I should be posting the Trenches results from the five week-three games later today. Starting with week four, all games will be reviewed with what could be called Blocker Rating 2.0. I hadn't realized how close my "Simple Blocker Rating" was to the system used by NFL teams until I read a couple of recent articles at SI.com by former NFL lineman Ross Tucker. The way I'm rating the linemen now, however, is closer to the "Quality Rating" system I mentioned in an earlier post, which I believe will lead to even more accurate ratings. I'm considering referring to my stat from week four on as Blocker Effectiveness Rating. Not only would that differentiate it a bit from the rating used in the first three weeks, but it would emphasize that the rating system is privileging effectiveness over technique (Dr. Z talks about technique quite a bit when discussing his linemen ratings though he's never forthcoming with his exact "forumla.")
The first of Tucker's two insightful articles mentioned above is on his perceived lack of differences between elite and average linemen (hat tip: FO). On a related note, his other article ranks the offensive line positions by difficulty (hat tip: FO commenter mm). I believe Tucker overstates the amount of help given to right tackles as opposed to left tackles. In the brief time I was tracking double teams, left tackles did receive help on more plays than right tackles, but it wasn't a significant difference. As I have kept an eye on the situation in reviewing games since, I haven't noticed anything to change my mind. If anyone knows of any stats on this, please clue me in.
As I've alluded to in previous posts, centers and left guards definitely are involved in more double teams than anyone else on the line, often they're working together. Tucker definitely knows what he's talking about, but again I think he's overstating the issue to help make his case (and it is a tough case to rank the interior line positions by difficulty, since the differences in duties on any given play are so nuanced). The fact that the centers are double teaming on most plays can be a bit deceptive. Typically the center (or whichever other lineman he's working with) has to release from the double team and get to the second level. The ability to do this effectively is one of the main things that separates elite centers (e.g., Tom Nalen) from the rest. My rating system will ding the center if he is ineffective at either level. In pass protection, the center is seldom one-on-one, but on most passing plays it's he, more than anyone else, who must keep his head on the proverbial swivel. The best centers flow to where they're needed: out of the corner of their eye, they notice one of their guards (or, sometimes, one of the tackles) getting beat, and they shoot over there to help out, even if it's sometimes just to give the defensive linemen an all-important shove, thereby extending the play for the extra second it needs to have a chance.
If I was a GM divvying up my salary amongst the linemen from highest to lowest paid, my priorities would mirror Tucker's list pretty closely. I'd go left tackle, right tackle, then one of the guards (I'm not ready, yet, to take a stand on the left guard/right guard argument), center, then the other guard.
In his first article, Tucker brings up the issue of how many plays per game an elite lineman actually affects as opposed to an average lineman. He says "about one to two plays per game." My initial reaction was that it would be more like five to ten plays per game, but it's probably about five and not so high as ten. Anyway, Tucker has prompted me to run the averages by position earlier than I had expected to and I hope to look at those later today and see what they can tell us regarding effectiveness on a plays per game basis.
Although I disagreed with some of his points, Tucker's article is really great and has me more confident than ever concerning the efficacy of my rating system. I'm also thrilled that I decided to "try it out" on the 2007 Replay games because I've been able to iron out some of the wrinkles. I'm sure that when 2008 comes around and I have complete games to review and rate line play, I'll have a really good and consistent system to do that with. The only thing that can stop this project is the very really prospect of not having enough time to focus on it. We'll see what happens.
Showing posts with label offensive linemen. Show all posts
Showing posts with label offensive linemen. Show all posts
Sunday, April 06, 2008
Monday, March 31, 2008
In the Trenches: An Intro
For the past four Sundays the NFL Network has been "replaying" five games from each week of the 2007 season. I've decided to go back and look at the offensive line play in all of these games using the Blocker Rating system I introduced following the Super Bowl.
I hope that by charting the 2007 games I can work some of the kinks out of the system, and be more prepared to chart games and post about them throughout the 2008 season. I've made some slight modifications to my rating system, and really it's still in flux. I thought about a major overhaul, or possibly doing two ratings: an Effectiveness Rating, which would basically be the simple Blocker Rating I discussed in the Super Bowl post; and a Quality Rating, which would better account for "dominating" blocks that eliminate a defender from a play no matter what might happen, as well as successful blocks of more than one defender on a play, while discounting blocks that were part of a double team. But after charting one half of a game using the Quality Rating system, I could see that it would be so time consuming that there would be no way I could manage to keep at it over the course of a season. Instead, I'm experimenting with ways to account for some of this info in my simple rating system, without adding too much time to the process. And besides, I'm not entirely convinced that the Quality Rating would have yielded significantly better results.
Here are some of the key points to know about this rating system:
1. 10 is the highest rating, whether for a game or a season (in effect, a 10 is a "perfect game"). The early returns suggest that perfect pass blocking games will not be rare ("uncommon" would seem to be more accurate), but perfect run blocking games will be scarce.
2. Once I have enough data I will post the averages of each position to provide a context in which a player's perfomace/rating can more easily be judged. Eventually this may lead to a "value over average" rating.
3. Blocker Rating's main limitation is that it doesn't account for how much a lineman might be asked to do. For instance, a left tackle who effectively blocks, say, Dwight Freeney one-on-one on a pass play will score the same as the left guard and center who effectively double team Ed Johnson. Similarly, a right guard who whiffs on his pull block, allowing Gary Brackett to stuff his running back for no gain, will score as low as the right tackle whose backside cut block fails to take out Robert Mathis regardless of whether Mathis is able to assist on the tackle or not (the assumption being that anything can happen on a given play, and hence, every defender who can be taken out of the play should be in case, for example, the running back needs to cut back or a fumble needs to be recovered).
4. Something of a minor point, but in case someone wonders about this: run blocking and pass blocking are defined by whether a play is a run or a pass. That might seem obvious, but it could potentially be defined otherwise. For instance, the blocking on a play-action pass calls for run-blocking techniques, while a draw play typically has the linemen feigning pass protection. Screen passes are almost their own animal, calling for some linemen to pass block, while others only initially fake that before slipping out to get ahead of the receiver, calling for skills closer to those used in the running game.
5. The ratings are not defense adjusted. That is, there's no adjustment whether a guard is facing a tough task like, say, Albert Haynesworth, or if he's battling Claude Wroten all day. (I am going to collect some opponent data, and by the end of the season I should be able to comment on the results.)
6. After charting a couple of week one games I changed it so that passing plays of 2 or fewer seconds are no longer included. Aborted plays are still not counted and neither are false starts, but plays with offensive holding are now charted, as are any plays where the penalty is on the defense (although if an offensive lineman is beat because the defensive man was offsides, it's not counted against him).
7. Games charted using NFL Replay are a necessity as long as I have a day job to attend to. The final Blocker Ratings from these games are not "official" in that these broadcasts don't show every play. However, enough plays are shown to get a representative sample of a player's performance, especially as the total plays accumulate over the course of a season.
8. I can't reiterate enough: This is a very simple rating system, so that I actually have enough time to chart as many games as I can record. Although quite a bit is left unaccounted for, I've seen enough already to believe that my Blocker Rating gives a good indication of how well, or poorly, a linemen has played. If I continue to be pleased with the results, I might eventually recruit other volunteers to chart games so that only full games are charted and so that a season can be completely captured. But I won't be doing that until I'm fully comfortable with this and have had more time to tweak the rating system.
There are probably other key points escaping me right now. I'll update this post as they occur to me.
I hope that by charting the 2007 games I can work some of the kinks out of the system, and be more prepared to chart games and post about them throughout the 2008 season. I've made some slight modifications to my rating system, and really it's still in flux. I thought about a major overhaul, or possibly doing two ratings: an Effectiveness Rating, which would basically be the simple Blocker Rating I discussed in the Super Bowl post; and a Quality Rating, which would better account for "dominating" blocks that eliminate a defender from a play no matter what might happen, as well as successful blocks of more than one defender on a play, while discounting blocks that were part of a double team. But after charting one half of a game using the Quality Rating system, I could see that it would be so time consuming that there would be no way I could manage to keep at it over the course of a season. Instead, I'm experimenting with ways to account for some of this info in my simple rating system, without adding too much time to the process. And besides, I'm not entirely convinced that the Quality Rating would have yielded significantly better results.
Here are some of the key points to know about this rating system:
1. 10 is the highest rating, whether for a game or a season (in effect, a 10 is a "perfect game"). The early returns suggest that perfect pass blocking games will not be rare ("uncommon" would seem to be more accurate), but perfect run blocking games will be scarce.
2. Once I have enough data I will post the averages of each position to provide a context in which a player's perfomace/rating can more easily be judged. Eventually this may lead to a "value over average" rating.
3. Blocker Rating's main limitation is that it doesn't account for how much a lineman might be asked to do. For instance, a left tackle who effectively blocks, say, Dwight Freeney one-on-one on a pass play will score the same as the left guard and center who effectively double team Ed Johnson. Similarly, a right guard who whiffs on his pull block, allowing Gary Brackett to stuff his running back for no gain, will score as low as the right tackle whose backside cut block fails to take out Robert Mathis regardless of whether Mathis is able to assist on the tackle or not (the assumption being that anything can happen on a given play, and hence, every defender who can be taken out of the play should be in case, for example, the running back needs to cut back or a fumble needs to be recovered).
4. Something of a minor point, but in case someone wonders about this: run blocking and pass blocking are defined by whether a play is a run or a pass. That might seem obvious, but it could potentially be defined otherwise. For instance, the blocking on a play-action pass calls for run-blocking techniques, while a draw play typically has the linemen feigning pass protection. Screen passes are almost their own animal, calling for some linemen to pass block, while others only initially fake that before slipping out to get ahead of the receiver, calling for skills closer to those used in the running game.
5. The ratings are not defense adjusted. That is, there's no adjustment whether a guard is facing a tough task like, say, Albert Haynesworth, or if he's battling Claude Wroten all day. (I am going to collect some opponent data, and by the end of the season I should be able to comment on the results.)
6. After charting a couple of week one games I changed it so that passing plays of 2 or fewer seconds are no longer included. Aborted plays are still not counted and neither are false starts, but plays with offensive holding are now charted, as are any plays where the penalty is on the defense (although if an offensive lineman is beat because the defensive man was offsides, it's not counted against him).
7. Games charted using NFL Replay are a necessity as long as I have a day job to attend to. The final Blocker Ratings from these games are not "official" in that these broadcasts don't show every play. However, enough plays are shown to get a representative sample of a player's performance, especially as the total plays accumulate over the course of a season.
8. I can't reiterate enough: This is a very simple rating system, so that I actually have enough time to chart as many games as I can record. Although quite a bit is left unaccounted for, I've seen enough already to believe that my Blocker Rating gives a good indication of how well, or poorly, a linemen has played. If I continue to be pleased with the results, I might eventually recruit other volunteers to chart games so that only full games are charted and so that a season can be completely captured. But I won't be doing that until I'm fully comfortable with this and have had more time to tweak the rating system.
There are probably other key points escaping me right now. I'll update this post as they occur to me.
Sunday, February 10, 2008
Super Bowl 42: Of the MVP and Blocker Ratings
When a quarterback engineers a dramatic Super Bowl-winning drive there's a pretty good chance he'll be the game's MVP. Eli Manning's 55% completion percentage is nothing to get excited about, but his 7.5 yards per attempt is better than average. He accounted for two touchdowns against only one interception, which wasn't even his fault (although he did throw at least a couple of other passes that could have been intercepted and would have been his fault).
But Manning wouldn't have gotten my MVP vote if I had one. The difference in the game wasn't the Giants offense which scored 17 points, about what many expected they would, but the defense that held the Patriots offense, in perfect conditions, to only 14 points, about twenty fewer than many expected.
The biggest reason for the Giants defensive success was the pressure they put on Tom Brady. New York's five sacks came from four different players and there were others, from Osi Umeniora to Aaron Ross, involved in the hurries and hits. Justin Tuck led the way with his two sacks and a forced fumble. He also contributed five solo tackles and an assist. While that might not be a huge game exactly, it was the defense's stand out performance, and I would have voted for Tuck as a way to recognize what the Giant defense accomplished.
The Giants seemed to dominate the Pats offensive line and it got me thinking about a blocker rating system I kicked around in my head for a few months during the season. I never got around to field testing that idea, but now, with a Super Bowl that in many ways came down to line play, I've set down some parameters and charted my first game using my Simple Blocker Rating system (SBR).
I'm sure there are others, but off the top of my head the only offensive line ratings (player-specific, not as a unit like Football Outsiders' Adjusted Line Yards) I can think of that fans have at least some access to are: Dr. Z's mysterious method which he uses in determining his All-Pro team each year; Yakuza Rich's statistical ranking based on yardage; the percentage-based “grading out” popular with college sites.
SBR is based on success and not excellence. A cut block on the backside of a run that may or may not have been enough if the run had broken differently is weighted the same as a block by a pulling guard that pancakes his man and opens a hole for the running back. This is one of the things that makes this rating “simple.” But I have chosen to keep it this way in order to limit the value judgments needed on my part. There are two main benefits for this: first, it makes the rating more objective by lessening the need to interpret responsibility or pinpoint the exact effectiveness of a block; second, it cuts down the time required to chart a game, which means this is more likely something I can continue to do.
Again, SBR doesn't consider technique, only success. In determining success it errs on the side of giving credit to the blocker. As such SBR is almost entirely a descriptive stat; it's probably limited in its ability to predict future results. Just as a quarterback's passing stats are affected by the quality of his receivers, his blockers, and even the threat posed by his running back, a lineman's blocker rating will be affected by things outside of his control such as the kinds of plays called (i.e, a lineman will rate better if his team calls a lot of wide receiver screens and quick slants) and the quality of the defender he's working against,
Since this is the first and only game I've charted using SBR there's no context for the numbers. I expect that over the course of a season certain positions, such as left tackle, would have lower numbers since they're often going up against the best pass rushers and are less likely than their linemates to be part of a double team.
Finally, before presenting the table, a few notes and observations. 10 is a perfect score. I charted tight ends and fullbacks but not running backs or wide receivers. All-Pro guard Logan Mankins struggled as much with Tuck, et al., as it seemed he did. Pats backup guard Russ Hochstein wasn't quite as bad as I thought while watching the game, but he was a huge drop off from Stephen Neal who was pitching a shutout. New York's Shaun O'Hara had been perfect in pass blocking before struggling a bit on the last drive. Similarly David Diehl and Rich Seubert struggled in pass protection late in the game, but played solidly prior to that. The Giants' tight ends were much better, with Kyle Brady in particular hurting the Pats in the running game. As might be expected, the Pats kept Ben Watson in to pass protect much more later in the game. NY Fullback Madison Hedgecock was less effective than I would have guessed.
But Manning wouldn't have gotten my MVP vote if I had one. The difference in the game wasn't the Giants offense which scored 17 points, about what many expected they would, but the defense that held the Patriots offense, in perfect conditions, to only 14 points, about twenty fewer than many expected.
The biggest reason for the Giants defensive success was the pressure they put on Tom Brady. New York's five sacks came from four different players and there were others, from Osi Umeniora to Aaron Ross, involved in the hurries and hits. Justin Tuck led the way with his two sacks and a forced fumble. He also contributed five solo tackles and an assist. While that might not be a huge game exactly, it was the defense's stand out performance, and I would have voted for Tuck as a way to recognize what the Giant defense accomplished.
The Giants seemed to dominate the Pats offensive line and it got me thinking about a blocker rating system I kicked around in my head for a few months during the season. I never got around to field testing that idea, but now, with a Super Bowl that in many ways came down to line play, I've set down some parameters and charted my first game using my Simple Blocker Rating system (SBR).
I'm sure there are others, but off the top of my head the only offensive line ratings (player-specific, not as a unit like Football Outsiders' Adjusted Line Yards) I can think of that fans have at least some access to are: Dr. Z's mysterious method which he uses in determining his All-Pro team each year; Yakuza Rich's statistical ranking based on yardage; the percentage-based “grading out” popular with college sites.
SBR is based on success and not excellence. A cut block on the backside of a run that may or may not have been enough if the run had broken differently is weighted the same as a block by a pulling guard that pancakes his man and opens a hole for the running back. This is one of the things that makes this rating “simple.” But I have chosen to keep it this way in order to limit the value judgments needed on my part. There are two main benefits for this: first, it makes the rating more objective by lessening the need to interpret responsibility or pinpoint the exact effectiveness of a block; second, it cuts down the time required to chart a game, which means this is more likely something I can continue to do.
Again, SBR doesn't consider technique, only success. In determining success it errs on the side of giving credit to the blocker. As such SBR is almost entirely a descriptive stat; it's probably limited in its ability to predict future results. Just as a quarterback's passing stats are affected by the quality of his receivers, his blockers, and even the threat posed by his running back, a lineman's blocker rating will be affected by things outside of his control such as the kinds of plays called (i.e, a lineman will rate better if his team calls a lot of wide receiver screens and quick slants) and the quality of the defender he's working against,
Since this is the first and only game I've charted using SBR there's no context for the numbers. I expect that over the course of a season certain positions, such as left tackle, would have lower numbers since they're often going up against the best pass rushers and are less likely than their linemates to be part of a double team.
Finally, before presenting the table, a few notes and observations. 10 is a perfect score. I charted tight ends and fullbacks but not running backs or wide receivers. All-Pro guard Logan Mankins struggled as much with Tuck, et al., as it seemed he did. Pats backup guard Russ Hochstein wasn't quite as bad as I thought while watching the game, but he was a huge drop off from Stephen Neal who was pitching a shutout. New York's Shaun O'Hara had been perfect in pass blocking before struggling a bit on the last drive. Similarly David Diehl and Rich Seubert struggled in pass protection late in the game, but played solidly prior to that. The Giants' tight ends were much better, with Kyle Brady in particular hurting the Pats in the running game. As might be expected, the Pats kept Ben Watson in to pass protect much more later in the game. NY Fullback Madison Hedgecock was less effective than I would have guessed.
| Name | Team | Pos | SBR | Run | Pass | Plays |
| David Diehl | NYG | LT | 8.19 | 7.82 | 8.42 | 61 |
| Matt Light | NE | LT | 8.55 | 6.66 | 9.07 | 69 |
| Rich Seubert | NYG | LG | 8.85 | 8.69 | 8.94 | 61 |
| Logan Mankins | NE | LG | 8.55 | 8.66 | 8.51 | 69 |
| Shaun O'Hara | NYG | C | 8.85 | 8.26 | 9.21 | 61 |
| Dan Koppen | NE | C | 9.27 | 8.66 | 9.44 | 69 |
| Chris Snee | NYG | RG | 9.18 | 9.13 | 9.21 | 61 |
| Stephen Neal | NE | RG | 10.00 | 10.00 | 10.00 | 19 |
| Russ Hochstein | NE | RG | 8.80 | 10.00 | 8.60 | 50 |
| Kareem McKenzie | NYG | RT | 9.50 | 9.13 | 9.73 | 61 |
| Nick Kaczur | NE | RT | 8.98 | 9.33 | 8.88 | 69 |
| Kevin Boss | NYG | TE | 9.62 | 9.50 | 10.00 | 27 |
| Ben Watson | NE | TE | 8.33 | 6.66 | 10.00 | 24 |
| Michael Matthews | NYG | TE | 9.16 | 9.09 | 10.00 | 13 |
| Kyle Brady | NE | TE | 6.19 | 5.33 | 8.33 | 21 |
| Mike Vrabel | NE | TE | 10.00 | 10.00 | ---- | 3 |
| Madison Hedgecock | NYG | FB | 7.69 | 7.22 | 10.00 | 13 |
| Heath Evans | NE | FB | 5.00 | 5.00 | ---- | 4 |
Saturday, February 02, 2008
Round Up
In honor of Black History Month, pro-football-reference.com's Chase Stuart kicks off a series looking at the history of black quarterbacks in the NFL. It's a great first post on a topic that shouldn't be relegated to the dustbin of football history just yet. It's amazing how much has changed in the thirty years I've been watching.
Yakuza Rich plays a First Round Draft Game, and takes a quick look at the correlation between first round draft pick performance and team success. Pacifist Viking looks at the importance of the draft anecdotally by considering the downfall of the Cowboys in the 1990s and the sustained success of the 49ers through the 80s and 90s.
Also worth checking out at Yakuza Rich's blog are his offensive linemen rankings. I don't think the rankings equate to a listing of the best linemen of the year exactly, but it's certainly a measure of success and something to consider when discussing the best linemen in the game.
Matt Mosley on Roger Goodell, Senator Arlen Specter, and Spygate. I can't believe a Senator is actually wasting his time, and therefore our tax dollars, on this. I understand that because of the NFL's Antitrust exemption the government has an interest in the league, but this issue has been blown out of such proportion over the course of the season, I get queezy every time it's brought up.
NFL Stats asks if red zone performance is real?
NFL Draft Guys might be the only site covering the Texas vs the Nation all-star game this week, providing relatively detailed practice reports and complete, downloadable weigh-in information. I'll be watching the game this afternoon and writing about it, possibly later tonight if I can make time. The official site provides some info on players from last year's game who went on to be draft selections.
Michael Davis Smith questions Roger Goodell about the NFL's outrageous rookie salary structure and its fairness to proven veterans.
Yakuza Rich plays a First Round Draft Game, and takes a quick look at the correlation between first round draft pick performance and team success. Pacifist Viking looks at the importance of the draft anecdotally by considering the downfall of the Cowboys in the 1990s and the sustained success of the 49ers through the 80s and 90s.
Also worth checking out at Yakuza Rich's blog are his offensive linemen rankings. I don't think the rankings equate to a listing of the best linemen of the year exactly, but it's certainly a measure of success and something to consider when discussing the best linemen in the game.
Matt Mosley on Roger Goodell, Senator Arlen Specter, and Spygate. I can't believe a Senator is actually wasting his time, and therefore our tax dollars, on this. I understand that because of the NFL's Antitrust exemption the government has an interest in the league, but this issue has been blown out of such proportion over the course of the season, I get queezy every time it's brought up.
NFL Stats asks if red zone performance is real?
NFL Draft Guys might be the only site covering the Texas vs the Nation all-star game this week, providing relatively detailed practice reports and complete, downloadable weigh-in information. I'll be watching the game this afternoon and writing about it, possibly later tonight if I can make time. The official site provides some info on players from last year's game who went on to be draft selections.
Michael Davis Smith questions Roger Goodell about the NFL's outrageous rookie salary structure and its fairness to proven veterans.
Subscribe to:
Posts (Atom)
