In 2006 (Michael Vick’s) last season he had an NFL QB rating of 75.7, which is way below average. So was Vick a below average QB in 2006? Looking at our new method of evaluating players, we determined how many points were generated on each passing play involving Vck in 2006. We found his passing plays generated 7 points per 100 plays which is far below the league average of 17 points per 100 passing plays. This ranked Vick 27th among 36 QB’s who were involved in at least 200 passing plays.

    The key to Michael Vick, however, is his great running ability. On 122 rushes in 2006 Vick generated an amazing 60 points per 100 rushes. Among all rushers who rushed at least 100 times this was by far the best performance in the league(including running backs!). When we factor in a QB’s rushing attempts we see that  Vick created 18 points per 100 plays which paced him 13th out of 36 QB’S.   Our new analysis indicates that in 2006 Vick was an above average QB.

This is another great example of how traditional NFL stats are incredibly misleading .

For more details of how to use nontraditional stats to evaluate football, baseball and basketball players pick up a copy of my new book Mathletics.

The NFL’s current system for rating QB’s is a complex. nonsensical formula that combines a QB’s completion percentage, Yard per Pass Attempt, TD Pass Percentage, and Interception Percentage into a single incomphrensible number.

    Wouldn’t it be nice (as the Beach Boys said) if we could assign a point value to each pass thrown by a QB and then evaluate QB’s based on points generated per 100 pass  attempts?.  Using a technique described in my new book Mathletics, for the NFL 2008 season here are the top 5 QB’s (numbers adjusted based on strength of opponents passing defenses:

1. Peyton Manning 37 points.
2. Drew Brees 36 points.
3. Matt Ryan 34 points
4. Phillip Rivers 32 points.
5. Kurt Warner 29 points,

Of other interest, Brett Favre only generated 10 points per 100 pass attempts with the Jets in 2008. If the Vikings knew this they might have thought twice about signing him. Thus we see that if Peyton and Brett each threw 500 passes Peyton would generate 135 more points than Brett.

Of course, when we evaluate a QB we are evaluating him based on his set of receivers and offensive line.  Properly partitioning the credit for a pass play between a QB, receiver and offensive line is a difficult, if not impossible task. But at least we can evaluate a team’s passing game with a single meaningful and easy to understand metric.

In our next post we will rank the NFL’s top running backs during the 2008 season.

For over 40 years baseball experts have realized that when an average batter is up, bunting is a bad idea. Yet at the Indianapolis Indian AAA game which I attended last night (had a great time by the way) 4 bunts were attempted! Several of these bunts involved the pitcher (probably a poor hitter) but to enlighten those of you who do not know why bunting is usually a bad idea, let’s go through the simple logic.

   Suppose there is a runner on first on none out. How many runs does an average MLB team score in an inning is this situation? The answer is 0.93 runs. Now let’s bunt and suppose the bunt succeeds. Now we have a runner on second base and one out. How many runs does an average major league team score in an inning with this situation? The answer is 0.71 runs. Therefore the “success” of the bunt has cost our team ) 0.22 runs. This should make it clear why the bunt is usually a bad idea.

   Of course, if the batter is a poor hitter (like a pitcher) the bunt might make sense or if the score is tied and our goal is to simply score a single run the bunt might be a good idea. For more discussion of baseball decision-making see chapter six of my new book Mathletics, which will be released by end of  August.

As described in our August 16 post the best way to evaluate a team’s defense is to assign a point value to every play the defense gives up. For example giving up 3 yards on 1st and 10 at midfield has a value of -0.01 points from the standpoint of the offenseive team while gaining 3 yards from midfield on 3rd down and 3 has a value of +0.70 points from the standpoint of the offensive team. Turnovers usually hurt the offense by around 4 points.

  For the 2008 season here are the Top 5 defenses (ranked in points given up per 100 plays relative to  NFL average defense).

1. Pittsburgh -17 points

2. Baltimore - 15 points

3. Eagles -14 points

4. Vikings -12 points

5. Titans -11 points.

Thus for example, Pittsburgh’s defenses gives up 17 fewer points per 100 plays than an average defensive team.

Unsurprisingly the worst defensive team was the Lions who gave up 16 .1 points more per 100 plays than an average team. Next worse defense was Denver giving up 15.6 points more per 100 plays than an average team.

Our metric for measuring defenses is very simple to understand but much better than yards per game which is the usual NFL metric.

By the way the Vikings had the best rushing defense and the Steelers  the best passing defense.  The Rams had the worst rushing defense and the Lions the worst pass defense.

By the way our results are adjusted for the strength of the passing and rushing attacks faced by the team’s defense.

In our next post we will look at the NFL’s best and worst quarterbacks.

My book Mathletics (out on August 28) describes our methodology in more detail and also talks about rating NBA and MLB players as well as sports gambling.

How should we evaluate the offensive efficiency of an NFL team? Just because a team scores lots of points does not necessarily mean they have a great offense. It may be that their great defense gave the team the ball in great field position. The best way to evaluate a team’s offense is to look at the average number of points generated by the team’s offensive plays.  We have done this for the 2008 season. Our metric gives you the number of points better than average a team’s offense is than an average NFL offense. (per 100 plays) We adjust for strength of defenses faced. We find that during 2008 the three best offenses were

1. Colts +13
2. Giants +11
3. Saints +10

In other words per 100 offensive plays, the Colts offense generated 13 more points than an average NFL team.

The three worst offenses were

30. Raiders -10

31. Browns -12

32. Rams -16.

Thus Rams offense generated 16 points less than average per 100 plays.

What are the three best passing offenses?

1. Colts +22
2. Saints +19
3. Chargers +16

What are the three best rushing offenses?

1. Giants  +16
2. Patriots +11
3. Redskins +6

In our next post we will examine NFL defensive efficiency and we will see how amazing the Steelers defense was in 2008!

Note that NFL evaluates offenses on yards per game Our approach is better  because our metric is closely tied to points generated by the offense rather than yards. Points generate wins!  Also yards per game does not factor in the true cost of a turnover which is between 3 and 4 points in most cases.

In spring a young man’s fancy turns to love, but in the fall it turns to NFL football!  How should we evaluate a running back or QB? Most running backs are judged on yards per carry. QB’s are juded on a ridicuously complex absurd system.  

   Really the value of say a 3 yard gain depends on the situation.A 3 yard gain on 3rd and 3 at midfield is much better than a 3 yard gain on 3rd and 10 at midfield. My colleague Jeff Sagarin and I have solved this problem by determing a point value for any possible play. See my new book Mathletics  for details of how point values are determined.

For example , a 3 yard gain with 3rd and 3 at midfield is worth around 1.2 points while a 3 yard gain with 3rd and 10 at midfield is worth around -0.40 points.

Using every play for the 2008 NFL season we found that a passing play on  average generated 0.19 points of value while a running play on average generated only 0.08 points of value. Teams pass around 55% of the time, so they seem to realize that the pass is on average a better call.

     The pass, however, is riskier than the run. The standard deviation of points gained per passing play is 1.56 points the standard deviation of points gained per run is only 1 point.

  So here we have an “asset allocation” problem reminiscent of portfolio optimization in finance. Stocks are on average a better investment than bonds but bonds are less risky so people usually hold both investments in their porffolio.

Tune in later this week for a listing of last season’s best passing and running offenses and defenses and a list of the NFL’s best QB’s and running backs by this new measure of effectivness.

Based on the scores of all games through August 15, Here are MLB’s top 10 teams. These ratings are adjusted for difficulty of schedule.

1.       Yankees 1.17 runs better than average.

2.       Red Sox 1.01 runs better than average.

3.       Dodgers 0.81 runs better than average.

4.       Angels 0..81 runs better than average.

5.       Rays 0.80 runs better than average.

6.       Blue Jays 0.62 runs better than average

7.       Rangers 0.56 runs better than average.

8.       Phillies 0.50 runs better than average.

9.       Rockies 0.49 runs better than average

10.    Twins 0.27 runs better than average

Note the AL East has 4 of baseball’s 6 teams

                What makes your favorite baseball team win or lose? Inquiring minds want to know! The wonderful website Fangraphs.com can help you decipher the keys to success for any major league baseball team. Fangraphs.com gives you traditional statistics such as Batting Average and E.R.A., but these really do not tell you the keys to team success. We know that most of a team’s success is due to their hitting, fielding and pitching abilities. Fangraphs.com makes it easy to determine how any team’s performance in these areas affects a teams win-loss record.

                Let’s look at the 2009 Yankees. Select them from teams and select Batters. Next choose Win Probability. The key number is the WPA ( Win Probability added) column. Johnny Damon, for example, has 3.72, indicating that in his plate appearances he added 3.72 wins more than an average player.  The total of the WPA column for all Yankee batter’s was 10.33 wins, indicating that Yankee hitters contributed 10.33 more wins than an average hitting team.

                Next select Pitchers and chose win probability. We can now see how Yankee pitchers contributed to Yankee wins. In his time on the mound Rivera contributed 2.82 more wins than an average pitcher while Chien-Ming Wang contributed 1.97 fewer wins than an average pitcher. Overall Yankee pitchers contributed 4.73 more wins than a team made up of average pitchers.

                Finally, let’s look at Yankee fielding. Choose Fielders and look at the UZR (Ultimate Zone Rating Column). This tells us how many runs better than average a fielder is. For example, Johnny Damon’s fielding cost the Yankees 7.7 runs while Brett Gardner’s fielding saved the Yankees 8.5 runs. Basically 10 runs = 1 win. The Yankees fielders UZR total to -14.8. This indicates that Yankee fielding cost them 1.48 wins. Now adding up the contributions from the Yankee’s batting, pitching, and fielding indicates that the Yankees (as of August 8) were

10.33 + 4.73-1.48 = 13.58 wins better than average. The Yankee’s record on August 8 was 68-42, which is 13 games above .500!

                We can gain unlimited insights from Fangraphs.com. For example, looking at the Devil Ray’s 2007 and 2008 fielding we can see that the Ray’s amazing 2008 season was primarily due to their amazing improvement in fielding. We can also find that Mike Lowell’s hip injury has cost the Red Sox 2 games this year due to his diminishing fielding ability.

                Of course, general managers can use this information (and do) to evaluate trades and make salary decisions. For more explanation of how Win Probabilities and fielding ratings are computed see my book Mathletics which will be out in September 2009.