Welcome Christian Covington

[DragonIce]

I am very convinced that PFF would agree with my point as it is completely obvious.

I'm still new here, but my early impression is that you speak more dogmatically about things than is merited.

You think PFF's single-season grade emphatically proves that one player is better than another? Why don't you try asking PFF about how single season scores compare with career averages for predictive value. How about you ask them about regression to the mean? How about you ask them about the impact of coach and scheme on player scores.

[Bearded14YourPleasure]

On rare occasion, stats can be misleading. Generally, though, they are reliable. If you examine the stats correctly, they almost always do tell the story and tell it much better than confusing team results with individual player performance, which is the single greatest analytical error perpetrated by fans and pundits that follow NFL football. Trent Dilfer was not better than Dan Marino. Jim Plunkett was not better than Dan Fouts, nor was Terry Bradshaw. Eli Manning was not better than Philip Rivers. Do you know why? Because their stats say so.

There are 3 kinds of lies: lies, damn lies, and statistics.

Especially in football, to pretend that statistics alone tell the whole truth is silly. The situation in which a player finds themselves can directly impact the statistics they have. That is why FA is such a hit or miss business, changing the situation and/or role of a player can turn them into a no brainer of a pickup into a huge mistake that sets the a franchise back for years. It’s also why the later you get in the draft, the more variation you see in how pundits and fans view player rankings compared to where a player is drafted. If statistics alone told the whole story would KJ Hill, the all time leading receiver at Ohio State, be available in the 7th round?

Statistics have a measure of meaning to them but to say statistics alone are the full measure of a player is laughable, even more so if you want to use a subjective rating system like PFF as a ‘statistic’.

[Guest]

I'm still new here, but my early impression is that you speak more dogmatically about things than is merited.

You think PFF's single-season grade emphatically proves that one player is better than another? Why don't you try asking PFF about how single season scores compare with career averages for predictive value. How about you ask them about regression to the mean? How about you ask them about the impact of coach and scheme on player scores.

No, that is not what I said.

I said that the greater the difference between two players' PFF numbers in a single season, the more likely it is that the player rated higher was actually better than the player rated lower in the season in question. I have not stated anything else about the PFF numbers (other than that I do not accept exact numbers as exact measurements of player performance) and have only applied that basic premise to the comparison of Square and Covington where there was a difference of approximately +13 in Square's favor for 2020, which is a medium difference between the two players, neither small, nor huge.

If the players were closely scored, then it would only take a slight error in the wrong direction to rate the better player lower. But the greater the separation in scoring, the greater that error has to be to produce the wrong better player versus worse player result for that one season.

Simply put, a small net error is much more likely than a significantly larger one.

It is not that I am dogmatic about that. Rather, it is that my limited statement is so obvious that it is a virtual truism.

[Guest]

There are 3 kinds of lies: lies, damn lies, and statistics.

Especially in football, to pretend that statistics alone tell the whole truth is silly. The situation in which a player finds themselves can directly impact the statistics they have. That is why FA is such a hit or miss business, changing the situation and/or role of a player can turn them into a no brainer of a pickup into a huge mistake that sets the a franchise back for years. It’s also why the later you get in the draft, the more variation you see in how pundits and fans view player rankings compared to where a player is drafted. If statistics alone told the whole story would KJ Hill, the all time leading receiver at Ohio State, be available in the 7th round?

Statistics have a measure of meaning to them but to say statistics alone are the full measure of a player is laughable, even more so if you want to use a subjective rating system like PFF as a ‘statistic’.

Nobody has suggested that PFF is a statistic. I use PFF as a general imprecise qualitative evaluator of players under the theory that PFF may be off some, but is not likely to be miles and miles off. It tends to point in the right general direction, but is worthless as an exact valuation of a player.

Nobody has suggested that college stats should be used to value a player at the NFL level, though I have suggested the opposite with respect to numerous players as there may be reasons why college production does not translate to the NFL level, a point I have repeatedly emphasized. K.J. Hill is a perfect example of my point. My take on him is that he is just not an NFL caliber player, which is why I think it is okay if Palmer makes the team as Palmer is more gifted than Hill as a player and probably more so than Reed as a receiver, but probably not overall when Reed's return and running games are considered.

NFL statistics, much more than anything else, do tell the story. I completely agree that the situation in which players find themselves can impact their statistics. I can only guess about the number of passing TDs lost in 2006 and 2007 for Rivers because the team decided to have LT run for the TD instead of having Rivers pass for that same TD. But we cannot give credit to a player for what he did not do even if he could have done it.

I definitely think that Brady's stats have been greatly aided by his teammates, head coaches, offensive coordinators and GMs. Brady was drafted in round 6 because he was not very physically talented, but he has been the ultimate "in the right spot at the right time" player. He deserves credit for the stats he has produced, which are some of the best ever for a QB. His teams deserve credit for being Super Bowl champions as that is a team accomplishment, not an individual accomplishment.

[DragonIce]

Chain...you are a piece of work.

It is not that I am dogmatic about that. Rather, it is that my limited statement is so obvious that it is a virtual truism.

You crapped your own bed and still haven't noticed?

:LOL:

[Steve]

BY themselves, statistics often don't mean what you think they do. They need some context to really help you understand what they do mean. That is true of ALL statistics. That is true of all branches of mathematics for that matter.

PFF has a grading scheme that is decidedly NOT a statistic. They have people grade film watching each player on each play, and providing a grade. They tally them for every game, and then the entire season. The problem I have always had is the assumptions being made. PFF has always said that they assume that NFL player know their assignments, and also they don't make any effort to grade technique and just grade the results. This is pretty flawed in 2 areas. There is no question that a lot of NFL players, maybe even the majority, make a lot of mental mistakes including but not limited to simply not knowing their assignments on any given play. They also claim that only the results matter, but it is hard to tell what a player or scheme is asking of a player, which is the justification for only grading on results, but it is tragically flawed, as well as being intellectually lazy. PFF grades can be useful, but there are a lot of hidden biases. For example, zone CB tend to have much higher grades than players who play more man coverage. Zone OL tend to get much lower grades than man to man or gap schemes. 2 gapping DL and read and react LB are also pretty under values.

As far as other statistics, you have to understand some of the underlying assumptions and mechanics to gleem much value out of them. a good starting point is the hidden game of football (https://www.amazon.com/Hidden-Game-F.../dp/0446514144).

[Guest]

Chain...you are a piece of work.


You crapped your own bed and still haven't noticed?

:LOL:

You assertion is ridiculous and even comical in light of your failure to grasp something that is almost as simple as 2+2 being equal to 4.

Again, and Lord please give me patience, PFF scores every play, right? Most of the plays are scored correctly, but there are inevitably going to be some errors, right? Knowing no other facts and assuming that an error has taken place, a scoring error on a single given play is just as likely to be too low as it is too high, right? Those are the two ways in which PFF can make a scoring error, right?

If both possibilities are equally likely, what is the expected net value of those scoring errors? Generally speaking, where the large majority of plays are scored zero, the answer is zero. It is a coin flip as to whether the error will be by giving too much credit or too little credit.

Now, just because the expected result is zero does not mean that we will always get an exact offset with our heads versus tails or our scored too high or too low plays. However, over a number of errors, the most likely outcomes are those that are zero or closer to zero either way because an error one way offsets an error the other way.

Yes, it is theoretically possible to get a wildly skewed result (for example, you can flip 20 coins and have 16 of them come up heads), but those examples are going to be statistical outliers that are relatively rare. Far more common will be a small net error.

So, all I have done is to say that I accept the (obvious) notion that a small net error in a PFF grade is far more likely than a large one. It is more likely that a PFF grade is 3 points off than it is that it is 30 points off.

Then, with that concept in mind, I have looked at the scores for Square and Covington for the 2020 season only. (My analysis has absolutely nothing to do with other seasons, so if you are talking about other seasons, you are talking about other concepts that I am not discussing, concepts about which you may be right or wrong.) When I did so, I saw roughly a 13 point difference, which in PFF terms is neither a small nor a large scoring difference. It is kind of a medium difference.

And because it is a medium difference, I have stated that Square was more likely the better player in 2020 under a correct (no errors) PFF analysis because the difference in scoring would require a net scoring error that is less likely than not to be present. Again, the basic premise is that the farther apart the scores are, the less likely it is that that scoring difference is accounted for by a net scoring error. I have not stated that it was impossible, just that at that magnitude of scoring difference, the likelihood of a net scoring error accounting for the difference is less likely.

That is pretty much all I have said. If you have a problem with any of that, I think maybe you should wake up from your dream and smell your surroundings carefully before you roll over in the bed.

[Guest]

BY themselves, statistics often don't mean what you think they do. They need some context to really help you understand what they do mean. That is true of ALL statistics. That is true of all branches of mathematics for that matter.

PFF has a grading scheme that is decidedly NOT a statistic. They have people grade film watching each player on each play, and providing a grade. They tally them for every game, and then the entire season. The problem I have always had is the assumptions being made. PFF has always said that they assume that NFL player know their assignments, and also they don't make any effort to grade technique and just grade the results. This is pretty flawed in 2 areas. There is no question that a lot of NFL players, maybe even the majority, make a lot of mental mistakes including but not limited to simply not knowing their assignments on any given play. They also claim that only the results matter, but it is hard to tell what a player or scheme is asking of a player, which is the justification for only grading on results, but it is tragically flawed, as well as being intellectually lazy. PFF grades can be useful, but there are a lot of hidden biases. For example, zone CB tend to have much higher grades than players who play more man coverage. Zone OL tend to get much lower grades than man to man or gap schemes. 2 gapping DL and read and react LB are also pretty under values.

As far as other statistics, you have to understand some of the underlying assumptions and mechanics to gleem much value out of them. a good starting point is the hidden game of football (https://www.amazon.com/Hidden-Game-F.../dp/0446514144).

I agree that PFF is a grading system and not a statistic. Back in school days, though, I could tell that one student did better than another because his A was better than the other guy's C.

Also, I question your statement that they "just grade the results". As I understand it, QBs that make good throws are not downgraded if the throw goes right through the WR's hands and into a defender's hands, so that seems to be different than what you are saying.

That said, I think there are fair criticisms of their approach and I certainly do not think that their grade exactly measures a player's performance. As I have stated repeatedly, PFF is good for pointing readers in the right direction and is useful in telling people which player is better in cases where the scored difference between two players is larger, but not so much when the score is close. In those latter cases, it is not very useful at all. (Nobody is saying that PFF can be relied upon to establish that a player graded, for example, 80.2 is better than a player graded 79.8.)

[equivocation]

A drop is a bad result for a WR (drop) and good result for a QB (delivered accurate pass). They also grade results of individual battles without the ball (so WR got open vs DB covered) or blocked vs not blocked. This is why they had Palmer as the 70 overall prospect.

They don't grade individual technique even if both players do a good job which biases certain schemes and can make them succeptible to the Peter Principal (Lamp had a super high grade in small ball college)

This often results in high year to year variation especually in the trenches. To continue your analogy, if a student got an 'A' on one test and a 'C' on another, how do you predict the next test? Grading on only the latest test introduces recency bias.

[Guest]

A drop is a bad result for a WR (drop) and good result for a QB (delivered accurate pass). They also grade results of individual battles without the ball (so WR got open vs DB covered) or blocked vs not blocked. This is why they had Palmer as the 70 overall prospect.

They don't grade individual technique even if both players do a good job which biases certain schemes and can make them succeptible to the Peter Principal (Lamp had a super high grade in small ball college)

This often results in high year to year variation especually in the trenches. To continue your analogy, if a student got an 'A' on one test and a 'C' on another, how do you predict the next test? Grading on only the latest test introduces recency bias.

That is how I understood PFF grading as well.

A PFF grade is for a whole season, so I think the latest trend should take on increased significance relative to earlier seasons absent reasons to discount it.

In my student example, you only assumed that it was one test when I discussed the grade. I actually meant the final class grade, which fits more closely with a final grade for an NFL season.

I think the label of "recency bias" suggests undue weight being given to the recent. The weight given to the more recent may be nothing more than appropriate due consideration. It may be the most predictive indicator.

[equivocation]

PFF has high year to year variance especially for interior line play. Taking ANY ONE YEAR as a better indicator than any other makes no sense if you're being objective. You need to balance that against overall trend, career, line stats, and film. Literally the only advantage Square has over Covington is 2020 PFF grade. Which again, has high variance.

[Steve]

I agree that PFF is a grading system and not a statistic. Back in school days, though, I could tell that one student did better than another because his A was better than the other guy's C.

Also, I question your statement that they "just grade the results". As I understand it, QBs that make good throws are not downgraded if the throw goes right through the WR's hands and into a defender's hands, so that seems to be different than what you are saying.

That said, I think there are fair criticisms of their approach and I certainly do not think that their grade exactly measures a player's performance. As I have stated repeatedly, PFF is good for pointing readers in the right direction and is useful in telling people which player is better in cases where the scored difference between two players is larger, but not so much when the score is close. In those latter cases, it is not very useful at all. (Nobody is saying that PFF can be relied upon to establish that a player graded, for example, 80.2 is better than a player graded 79.8.)

PFF can be maddening because they ignore a lot of the context they are trying to add.

For example, they tend to give high grades to players who matchups are making a lot of mistakes. Is a DE who gets a bunch of sacks really playing all that well, if the OL opposite him keeps trying to block someone else? How about the OLB who gets his sacks/pressures off of a TE or RB? Those are gimme plays, and there are a lot of people who don't think you should give a top grade for that kind of thing. PFF has never released their grading guidelines, but the discussions they have about it (only grading production and assuming that players know what they are doing) seems to indicate they give a player top marks for "gimme" plays.

The bigger issue is the coverage thing for the zone pass coverage CB, zone blocking OL and 2 gapping DL. Those are all examples of cases where the coaches are making decisions about the team and helping some players at the expense of others. That is going to be particularly true since the Chargers are now using a 2 gapping DL scheme. How do we credit a DL who stands his man up, forces the other OL to help out, but doesn't make the tackle and is slow on the pass rush? What about the little guys behind him who are now free to run to the ball, with the extra step or two that they have on the blockers? The DL may not make the play, but the guy who did is being given a "gimme".