calvin Johnson TD examples are both perfect. Way to miss most of the post. Like I said, either you totally misunderstand the concept, or you get the concept and were purposely obfuscating things to prove a point. The guy earlier, who "pegged it" was wrong. He doesn't apply the concept correctly. P.S. your example was not misunderstood, it was a ####ty example. You listed it in the same format as the previous five examples, all of which are players who had an established baseline of production, clearly exceeded it in one season, and then returned to that established baseline of (in some cases, excellent) production. You tried to compare an orange to five apples. Again, I know from the past that you're kind of a jerk in your posts, but I also thought you were a decently bright guy. It seems to me that either I misjudged that/recall incorrectly...or you're purposely being obstinate to try and prove a point. If it's the latter, you aren't doing a very good job of it. ETA: Further, the examples provided by that other poster also seem to show an understanding that regression to the mean is an individual phenomenon. Manning didn't regress to normal QB passing standards, he regressed to more normal Peyton Manning passing stats. You're trying to say you understand the concept, but demonstrating in your posts that A) you're a #### or B) you clearly don't understand the concept.

 

[Kool-Aid Larry]

hmmmm......maybe if I sprinkle my posts with more name calling and filtered words I could make my point a little better, you ####### ####.

 

[Chase Stuart]

It's a valid concept, but people need to be clear just whose mean we are referring to. Regression to the mean doesn't mean regression to the league average. Jimmy Graham and Tony Scheffler are not regressing to the same mean.

To illustrate using a baseball analogy: let's say you have a .300 hitter. Let's say over 10 at bats, he goes 8-for-10. How many hits should we expect over the next 10? Right around 3. What if he goes 0-for-10 over those 10, what should we expect over the 10 after that? Right around 3 hits. If he starts out the first two weeks of the season hitting .500, where should we expect him to finish the season? Right around .300. Why? Because the guy's mean is .300, and he'll tend to regress to it. A .200 hitter will regress to something completely different.

Take Calvin Johnson last season. Over the first 4 weeks, Calvin scored 2 TDs every week. Everyone started speculating about how many TDs Calvin would manage to put up, taking it as foregone conclusion that he was going to set the TD record. And then a funny thing happened on the way to the record books- over the last 12 weeks of the season, Calvin's TD-per-game rate was essentially identical to his TD-per-game rate from 2008-2010. His hot start meant nothing- he regressed to pretty much his exact career averages. Regression to the mean at work!

Players can improve, and their "true mean" can change, but the concept of regression to the mean really only means that any player can (and will) play over (or under) their head over a short enough timeline, but as you extend the timeline they'll always return to their "true" performance level. It's up to us to try to determine whether, for instance, Cam Newton's 14 rushing scores, or Aaron Rogers' 45 passing scores represent their true mean or an example of them playing over their heads over a short timeline.

yeah, why is it nobody ever talks about 'progression to the mean' for some guy like jamarcus russell?

Do you think Chris Johnson will be better this year?

Chris Johnson's drop off last year was due to specific reasons - he held out and didn't get the work in camp, and the offensive line was playing poorly. Yes, he will do better this year, but not because of this mystical regression/progression to the mean. He will do better because he will be in camp, he is working harder, and the offensive line will play better.

So you're saying if we replayed Chris Johnson's 2011 season a million times, in some of them he wouldn't hold out, in some of them he would work harder, and in some of them his offensive line would play better?We also don't *know* why Johnson's stats sucked last year. Sometimes, a player just has a bad year because of the small sample size in an NFL season. http://www.pro-football-reference.com/articles/roboedge.htm