This blog has been a little slow lately because of preparation for the 2009 season.  It will be more updated for 2009 soon.

I want to point out now an awesome application at Advanced NFL Stats for the NCAA tournament.  It shows a graph for each game with the current probability that either team will win.  It’s hard to describe, but you can check it out here: Live tournament win probability.

The team I picked to win it all, Memphis, almost blew it.  You can see in the time-series win probability that most of the game they had a below 50% chance of winning until the last few minutes.

Earlier in the NFL season, I thought that Drew Brees would break Dan Marino’s record, which obviously didn’t happen.  He came close though, and it came down to his last pass of the season, which fell incomplete.

Based on Brees’ strength of schedule, I predicted that he would finish with 5,111 yards.  He finished with 5,069, which makes the prediction off by only 42 yards, or 0.8%.  The prediction happened after he had already played 6 games, so it’s not too surprising that it was very close.

Even though he didn’t break the record, I still think it’s important to try to predict whether a record gets broken.  I might try to develop a probability for the prediciton, instead of just a yes/no prediction.

Since I had predicted he would get just barely above Marino’s record, I should have given him something like a 52% chance of breaking the record, instead of 100%.

We’ll have to wait until next year to see if anyone is close to breaking another record though.

Way back in October I predicted that Drew Brees would pass Dan Marino’s single-season passing record from 1984.

Although he’s slowed down in recent weeks, Brees is still on pace.  Through 14 games in 1984, Dan Marino had 4,340 yards. Drew Brees has 4,332 yards.

Only 8 yards behind.

Marino finished his season well with 404 and 340 yards in the final two games.  Below is a graph comparing the two players’ cumulative yardage totals over the course of their seasons:

You can see that Brees has just recently fallen to Marino’s pace while he has been solidly ahead of it for most of the year.

Back in October, I estimated Drew Brees would end up with 5,111 yards, just 27 yards ahead of the record.  He has an easy matchup this week with the Lions, but a tough one against the Panthers to finish the season.

It should be a close one.

Everyone knows that it’s important to protect the quarterback.  Teams spend millions of dollars on the best offensive linemen in hopes that they will give their quarterback enough time to tear up the opposing secondary.

Since passing is so important in the NFL today, how can we measure the effectiveness of protecting the quarterback?

Sacks and Win Percentage

As a simplistic measure, let’s use sacks as a measure of protecting the quarterback.  Below is a plot of win percentage for every quarterback in the league that’s played at least 4 games versus their average sacks per game:

There seems to be a clear trend that fewer sacks per game improves a team’s chances of winning.  The R-squared value for that trend is 0.29, which isn’t large, but is significant as far as football statistics go.

Sacks and QB-Rating

On the other hand, there is no correlation between sacks per game and quarterback rating:

On the extremes here are Kerry Collins with only 1 sack in 7 games with a 74.2 QB rating and J.T. O’Sullivan with a bone-crushing 29 sacks in 7 games and a 75.5 QB rating.

Using sacks per game may not be the best measure of how well an offensive line protects the quarterback.  I used it here because it’s the easiest to measure. A statistic that incorporates sacks, hurries, and knock-downs would be a better indicator of how well a quarterback is protected.

Quarterbacks Perform Independent of O-Line?

If J.T. O’Sullivan were traded to Tennesee and played behind their offensive line, would he be transformed into a superstar?  I don’t think so.  My guess is that his QB rating would have a slight but negligible increase.

This isn’t a conclusive study, but I think this indicates that a quarterback’s performance is almost entirely independent of their offensive line.  There are a lot of uncontrolled variables here such as the quality of the rest of the offense and the strength of schedule, but I’m not too surprised by this result.

O-Line and Running Backs

What about Frank Gore?  Would his yards per carry increase if he were traded to Tennesee?  I bet it would. Running backs seem to be more interchangeable than quarterbacks and depend more on their offensive line.

Sacks allowed is somehow related to win probability, but perhaps not through quarterback performance.  I’ll look into the effect of offensive line on the running game in a later post.  There is probably a reason that teams invest heavily in offensive lines, and there should be a way to measure how effective that strategy is.

(Follow-up 12/19: Drew Brees is still close to Marino’s record.)

In my last post, I outlined four players on track to break records this year.  In just two short weeks, however, that list was cut down to one:

Drew Brees

Brees is holding steady at 332 yards per game, which will give him 5,315 for the season if he can continue that pace.  But can he?

The Saints have had an easy schedule so far.  He’s already lit-up the league-worst Broncos for 421 yards, and  the toughest pass defense he’s seen so far was Washington, the 13th best in the league.

Adjusting for Strength of Schedule

The graph below shows Brees’ passing yards versus his opponent’s defensive passing rank.  The blue dots show his actual performance in the past six games and the red dots show his predicted performance for the next 10 games.

As you can see, not only has he faced a lot of bad defenses, he’s scheduled to play even more bad defenses including San Diego (31), Detroit (30), and Atlanta twice (24).  He only has three games against above average defenses: Carolina twice (2) and Green Bay (7).

Adjusting his average yards per game for opponent difficulty, he is projected to finish the year with 5,111 yards.  That’s a hair more than Marino’s 5,084 yards in 1984.

Conclusion

Drew Brees has played a cream-puff schedule so far and has a cream-puff schedule ahead of him.

After adjusting for strength of schedule, he is on pace to break Dan Marino’s single-season passing record, but just barely.

Don’t expect a big game from him this week against Carolina, but he will stay within reach of Marino’s record down the stretch.

It’s difficult to know when to sell high on a player. They may continue to perform well or they may just be on a hot streak.

However, when a player is on pace to break an NFL record, that might signal that they will drop off soon.  So far this season, there are four quarterbacks and receivers on track to break single-season records.

The fact that no running backs are at record-setting pace indicates that the league is continuing its new passing trend that I wrote about earlier.

Here are the outstanding passers and receivers so far. The projections below are their per game averages multiplied by 16 games:

Passing Yards

NFL Record: 5,084 (Dan Marino)

Both Brees and Cutler have looked good so far, but there is a good chance they will slow down to less Marino-like paces.

Receiving Yards

NFL Record: 1,848 (Jerry Rice)

Obviously Marshall’s performance is closely tied to Cutler’s above, and they’re both on track for big seasons. However, just one or two bad games will throw them off this pace.

No Running Backs

It’s not surprising that after only four weeks some players are above record-setting pace. What is surprising though is that they are all in the passing game. Running backs are completely absent.

The leading running back in the league is Michael Turner, who has 422 yards through 4 games. That’s only on pace for 1,688 yards, which is well off the record of 2,105 yards.

Sell High or Let It Ride?

Is it time to sell high on Brees, Cutler, Marshall, and Jennings? Or do some of them have a legitimate chance at maintaining their pace?

Perhaps Jay Cutler and Brandon Marshall can knock off the passing and receiving yardage records similar to how Brady and Moss broke the single-season touchdown records last year.

However, if you think they are more likely to slow down, then now is the time to trade them away.

Tom Brady’s injury is devastating to the teams that drafted him. The effect is a net loss of 14 to 20 fantasy points per week.

That number was calculated by looking at the predicted fantasy points per game for quarterbacks before the Brady injury:

Tom Brady was expected to produce about 26 fantasy points per game; a lot more than any other quarterback.

Decent Backup

The effect of his injury depends on who you have to replace him. If you have a decent replacement or were able to pick up Matt Cassel, then your new QB will probably average 12 fantasy points per game. It’s still early to tell what Cassel will do, but with a great team surrounding him he’ll still be a solid replacement.

In this scenario with a decent backup, the net loss is 14 points per game due to Tom Brady’s injury. That’s a lot.

Waiver Wire

Unfortunately, that’s the best-case scenario for most fantasy owners. If you weren’t fast enough to grab Cassel and you didn’t draft a good backup then you’re looking at pulling someone from the free agent pool.

In general, there will be better players on the waiver wire in smaller leagues than in larger ones. Assuming there were two QBs drafted per team, a 10-team league will have the 20th-best QB available, such as Jon Kitna.

If you can grab the 20th-best QB, then your net loss from Brady’s injury is 17 points per week.

For a 12-team league the 24th-best QB, Matt Ryan, should be available and the net loss is 19 points per week.

For a 14-team league the 28th-best QB, Jamarcus Russell, should be available and the net loss is 20 points per week.

Ouch.

Trade! Trade! Trade!

If you think you can spot every team in your league 14 to 20 points per week and still make the playoffs, you’re dreaming.

To have any shot at winning, you need to get more points per game into your starting lineup. That means taking a risk and sacrificing depth at other positions. Package up your backup running backs and wide receivers to upgrade at quarterback.

The only exception is if you got lucky and had an amazing draft and the rest of your team can make up for the loss at quarterback. If you drafted Michael Turner and Willie Parker, who are both outperforming their ADP, then you can probably limp along with a scrub at QB.

All the other unlucky Brady owners, however, need to start looking for teams to trade with. It’s going to take a lot of work to get back into playoff contention.

One of the most valuable articles in Pro Football Prospectus 2008 describes a new way to predict the future performance of rookie running backs.

In “Five Seconds Can Be a Lifetime,” Bill Barnwell describes a simple equation that combines a rookie’s weight and 40-yard dash time into one number that is better than any other number in predicting if that running back will be a success in the NFL.

He calls that number the Speed Score and calculates it with this equation:

The faster a player’s time, the higher the Speed Score. Also, if two players run the exact same 40 time, the one that is heavier gets a higher score because he had to run the time with more weight on his body.

When Barnwell looked at past rookies and how they later performed, he found that many successes had a speed score greater than 100 and many busts had speed scores less than 100.

I don’t want to give away too much, but when you look at the list of players it picked to succeed versus those it didn’t is pretty impressive. It is slightly biased because a 40-yard dash time that isn’t weight adjusted is already a good indicator of success. However, their statistics show that the speed score is slightly better than the 40-time alone.

Can you guess which rookie this year posted the highest speed score at the combine? You’ll have to buy the book to find out, but I will say that it wasn’t McFadden.

I mentioned last week that I’m a fan of Pro Football Prospectus, and this year’s edition is definitely worthwhile.

One article that I find interesting is the study of how a wide receiver’s height and BMI can be an indicator of success.

The way they predict success is by determining whether or not a player falls into one of four sections on a chart of height versus BMI. The study pooled data from the past ten years and compared players by looking at average yards per game.

Here is a rough sketch of their complete graph with some representative players shown:

The theory is that if a player falls into one of the four boxes, they have a decent chance of being a top wide receiver. Conversely, if a player doesn’t have a body type that fits into one of the four boxes, then they have only a small chance of being an elite receiver.

If you’re wondering how big these relative differences in body type are, below I added myself for comparison. At 6′ 2″ and 170 lbs, I fall just a bit on the light side of a successful NFL wide receiver:

This article is by far the most extensive study I’ve seen on how well body type can predict success. I am a little hesitant at the moment to take the results as fact though.

The main reason is that I don’t think there is enough data to justify drawing four separate zones on this graph. The theory is probably overfitting the data. For example, the height difference between the top and bottom zones is only two inches.

Second, I’m not convinced there is a legitimate football-related reason why these narrow zones of body type would be successful and others wouldn’t. Especially that dead zone in the middle. Is there a reason why a player with that height/BMI combination is less likely to be successful?

The article suggests that the reason for these zones is that they identify which body types are able to get separation from cornerbacks and which are useful as large targets for a quarterback. They also propose that looking at the body types of defensive backs could help determine if these wide receiver body types create favorable mismatches.

Although I’m skeptical, I’m always on the lookout for hidden trends and new ways to predict successful players. I’m going to watch this one for the next few years to see if these body type trends continue.

If a new successful player emerges with a body type that is right in the middle of those four zones, I think this theory is busted. If not, and the trends continue, this could be a real way to predict success within a group of players that is usually just a crap shoot.

I received my copy of Pro Football Prospectus 2008 this week, and I’m excited.  I’ve read it for the past three years and it’s packed with lots of great analysis.  Being a numbers guy, this book hits the sweet spot for me.

If you’re not familiar with it, half the book is analysis and projections for NFL teams, and then the other half is projections for individual players.  They also compile projected fantasy points for each player and rank them by VBD at the end of the book.

The player projections are good because in general they are modest for almost everyone.  It helps me keep a level head when I’m really positive about a player and might draft him too early.  Their trends usually show players that have great years regress to the mean the next.

One player that they project to do well is Aaron Rodgers (if he gets a chance to play that is).  However,  I still remember a few years ago when they projected Kevin Jones and Julius Jones to be #1 and #2 in rushing yards.   Whoops!  I’ll let that one slide even though I took Kevin Jones in most of my leagues that year because of their advice.

Despite the occasional bad projection, it’s still a good book for fantasy football fans and especially those who like to analyze statistics.  I’ll try to write more about the book once I’ve finished it.  At over 500 pages, I know there’s plenty of material to cover.