Anthony Gonzalez and Wide Reciever Stats, Part 2
I explained how Catch% and YPR are very important tools for evaluating WRs and showed Anthony Gonzalez's success in both in part 1. Now I'll expand on the relationship between the two, and what it can tell us about WR performance.
YPR isn't a perfect measure of how deep a WR is catching the ball. A 15 yard pass could be a pass caught 3 yards deep and run for 12 more yards, or it could be a pass caught 15 yards deep with no YAC, or anything between. Since not all WRs are equally good at running after the catch, and certain routes allow for different amounts of YAC, YPR isn't the best measure of how deep a WR is catching the ball.
Luckily some places WR YAC is kept, allowing a calculation of the average depth a WR caught the ball. This should be the best measure of WR usage available without going through play by play.
There were 168 player-seasons with 50+ targets the last two years. Among these 168 their AirYPR and Catch% had a r squared value (the percent of the variance in Catch% accounted for by AirYPR) of .16 , meaning that 16% of the variation in catch% can be accounted for by YPR. That is a small correlation, but it is significant, especially when considering that it doesn't account for WR skill, defense ability or QB skill. The average depth a WR catches the ball at, accounts for 16% of the variation in the percent of the time he catches the ball. To relate this level of relatedness, this is a similar correlation to team wins vs Offensive Yards Per Carry and team wins vs team penalties (or more accurately, since the correlation for penalties is negative, team losses and penalties).
Just like committing penalties is a significant factor is losing games and running for a good YPC is related to winning, a WRs AirYPR is related to his Catch%.
So there's a solid relation between the two. Now the useful part. Mgrex03 lent me a hand to use a logarithmic regression instead of the a straight linear one I used before. Average AirAPR was 9.18 and average catch% was 58.8%. Combining these two with the logarithmic regression gives a formula for the catch% of an average receiver for a given AirYPR. We can compare this to the WRs actual catch% for a measure of how well a WR catches the football.
It's not an overall metric of WRs because YAC is excluded. A WR's job breaks down into Getting Open, Catching the Football and Running after the Catch. This covers none of the run after, pretty much all of the catching and some elements of getting open (more room to catch raising catch%). As a alternative to the rate stat, there's Receptions over Average, the WR's receptions minus (the expected catch% for his AirYPR muliplied by his targets). Giving how many more or less catches the player had compared to what would be expected for an average WR with the player's number of targets.
And now the results
Full spreadsheet here (download)
Top 10 Catch% over average
| Player | Targets | C% over Average |
| Ike Hilliard 08 | 58 | +17.8% |
| Anthony Gonzalez 07 | 51 | +16.2% |
| Jabar Gaffney 07 | 50 | +13.1% |
| Gonzalez 08 |
57 | +12.4% |
| Andre Johnson 07 | 86 | +12.3% |
| Steve Breaston 08 | 113 | +10.6% |
| Josh Reed 08 | 79 |
+10.3% |
| Bobby Engram 07 | 134 | +10.3% |
| Wes Welker 07 | 145 | +10.3% |
| Reggie Wayne 07 | 156 | +10.3% |
Bottom 10 Catch% over Average
| Player | Targets | C% over Average |
| Marty Booker 07 | 105 | -11.7% |
| Jerry Porter 07 | 102 | -12.1% |
| Ted Ginn Jr. 08 | 71 | -13.6% |
| Darrell Jackson 07 | 104 | -14.5% |
| Drew Bennett 07 | 73 | -14.8% |
| Justin McCareins 08 | 73 | -14.9% |
| Braylon Edwards 08 | 138 | -15.2% |
| Keary Colbert 07 | 69 | -15.6% |
| Roy Williams 08 | 82 | -16.0% |
| Brad Smith 07 | 67 | -16.5% |
Top 10 Receptions over Average
| Player | Targets | Rec over Average |
| Andre Johnson 08 | 170 | +16.4 |
| Wayne 07 | 156 | +16.0 |
| Welker 07 | 145 | +15.0 |
| Engram 07 | 134 | +13.9 |
| Breaston 08 | 113 | +12.0 |
| Colston 07 | 144 | +11.8 |
| Randy Moss 07 | 160 | +10.8 |
| Royal 08 | 129 | +10.7 |
| Mason08 | 58 | +10.6 |
| Andre Johnson 07 | 150 | +10.6 |
Bottom 10 Receptions over Average
| Player | Targets | Rec over Average |
| Keary Colbert | 69 | -10.4 |
| Arnaz Battle 07 | 104 | -10.8 |
| Drew Bennett | 73 | -10.8 |
| Justin McCariens 08 | 73 | -10.8 |
| Brad Smith | 67 | -11.1 |
| Marty Booker 07 | 105 | -12.3 |
| Jerry Porter 07 | 102 | -12.3 |
| Roy Williams 08 | 82 | -13.1 |
| Darrell Jackson 07 | 104 | -15.1 |
| Braylon Edwards 08 | 138 | -21.0 |
1 recs |
15 comments
|
Comments
ok so
For every season in the last two years where a WR had 50 or more passes thrown to them, I took the yards per catch, removed yards after catch, which gives the average distance past the line of scrimmage that the WR caught the football at. I also took the catch%. From this data I got a formula to predict a players catch% from the players AirYPR (the average distance past the line they caught the football).
So when a player outperformed the formula they showed above average hands for the type of routes they were running, when they underperformed it they showed poor hands.
To account for differences in number of targets and reward players that were good over a large amount of targets, I also did the receptions over average. The catch rate expected for the WRs AirYPR times their targets gives how many receptions you’d expect from a WR that ran that depth of routes, and got that many targets, so the expected receptions minus the actual ones compares their total catch production to what you’d expect.
I be ridin', just ridin' alone
With my daddy on my mind like you gotta be kiddin' me
How the hell you ain't here to see your prince do his thing
Sometimes I wanna drop a tear, but no emotion from a king
by shake n bake on Jun 21, 2009 9:54 PM EDT up reply actions 0 recs
Again dude
I think there are just wayyyyy too many variables. Lets take Arnaz Battle for instance. Now if you see 49er games you will notice Battle is actually a pretty darn good WR. He may not be great, but he has solid ability. Runs good routes and has good hands. The problem with Battle is not him the problem is that he plays for the 49ers who do not have a good QB or a good O-Line. You can see from the stats you posted that he gets targeted a lot, but does not get enough catches.
There is just too much that can skew the averages. What if a WR has a bad QB? What if the Oline is bad and results in hurried throws and thus more incompletions. There is just so much that a WR is dependent on. Too much that is beyond his control.
by MasterRWayne on Jun 21, 2009 11:26 PM EDT reply actions 0 recs
and those effect these more than conventional stats why?
I’m not claiming to have fixed every issue with player eval statistics, I’m just attempting to improve on what’s already out there.
I be ridin', just ridin' alone
With my daddy on my mind like you gotta be kiddin' me
How the hell you ain't here to see your prince do his thing
Sometimes I wanna drop a tear, but no emotion from a king
by shake n bake on Jun 21, 2009 11:51 PM EDT up reply actions 0 recs
Two things I really don't like about this "Watch the games" responce to stats
There are 267 games a year, that’s about 400 hours of football to watch at 90 minutes a game, if you are doing it on tape to just watch playtime and lining up. During the regular season that’s 23 hours of football played each week. Just watching every game once is a massive investment of time, and that’s assuming you can evaluate all 22 players on the field at the same time on your one run through. Realistically you probably can’t watch more than 3-4 players carefully on one playthough. So that’s about 6 times through a game to follow every player reasonably well (and DBs and WRs will be nearly impossible without coaches tape, which you can’t get unless you are in an organization, or are Ron Jaworski). 2,400 hours of tape time a year, That’s basically a full time job (8 hours a day, 300 days a year). 135 hours (5 1/2 days!) per week of regular season action.
If you are a rare person that has the means and the motivation to do that, then I bet you would have a better perception of player skill than any stat could ever give, but that’s not viable for anyone who isn’t making their living doing it. Stats are good for fans because they don’t have the time and resources for total immersion. Stats are good for organizations because they can uncover systematic biases (check out FO’s Lewin System stuff on college QBs, the gist is that scouts and GMs get sold on guys who either didn’t complete the ball consistently or that they don’t have a lot of starts, so don’t have a ton of tape to judge the player off of, which is shown by most 1st round busts having a low college completion% and/or relatively few college starts).
I’ll adapt a quote from Moneyballto kick off point two, The difference between Jerry Rice and Issac Bruce is a catch every other game. Could anyone simply watching their careers pick up on that without the benefit of adding up their statistics?
Peoples perceptions aren’t perfect. Even if you spent the couple thousand hours of tape to break down every player on every play, you wouldn’t remember them all and because of that the stuff that got excluded might not be entirely random. People tend to better remember information that fits with their previously held beliefs. If you watched every play from say Ryan Diem previously believing that he has trouble with speed rushers, when you afterwards think about Diem the plays that come to mind are likely to have a more than representative sample of him struggling against speedy pass rushers.
So in “Watching the games” you are more than likely only seeing a small sample of that player/teams performances, and all that you do see is subject to all of your personal and subconscious biases and memory shortcuts.
Stats gather up all the information that would take thousands of hours, and then with some statistical tools can disentangle some of the factors that can skew certain stats (like how AirYPC skews Catch%), to give a objective measure of what happened. You still have to make a subjective jump to go from what a player did, to how good they are, but went you make that jump with a solid base of statistics you only run your conclusions through your biases once, instead of twice, on both gathering the information and on going from performance to ability.
I be ridin', just ridin' alone
With my daddy on my mind like you gotta be kiddin' me
How the hell you ain't here to see your prince do his thing
Sometimes I wanna drop a tear, but no emotion from a king
by shake n bake on Jun 22, 2009 12:43 AM EDT up reply actions 0 recs
That sounds great...
The problem is that there are far too many variables that affect the statistics for them to accurately represent skill/ability. Statistics are representative only of a player’s past performance (as accurately as that performance can be recorded by those recording the numbers). The interpretation of what those numbers mean is subject to endless intellectual bias or “run-throughs” on behalf of those who are analyzing them to “prove” their theory.
In order to accurately analyze a WR, for example, you need to consider all of the following variables, which will be found inside the numbers of any statistic you could come up with:
Pass Accuracy – Weather – Field Conditions – Pass Coverage (individual defender and team defense) – Field Position – Injuries – Offensive Scheme – Other Receiving Options (pass-catching teammates) – Offensive Line Strength – Defensive Line Strength – Time of Possession – Game Situations (when was the reception made, under what game conditions) – Defensive Penalties (passing interference, holding, etc.) – Time/Placement/Impact of Tackle – etc.
1) Pass Accuracy – Any uncatchable pass, overthrow, underthrow, early throw, late throw, or pass errant in any way would have to be a part of the analysis. (not recorded)
2) Weather – The number of passes, receptions, and ability to perform as a receiver will obviously be affected by the weather. (functionally not recorded)
3) Field Conditions – Similar to weather, if you can’t run routes, if you trip on a choppy field, etc. it can affect your ability to perform. (not recorded)
4) Pass Coverage – Is your opponent one of the best pass defenders in the league, how often are you covered by high caliber pass defenders compared to other receivers, what kind of defense are you facing, does the defense prevent deep passes or gamble and give up deep passes, all of these things will affect a receivers ability to perform. (not recorded)
5) Field Position – How often were you on a long field compared to a short field. How often were you on the thin side of the field vs. the wide side of the field. (functionally not recorded)
6) Injuries – How often were you on the field and under what physical condition, were you hobbled, incapable of make cuts, incapable of catching passes you would otherwise have been able to reel in (functionally not recorded)
7) Offensive Scheme – Do you play in a the west coast offense, do you play in a vertical threat offense, does your offense key on the run or the pass, etc. (not recorded)
8) Other Receiver Options – Do you play next to other outstanding receiving options, like maybe a Reggie Wayne, Dallas Clark, Marvin Harrison, etc. (do you play on the same team with all of them, 1 or 2 of them), how do those receiving options impact your role in the offense (position, routes, etc.). (functionally not recorded)
9) Offensive Line Strength – How much time does your offensive line give your quarterback to look at the field and give you to find a way to get open, does the quarterback have to regularly rush throws, do you regularly have to make a play on balls arriving too soon/late in your route, etc. (functionally not recorded)
10) Defensive Line Strength – Conversely related to 9. (functionally not recorded)
11) Time of Possession – Does your team rarely have the ball or hold the ball for most of its games, how does this affect your role or the offensive play-calling. (functionally not recorded)
12) Game Situations – Did you miss a catch that was a hail mary at the end of a ball game which was a free for all for the players waiting in the end zone, did you make a clutch catch when time was about to expire to win the game, etc. (functionally not recorded)
13) Defensive penalties – How many passes did you “drop” or miss due to penalties, called or uncalled, like being held or interfered with, due to incidental contact, maybe a bad pick or broken route. (functionally not recorded)
14) Time/Placement/Impact of Tackle – When you made a play on the ball, when did you get contacted by the defender, where were you hit, how hard, in what position was your body, how long had the ball been secured in your hands (if at all) before the blow was delivered. (not recorded)
I will stop with this list and simply say I have no doubt that I could make it longer. No stat, not mgrex’s stats, yours, or anyone else’s will ever be able to take all of this into account. If they could, then maybe we would be close to having a statistical representation of a player’s actual ability… and only then they would still be subject to individual interpretation of the numbers.
Oh but keep in mind, each variable you add to your statistical representation/analysis will also be subject to individual interpretation of the data. Accordingly, even if you could come up with the list that included all of the variables I listed, and the rest of the variables that should be included, it would be such a muddled omelet of individual interpretation, bias, and opinion that it would be entirely useless as a representation for anything. Doesn’t that suck? I think it sucks, the world would be simpler if this weren’t true and everyone could be a football expert just by studying statistics. Oh well, I guess we’re just going to have to figure out a way to know a good football player when we see one.
Without the use of statistics at all consider my following opinion based on watching every game Anthony Gonzalez has played in for the Colts:
Solid hands (susceptible to the occasional drop in traffic or when he tries to run before the catch)
Excellent feet (does a great job of keeping his balance and making cuts when he needs to)
Solid field awareness (excels more at knowing where the sidelines are and could use some improvement on where the first down marker is)
Excellent quickness (makes most of his runs after the catch due more to his initial quickness than to his blazing timed speed)
Solid speed (could be really fast but focuses more on fundamentals, making the catch, running a solid route, most of the time)
Excellent mind (very smart on and off the field)
Excellent conditioning (goes above and beyond to stay in peak physical condition)
Average blocking (not much of a blocker down-field but will always make an effort)
Entering his first role as a unquestioned starting wideout, in his third season with Peyton Manning (2nd off-season of work with Manning), Manning throws for 4,000 yards + almost every year.
Assessment – Someone has to catch Peyton Manning’s passes this year with Harrison gone. Gonzalez will get the opportunity to fill that position and has the hands, route running ability, speed, quickness, and work ethic which suggest he will do well. With Peyton back to full health it makes sense for Gonzalez to see 20-40 more passes this year than he did last year, and probably an increase in another 300-500 receiving yards. My bet is that he catches 85-95 passes and eclipses 1,000 yards for the year with five or more TDs (so long as he stays healthy).
This is not based on “stats” so much as it is me watching the games and reaching my own conclusions about Gonzalez based on my “football eye.” Fallible? Sure, all opinions/assessments are. More fallible than a statistical interpretation? I don’t think so. You might. That’s fine, if numbers make you feel better about your opinions, go for it. To me, they’re largely questionable as tools to assess player ability or to even predict how a player will produce in the future.
In short, number are RAW, any additional weight given to them is created by the person analyzing them.
by bamock on Jun 22, 2009 9:36 AM EDT up reply actions 0 recs
that's a long way of repeaing what I said in my last paragraph
You do have to make subjective judgments, but I think it’s a lot better to make them once by having a good statistical base for what happened, then accounting subjectively for the things you know that stat doesn’t consider over can’t disentangle.
Change these hundreds for me cashier, Cuz I ain't made it yet, but I'm better off than last year
And what it look like hun', I ain't never made it rain but it look like fun
-Drake, Still Drake
by shake n bake on Jun 22, 2009 11:58 AM EDT up reply actions 0 recs
This is a great article
But I’m curious, where do you find the Air Yards and Catch%?
by njmetfan12 on Jun 22, 2009 2:31 AM EDT reply actions 0 recs
Football Outsiders has target data (receptions/targets=catch%)
and Yahoo had YAC per reception (YPR – YAC per reception= air yards per reception)
Change these hundreds for me cashier, Cuz I ain't made it yet, but I'm better off than last year
And what it look like hun', I ain't never made it rain but it look like fun
-Drake, Still Drake
by shake n bake on Jun 22, 2009 12:02 PM EDT up reply actions 0 recs
That was nicely said
Plus you laid out a great retirement plan for me. If I can somehow retire 20 years early and convince my wife to garden 2,400 hours a year. Might be tough in the winter…. when I am most “needed” at the DVR.
I hate Joe Namath. That's how long I've been a Colts fan.
by Bobman on Jun 22, 2009 3:15 AM EDT reply actions 0 recs
Great job
Do you think that it would be useful to utilize AirYPR to evaluate a player’s YAC? It seems that would help separate the talent level(at least as much as it is relevant to their position) of receivers that are used in any of different ways. Also, is there any place to get Success rates for receivers? It seems that this might be one place that Gonzo might fall inline with the average – as he seems at times to misread distance when he is in space, though he does move to gain the yards needed when he runs the appropriate route.
Also, what do we call this stat, CPOA?
One last note, it would be great if we could use a receiver’s QB’s VOA when he targets his other receivers and at least include that into a statistical analysis of a receiver.
I also wish I had time and motivation to blog at Speed Blue Nation
by Bullard47 on Jun 22, 2009 4:13 AM EDT reply actions 0 recs
Oh
In the first part of above I meant to create a YACOA using a formula similar to CPOA.
I also wish I had time and motivation to blog at Speed Blue Nation
by Bullard47 on Jun 22, 2009 4:14 AM EDT up reply actions 0 recs
I've got that data to look at YAC vs AirYPR
I’ll see if it correlates. Though I’m not optimistic since it won’t distinguish between a 2 yard screen pass, which is designed for piles of YAC and a 2 yard quick slant which is usually just a grab for those 2 yards.
Change these hundreds for me cashier, Cuz I ain't made it yet, but I'm better off than last year
And what it look like hun', I ain't never made it rain but it look like fun
-Drake, Still Drake
by shake n bake on Jun 22, 2009 12:07 PM EDT up reply actions 0 recs
with a polynomial equation
it borders on a small correlation .093

I’m planning on adding more data so I’ll keep an eye on it. The parabola trendline does make some sense. The two most likely types of passes to gain YAC in my mind are screen/dumpoffs and deep balls, while the middle stuff is usually being caught in the middle of the D.
Change these hundreds for me cashier, Cuz I ain't made it yet, but I'm better off than last year
And what it look like hun', I ain't never made it rain but it look like fun
-Drake, Still Drake
by shake n bake on Jun 22, 2009 12:26 PM EDT up reply actions 0 recs

by 






















