Jump to content

Recommended Posts

Posted

I dislike his charts. The NFL plays a 16 game schedule, compared to the NBA, MLB, and NHL, that's basically a month. The difference between a 4 win and 8 win team is stark, and he lumps them into one group. He could look at numbers of teams within a standard deviation, but doesn't. Or size/spread of deviation. He didn't really prove any lack of parity. If he wants to talk about lack of pay in the NFL given their part of revenue generation, career length, and injury risk, he should do so, but placing it on the back of parity I think is flawed.

Posted

Canisius HS, Penn State grad and Ravens OL. Great read, he's brilliant.

 

The Parity Ideal

 

http://www.theplayerstribune.com/john-urschel-ravens-parity-ideal/

Best sports article I've read in many a moon. Thanks for posting!

 

Really true too.

I dislike his charts. The NFL plays a 16 game schedule, compared to the NBA, MLB, and NHL, that's basically a month. The difference between a 4 win and 8 win team is stark, and he lumps them into one group. He could look at numbers of teams within a standard deviation, but doesn't. Or size/spread of deviation. He didn't really prove any lack of parity. If he wants to talk about lack of pay in the NFL given their part of revenue generation, career length, and injury risk, he should do so, but placing it on the back of parity I think is flawed.

Disagree - the percentages should still work out to be the same. The sample size is still certainly large enough. Re: Urschel, he has an MA in math: 'In 2015, Urschel co-authored a paper in the Journal of Computational Mathematics.[12] It is titled "A Cascadic Multigrid Algorithm for Computing the Fiedler Vector of Graph Laplacians" and includes "a cascadic multigrid algorithm for fast computation of the Fiedler vector of a graph Laplacian, namely, the eigenvector corresponding to the second smallest eigenvalue."[13]'

Posted (edited)

I dislike his charts. The NFL plays a 16 game schedule, compared to the NBA, MLB, and NHL, that's basically a month. The difference between a 4 win and 8 win team is stark, and he lumps them into one group. He could look at numbers of teams within a standard deviation, but doesn't. Or size/spread of deviation. He didn't really prove any lack of parity. If he wants to talk about lack of pay in the NFL given their part of revenue generation, career length, and injury risk, he should do so, but placing it on the back of parity I think is flawed.

 

I'm with you, this is not a very useful article and a rather poor, biased one considering Urschel's intelligence. But being able to do higher level calculus and analyze a real world situation are obviously two very different things.

 

Urschel seems to assert that there is parity in all 4 leagues and salary structure does not make a difference. To do this he defines parity as performing to a 0.500 record and first takes every team's record over a decade and bins it into a league-wide bar graph. Is this appropriate? What about charting individual teams year to year according to the mean?

 

In an ideal world, the mean would equal the median at 0.500 and there would be a normal (symmetric) distribution about the mean. What are the actual means/medians and corresponding errors for each league? There is a negative skew in every league so can he justify all these numbers being close enough to 0.500 to give parity? If not, what is the reason?

 

In the last 10 years (2015-2006)the Bills' avg # of wins is 6.6 +/- 1.35

the New England Patriots: 12.4 +/- 1.65 Do these fit the definition of parity?

 

IMO, his best attempt to answer this is looking at year to year movement, i.e. the pie charts showing "stays put/moves towards the mean" vs "moves away from the mean". Even this seems weak as first of all you would want to split out the "stays put" group and these metrics would probably be better assessed with individual teams over the 10 years. How many individual teams fit a regular distribution about the mean? This would seem to me to be an interesting test of parity.

 

So, whether or not there is parity, how do you justify salary structure as a significant factor as compared to quality of coaching, quality of talent selection and acquisition, injuries, etc? In each league, has the salary structure changed over time to any measurable effects on parity? Is there any correlation of baseball payrolls with winning? What comparisons between baseball and football, who have the biggest comparative divergence in salary structure?

 

Really fascinating questions to me but Urschel's article really doesn't enlighten much.

Edited by GaryPinC
Posted (edited)

Unlike a couple of people here, I really thought the piece was convincing. That said, it has one important flaw, and it relates to baseball. Whereas NFL teams start the same players every game (barring injury), baseball teams have essentially five squads (i.e., five different starting pitchers), and no one team has a flawless pitching staff (i.e., the equivalent of Manning or Brady in their prime at QB). Hence for every Zack Grienke who goes 20-3 with a 1.66 ERA, a staff will have a couple of inning eaters who have a high-3 or low-4 ERA. 108-54 in baseball is staggeringly good, and it's basically 11-5. 91-71 is very good too, and it usually (not always) gets you to the playoffs. But it's the same as 9-7. There are no 130-30 (13-3) teams in MLB. However, you do get that sort of thing in the NBA, where dominant teams often win 60-65 games in their best seasons.

Edited by dave mcbride
Posted

Unlike a couple of people here, I really thought the piece was convincing. That said, it has one important flaw, and it relates to baseball. Whereas NFL teams start the same players every game (barring injury), baseball teams have essentially five squads (i.e., five different starting pitchers), and no one team has a flawless pitching staff (i.e., the equivalent of Manning or Brady in their prime at QB). Hence for every Zack Grienke who goes 20-3 with a 1.66 ERA, a staff will have a couple of inning eaters who have a high-3 or low-4 ERA. 108-54 in baseball is staggeringly good, and it's basically 11-5. 91-71 is very good too, and it usually (not always) gets you to the playoffs. But it's the same as 9-7. There are no 130-30 (13-3) teams in MLB. However, you do get that sort of thing in the NBA, where dominant teams often win 60-65 games in their best seasons.

 

It's a great point about baseball and only one of many flaws of what he is trying to assert, IMO. Is it really so groundbreaking to take the W-L of every team in the league for a decade, throw them all together and show a regular distribution about a mean and median? Every league has a negative skew though, is it small enough to keep with parity? If the same teams are always to the left or right of the mean is that truly parity?

Posted (edited)

 

I'm with you, this is not a very useful article and a rather poor, biased one considering Urschel's intelligence. But being able to do higher level calculus and analyze a real world situation are obviously two very different things.

 

Urschel seems to assert that there is parity in all 4 leagues and salary structure does not make a difference. To do this he defines parity as performing to a 0.500 record and first takes every team's record over a decade and bins it into a league-wide bar graph. Is this appropriate? What about charting individual teams year to year according to the mean?

 

In an ideal world, the mean would equal the median at 0.500 and there would be a normal (symmetric) distribution about the mean. What are the actual means/medians and corresponding errors for each league? There is a negative skew in every league so can he justify all these numbers being close enough to 0.500 to give parity? If not, what is the reason?

 

In the last 10 years (2015-2006)the Bills' avg # of wins is 6.6 +/- 1.35

the New England Patriots: 12.4 +/- 1.65 Do these fit the definition of parity?

 

IMO, his best attempt to answer this is looking at year to year movement, i.e. the pie charts showing "stays put/moves towards the mean" vs "moves away from the mean". Even this seems weak as first of all you would want to split out the "stays put" group and these metrics would probably be better assessed with individual teams over the 10 years. How many individual teams fit a regular distribution about the mean? This would seem to me to be an interesting test of parity.

 

So, whether or not there is parity, how do you justify salary structure as a significant factor as compared to quality of coaching, quality of talent selection and acquisition, injuries, etc? In each league, has the salary structure changed over time to any measurable effects on parity? Is there any correlation of baseball payrolls with winning? What comparisons between baseball and football, who have the biggest comparative divergence in salary structure?

 

Really fascinating questions to me but Urschel's article really doesn't enlighten much.

Excellent points, Mango and Gary. Urschel has an intriguing thesis: that the NFL sells its version of the salary cap as essential to maintain parity, but that this is essentially b.s.; it's really about maintaining an imbalance in bargaining power between management and labor. But his stats fall far short of the mark. Looking at fluctuations for individual teams (Season X to Season X+1) gives results that are consistent with random changes in situational luck. Baseball analysts have the benefit of better measures than W-L records; that's why they tend to focus on pythagorean records (expected W-L records based on run differential) which have much better predictive ability. And there's also a problem in comparing the NFL to the other 3 major North American pro sports leagues, 2 of which also have salary caps. (I don't know enough about European soccer leagues - do they have salary caps? If not, they might provide better comparisons to isolate the true effect of salary caps.). And then, of course, you need to account for the fact that the drafts being based on the prior season's record. In leagues where one great draftee can make a huge, immediate difference (think Andrew Luck or even Eichel as opposed to the #1 pick in the MLB draft, in which there's always a 1-4 year delayed impact at best) you'd expect some significant upward regression to the mean among the worst teams. And it's also very difficult to isolate the impact in the NFL of the salary cap from that of revenue sharing, which barely exists in baseball but which has a huge impact in the NFL. The NFL has explicitly linked the cap to revenue sharing -- a linkage that is by no means required. That attempt at linkage provides some anecdotal support for Urschel's thesis, but again, it muddies the statistical waters.

 

So ... it's kind of a mess. But I can't be too critical, because I am really interested in the notion that the NFL sells a salary-suppressing scheme as critical for the good of the league overall, and ultimately for the good of the players who participate in it. Is Urschel on Bernie Sanders' contributor list?

Edited by The Frankish Reich
  • 4 weeks later...
Posted (edited)

I dislike his charts. The NFL plays a 16 game schedule, compared to the NBA, MLB, and NHL, that's basically a month. The difference between a 4 win and 8 win team is stark, and he lumps them into one group. He could look at numbers of teams within a standard deviation, but doesn't. Or size/spread of deviation. He didn't really prove any lack of parity. If he wants to talk about lack of pay in the NFL given their part of revenue generation, career length, and injury risk, he should do so, but placing it on the back of parity I think is flawed.

 

Also the sample size is so small with the NFL, that a bad few weeks, an untimely injury, or a lucky or unlucky bounce a few times can be the difference between making the playoffs or not. The NHL and NBA play over 5 times more games so it's more unlikely that "luck" will play a huge factor, and with the MLB playing almost twice as many games as the other two leagues and 10 times more than football, luck plays very little part in the grand scheme of things.

 

Each year the NFL has 2-3 teams at least that are in the playoffs that likely wouldn't be if the NFL changed to an 18 game schedule just due to the random things that can happen and change the outcome of games that would be a much smaller factor if they played 82 games and almost no factor at 162 games...

 

So in the NFL you have 6-8 really good teams, 6-8 really bad teams and then the rest somewhere in the middle that luck plays a huge factor for determining if they end up in the playoffs or not. When they play a team(do they play them after their starting QB is out for the year or before, early in the year before they've hit their stride or later once they've gotten on a roll, etc), what players are injured when they play, a "bad bounce", dropped pass, replay review, missed/made FG, etc all can turn the outcome of probably 4-5 games for most of these teams and it's the teams who can go 5-0, 4-1, or 3-2 in these games that make the playoffs from this group of teams and the ones who go 0-5, 1-4, or 2-3 usually miss out...so obviously there is a lot of parity because of small sample size, randomness, and lack of time for regression to the mean to occur(i.e., those "bad breaks"/"good breaks" don't have a chance even out with time).

 

Each NFL game is 6.25% of a teams schedule. The NBA/NHL each game is 1.2% of the schedule or for the MLB 0.6% of the schedule. So if a player misses a week in the NFL it's the equivalent of them missing 5 games in the NBA/NHL or 10 games in MLB. So if Steph Curry misses one game for the Warriors, it's unlikely to effect their record very much. If Tom Brady misses one game for the Patriots, it has a much larger effect on their record.

Edited by matter2003
Posted

I dislike his charts. The NFL plays a 16 game schedule, compared to the NBA, MLB, and NHL, that's basically a month. The difference between a 4 win and 8 win team is stark, and he lumps them into one group. He could look at numbers of teams within a standard deviation, but doesn't. Or size/spread of deviation. He didn't really prove any lack of parity. If he wants to talk about lack of pay in the NFL given their part of revenue generation, career length, and injury risk, he should do so, but placing it on the back of parity I think is flawed.

 

An interesting article but I agree with Mango. You can't compare a short 16 game football season against the much, much longer seasons of basketball, baseball and hockey.

 

I guess you could compare the parity of the NFL versus the parity of the NBA, NHL and MLB after they were 16 games into a season.

 

But the real question is: would a soft cap, higher rookie pay, or any other changes give the NFL more parity. Urschel never successfully makes that point, though he seems to suggest it. Forget about what may or may not work for the other sports. I want to see a reasoned argument that there's a way to make the NFL better. This article doesn't have it.

Posted

Also the sample size is so small with the NFL, that a bad few weeks, an untimely injury, or a lucky or unlucky bounce a few times can be the difference between making the playoffs or not. The NHL and NBA play over 5 times more games so it's more unlikely that "luck" will play a huge factor, and with the MLB playing almost twice as many games as the other two leagues and 10 times more than football, luck plays very little part in the grand scheme of things.

 

Each year the NFL has 2-3 teams at least that are in the playoffs that likely wouldn't be if the NFL changed to an 18 game schedule just due to the random things that can happen and change the outcome of games that would be a much smaller factor if they played 82 games and almost no factor at 162 games...

 

So in the NFL you have 6-8 really good teams, 6-8 really bad teams and then the rest somewhere in the middle that luck plays a huge factor for determining if they end up in the playoffs or not. When they play a team(do they play them after their starting QB is out for the year or before, early in the year before they've hit their stride or later once they've gotten on a roll, etc), what players are injured when they play, a "bad bounce", dropped pass, replay review, missed/made FG, etc all can turn the outcome of probably 4-5 games for most of these teams and it's the teams who can go 5-0, 4-1, or 3-2 in these games that make the playoffs from this group of teams and the ones who go 0-5, 1-4, or 2-3 usually miss out...so obviously there is a lot of parity because of small sample size, randomness, and lack of time for regression to the mean to occur(i.e., those "bad breaks"/"good breaks" don't have a chance even out with time).

 

Each NFL game is 6.25% of a teams schedule. The NBA/NHL each game is 1.2% of the schedule or for the MLB 0.6% of the schedule. So if a player misses a week in the NFL it's the equivalent of them missing 5 games in the NBA/NHL or 10 games in MLB. So if Steph Curry misses one game for the Warriors, it's unlikely to effect their record very much. If Tom Brady misses one game for the Patriots, it has a much larger effect on their record.

The "luck" disparity you mention between the NFL and baseball can't be proven one way or another. You say luck plays little part in MLB in the grand scheme of things compared to NFL bc of the greater number of games involved. But that statement is directly dependent on the assumption that "good luck" or "bad luck" eventually even out over a MLB season specifically because of the duration of the season. That is nothing more than an assumption.

 

Something important here as well would be defining exactly what "luck" is. We will find such a task extremely difficult.

Posted

Disagree - the percentages should still work out to be the same. The sample size is still certainly large enough. Re: Urschel, he has an MA in math: 'In 2015, Urschel co-authored a paper in the Journal of Computational Mathematics.[12] It is titled "A Cascadic Multigrid Algorithm for Computing the Fiedler Vector of Graph Laplacians" and includes "a cascadic multigrid algorithm for fast computation of the Fiedler vector of a graph Laplacian, namely, the eigenvector corresponding to the second smallest eigenvalue."[13]'

 

Fiedler was a hack.

×
×
  • Create New...