Daily Dose of Statistical Minutiae, 2/28/13

Oklahoma converted on 3rd down 93% of the time that Jalen Saunders was targeted. Tied for second best nationally was teammate Kenny Stills (75%, tied with Darrin Moore, Texas Tech).

Daily Dose of Statistical Minutiae, 2/27/13

Conversion percentage by distance on 3rd down.

Daily Dose of Statistical Minutiae, 2/26/13

Three players with at least 50 targets caught at least 75% of those attempts while also averaging more than 15 yards per reception: Kerwynn Williams, Utah State (78.9% 15.5); Cody Latimer, Indiana (78.5%, 15.8);  Amari Cooper, Alabama (76.3%, 17.2)

Daily Dose of Statistical Minutiae, 2/25/13

Four players gained 10 or more yards on at least 25% of their carries, Gee Gee Greene, De'Anthony Thomas, Dri Archer, Lache Seastrunk

Daily Dose of Statistical Minutiae, 2/24/13

Yards per carry and YPC on 1st down only for running backs. 

Player	Team	YPC	YPC1st	Diff
Good on 1st Down
Burkhead, Rex (RB/SR)	Nebraska	6.89	8.68	-1.80
Lane Jr., Marlin (TB/SO)	Tennessee	5.48	7.19	-1.71
Malena, Ben (RB/JR)	Texas A&M	5.86	7.31	-1.45
Hill, Bronson (RB/SO)	Eastern Michigan	6.46	7.84	-1.38
Ware, Spencer (RB/JR)	LSU	3.90	5.11	-1.21
Bad on 1st down
Baker, Demetre (RB/SO)	South Alabama	5.04	3.95	1.09
White, James (RB/JR)	Wisconsin	6.45	5.22	1.23
Stephens Jr., Eric (RB/SR)	Texas Tech	5.39	4.16	1.23
Banks, Horactio (RB/FR)	Ball State	5.19	3.66	1.53
Anderson, C.J. (TB/SR)	California	6.27	4.66	1.61

Daily Dose of Statistical Minutiae, 2/23/13

Home teams averaged 4.57 yards/rush. Road teams averaged 4.16 yards/rush. Teams at neutral sites averaged 4.41 yards/rush.

Daily Dose of Statistical Minutiae, 2/22/13

When running against Stanford, opponents failed to get to the line of scrimmage 27.8% of the time. BYU, Syracuse, Ole Miss, Bowling Green, Tulsa, Florida, Western Kentucky had between 21% and 22% tackles for loss/rush.

Daily Dose of Statistical Minutiae, 2/21/13

Two players rushed for over 1,000 yards while also averaging more than 7 yards/carry: Lance Seastrunk (1,012/7.73) and Dri Archer (1,429/8.99). Kerwynn Williams fell just short of the 7 yards/carry mark (1,512/6.94).

Daily Dose of Statistical Minutiae, 2/20/13

Below I have charted average points per game and the Dow Jones Industrial Average (DJIA) since 1900. Besides some weird blips between 1910-1924 and 1945-1965, the fit is remarkable good.

Vegas Has Your Number Part II: Bowl Games and the Spread

Last week on this fine web page, I looked at how accurately the point spread predicted the results of regular season games in 2012. For those folks who are lazy like me and don't want to click the link, I'll summarize concisely. The spread did a fine job, on average missing the actual result by 12.10 points. But bowl games are a different beast right? All that extra time off, coaches moving up to better jobs, coaches being fired, players distracted by all their bowl game swag has to make those games more unpredictable right? Once again, conventional wisdom is dealt a crushing blow. In the last eight postseasons beginning in December 2005, the betting line has differed from the actual bowl game result by an average of 11.93 points over 265 games, all but equal to the 12.10 difference in the 2012 regular season. The table below gives the average spread difference for each of the eight postseasons I examined begging with 2005.

With the exception of 2011 when the point spread enjoyed an uncanny streak of predictability, the average difference has hovered within a three-point range from about 11 to 14. In addition, considering the sample size of bowl games for each postseason (between 28 and 35 for this study) is markedly fewer than that for most regular season weeks (all weeks featured at least 41 games except for the final weekend which saw just 19), the absence of any extreme variation is even more impressive.

Finally, as I did last week, I looked at how often the point spread came within a certain point toal of the actual result. I considered four ranges, and while they are arbitrary, I think they do a good job of sorting the numbers out. The ranges I used were 0-7 points (spread within one touchdown), 7.5-14 points (within two touchdowns), 14.5-20 points (almost three touchdowns), and 20 or more points (seemed like a good cutoff point).

Over 44% of the bowl games played since 2005 have come within one touchdown of the actual result! Consider that for the 2012 regular season, the betting line achieved that level of accuracy just over 37% of the time. Based on this look at the betting line and bowl games, I think it is safe to conclude that Vegas is just as accurate (if not a tiny bit more so) when forecasting postseason games as when doing the same for regular season contests. If you weren't aware of this already, a lot of expert information goes into determining the spread. My advice for retiring early? Give up on trying to beat Vegas and switch to Powerball.

Daily Dose of Statistical Minutiae, 2/19/13

The SEC scored 452 points in 12 games against the Sun Belt (37.7 per game). The Big 10 scored 432 points in 12 games against the MAC (36 per game). The Sun Belt scored 350 points in 9 games against Conference-USA teams (38.9 per game).

Daily Dose of Statistical Minutiae, 2/18/13

Nick Florence had 26 passes for 40 or more yards. Second best were Aaron Murray and Geno Smith with 19.

Daily Dose of Statistical Minutiae, 2/17/13

Two players had more than 50 receptions and 50 carries. Tavon Austin had 111 receptions and 73 carries. Robbie Rouse ran the ball 282 times and caught 63 passes.

Daily Dose of Statistical Minutiae, 2/16/13

SEC might have won its 7th straight national championship, but the Big East was 4-1 in head-to-head match-ups. Only the Big 10 (3-0) and Pac-12 (1-0) had winning records against the least deserving BCS conference.

Daily Dose of Statistical Minutiae, 2/15/13

Louisiana Tech scored 51.5 points per game in 2012. Only two teams have scored more: 1989 Houston (53.5) and 1995 Nebraska (53.2) - the latter was one of the best teams of all time. The Bulldogs edged out 2008 Oklahoma, who scored 51.1 points per game.

Daily Dose of Statistical Minutiae, 2/14/13

2012 was the highest scoring season in the history of college football. Teams in the top division scored 46,018 points. This outpaced previous leaders 2007 (43,086) and 2011 (43,776). One reason for the extra points is that we have more FBS teams in 2012 than in 2011 or 2007, but teams also scored an extra point per game in 2012 than in 2011.

An Ode to the BCS

We have but one year remaining in the BCS, and I thought it time to defend the undefendable:  the BCS is pretty good.  If the goal is to decide who the number 1 team is, the BCS serves its purpose.  Now, there's more to a post-season than just number 1, and it's all the other stuff (like NIU playing in the Orange Bowl) that tends to aggravate people.  But, for a moment, let's focus on the BCS's stated purpose of identifying the best team in college football.

Due to the fact that in a single season no football team plays more than 10% of the BCS teams, it's a difficult proposition to determine number 1.  Unlike other sports, because of the physical demands of the game, you can't dramatically increase the number of games to get a more accurate indication or create an extensive playoff system.  This is why Ken Massey has identified over 132 different ranking systems, each to varying degrees offering a unique portrayal of the state of college football.  However, the overwhelming majority of rankings had Notre Dame and Alabama in the top two, even if they differed on the team's ordering.  For the purposes of deciding who goes to the championship game, it doesn't matter who a ranking says is number one so long as it is in agreement on the top two.  The definitive number one determination is decided on the field as it should be.  While the BCS includes some of these rankings (though by no means the best ones) and two opinion polls, it is imperfect.  But so is every ranking, and the selection committee for the playoff in 2014-15 will be riddled with imperfections.  The BCS has performed admirably in determining the top two teams at the end of the regular season.

There have been rare occasions where the identification of the top two teams going into the bowl season was in dispute.  Undefeated Auburn, Utah, and Boise State in 2004 or undefeated Boise State in 2009 and 2006.  But the vast majority of methodologically rigorous rankings listed on Kenneth Massey's website agree that each of these teams didn't earn a national championship berth.  Even in the contentious years, the BCS was good enough, with most other approaches in agreement on the two teams that should compete for the Coach's trophy.  Below is a comparison of the BCS with my own Network Ranking, Kenneth Massey's, and Wesley Colley's in the highly contentious 2004 season.  Much to the dismay of War Eagle, there appears to be broad consensus on the top two.

Team	BCS	Network	Massey	Colley
USC	1	2	1	1
Oklahoma	2	1	2	2
Auburn	3	3	5	3
Texas	4	5	6	4
California	5	6	4	6
Utah	6	4	3	5
Georgia	7	10	11	8
Virginia Tech	8	18	8	12
Boise State	9	7	9	7
Louisville	10	9	7	13
LSU	11	8	12	11
Iowa	12	11	17	9
Michigan	13	21	20	14
Miami	14	17	10	16
Tennessee	15	13	21	15
Florida State	16	22	16	18
Wisconsin	17	16	22	20
Virginia	18	19	14	19
Arizona State	19	12	13	10
Texas A&M	20	14	15	17
Pittsburgh	21	42	38	29
Texas Tech	22	20	19	22
Florida	23	25	26	28
Oklahoma State	24	15	18	21
Ohio State	25	23	30	25
Oregon State		24	23	24
North Carolina		54	24	33
UCLA		44	25	39
Colorado		27	28	23

While I don't want to offend my friends from Auburn on the 2004 team (who even had "National Champion" rings made after their win in the Sugar Bowl), all four approaches are in agreement on the two teams that have earned the right to play in the national championship.  Auburn, Utah, and Boise State, while undefeated, lack the degree of impressive wins during their season to merit a national championship berth.

Interestingly, consensus is absent between these four approaches beyond the top two, which should cause concern for the mechanics of a post-BCS era.  In a four team playoff, do you still exclude undefeated Utah, as the BCS and Wes Colley would?  Can you really overlook Auburn, Kenneth Massey?  Is a selection committee going to allow more than one team per conference, as the BCS and Colley suggest both Oklahoma and Texas while Massey would recommend both USC and California?  If there had been a playoff in the 2011 season, the Network Ranking would have recommended that LSU and Florida join Alabama and Notre Dame in a playoff.  Unfortunately, it is unimaginable that a selection committee would grant 3 of 4 spots to the same conference, no matter how dominant the SEC may be.  Throughout its brief history, the BCS did a suitable job determining numbers 1 and 2.  But the problems of ranking college football teams only becomes more complex when you expand the number of teams the final ranking will effect. If you expand the pool of possible national champions from 2 to 4, you also expand the possibility for error, and the number of teams like Auburn, Utah, and Boise who will protest their exclusion.

Daily Dose of Statistical Minutiae, 2/13/13

Navy lost to San Jose State 12-0 on September 29. The last time Navy lost while allowing 12 points or fewer was 1997 to Air Force (10-7) and it is only the 6th such loss since 1970. Navy has lost to Army 9 times while allowing 12 or fewer points.

Vegas Has Your Number: A Look at the Spread

One of the facets of college football that I love, in addition to the pageantry, the emotion, and the sheer entertainment of the game is the point spread. Despite more than 100 teams playing at various levels of interconnectedness (Auburn and Ole Miss shared seven common opponents in 2012 while Alabama and Utah State shared no common opponents), the Vegas line gives a reasonable assessment of the quality of each team. I wanted to test this quality by determining exactly how accurate the line is. To do this, I looked at every IA football game in 2012 and determined how far the final margin was from the predicted margin based on the betting line. For example, on the opening night of the college football season, South Carolina was a six and a half point favorite at Vanderbilt. This line was eerily prescient, as the Gamecocks won by four, meaning the betting line and the actual margin were off by just two and a half points. A nearly antipodal result occurred two days later on the first Saturday of the college football season. Texas State played their first game as a IA program on the road against a Houston team that had finished in the top-20 a year before. The Cougars were 34 and a half point favorites over the IA neophytes. In one of the largest betting line upsets in college football history, the Bobcats not only beat the Cougars, they won by a comfortable margin (30-13), making the difference between the betting line and the actual outcome a remarkable 51 and a half points. Surprisingly, that was not the largest difference of the year. UCLA crushed Arizona 66-10 as a slight two and a half point favorite to finish 53 and a half points clear of the spread. So we've seen both ends of the spectrum, but how did the spread perform on average? Before we get to that, full disclosure: I only looked at games between IA teams. While some sportsbooks do make lines for the 'clashes' with IAA opponents, I chose to ignore those and focus only on games involving two IA teams. Overall, of the 697 regular season games involving IA opponents in 2012, the point spread differed from the actual result by an average of 12.10 points. While that may intuitively seem high considering the reverence with which the point spread is regarded, keep in mind that the smallest amount the point spread can differ from the actual result is zero, while the largest amount is theoretically infinite (if realistically about 55). Consider that it takes about six reasonably accurate spreads (say a difference of five points) to bring the huge UCLA/Arizona outlier down to the twelve point average.

Daily Dose of Statistical Minutiae, 2/12/13

In its inaugural year in the Football Bowl Subdivision, Massachusetts managed less than 1 point per possession, but the Minutemen scored 34 points against Ohio in a losing effort. Only one team scored more points against Ohio in 2012 (Ball State).

Daily Dose of Statistical Minutiae, 2/11/13

The 190lb. freshman running back from New Mexico, Jhurrell Pressley, gained a 1st down or touchdown 93.3% of the time when given the ball within 3 yards of the first down marker or goal line.

Daily Dose of Statistical Minutiae, 2/10/13

Utah State's Kerwynn Williams is 4th nationally in yards/target with 12.2. The next running back on this list is Dri Archer at 117th.

Daily Dose of Statistical Minutiae, 2/9/13

Johnny Manziel led all quarterbacks by gaining 25 or more yards once every 13.3 times he ran. Three running backs broke big runs more often: Dri Archer (Kent State) - 10.6, DeAnthony Thomas (Oregon) - 11.5, and Kerwynn Williams (Utah State) - 12.0.

Yards per Pass+

Aaron Murray was the country's most effective passer in 2012 and it wasn't even close.

One of the initial motivations behind CFBTN was to find new ways to compare teams and players across conferences and across the country. The unique challenge with college football is that there are many teams, few games, clustered schedules, and huge disparities across conferences. For example, how do you compare Johnny Manziel and Jordan Lynch when they don't play any of the same teams or even the same caliber of opponent?

To that end I have developed an algorithm that attempts to answer that question. Essentially, it attaches a value to offenses and offensive players and defenses and uses those values to predict an outcome - e.g., yards per rush. It then compares the predicted values with actual values from actual games and makes adjustments as necessary. The process repeats (dozens or hundreds of times) until the predicted values are as close to the real values as possible. We will never get a perfect fit, but the values generated by the process are the best possible estimates of a team's or player's proficiency in that particular area.

I call these "plus" statistics - e.g., YPP+ (yards per pass plus). Instead of reporting what the team or player actually did during the season, they report what they would have done against a schedule of average opponents. Unfortunately, for quarterbacks I cannot adjust for a player's own team, so a player on a better team (better offensive line, better receivers) playing against this schedule of average opponents has an advantage over another player on a lesser team.

There are many ways of measuring a passer's efficiency, but the one that offers the most bang for the buck is yards per pass, or, even better, yards per pass+. Below is the top 25 by YPP+ and their national ranking in YPP among players with at least 200 pass attempts. For comparison is the top 25 defenses. The dYPP+ means, for example, that Florida allows 2.51 fewer yards per pass than the average defense.

Against an SEC schedule, Aaron Murray averaged 10.1 yards per pass, but should have averaged 10.4 yards per pass against an average schedule, a full yard better than any other quarterback with a significant number of attempts. He faced two of the top four and three of the top 10 defenses by dYPP+. Manziel managed 8.5 YPP despite 4 of the top 10 defenses by dYPP+, but few teams were worse than Louisiana Tech (West Virginia was one of those teams).

More impressive was Taj Boyd, who faced three of the top four defenses by dYPP+ and averaged 9.1 YPP. Nick Florence's 9.3 yards per pass was inflated by his 51 attempts against West Virginia (#118), but brought back down by games against Oklahoma (#7), Kansas State (#13), Oklahoma State (#15), and TCU (#16). McCarron matched Florence with 9.3 YPP and YPP+.

Missouri's James Franklin wins the award for toughest schedule for throwing 51 passes against Florida, a third of his passes against Florida, South Carolina and Vanderbilt (all in the top 20) and half against those three and Georgia (who came in at 33rd). He missed games against Arizona State (#9), Alabama (#10), and Texas A&M (#27); Berkstresser didn't attempt enough passes to qualify for this list, but  otherwise could have claimed an even tougher schedule than his Missouri teammate.

The Best Single Game WR Performances of 2012

Having already posted the top QB performances of 2012, I suppose it's a little predictable that some of the QBs from that list will have helped some of the WRs on this list.  Either way, here's the list.

#10: Alex Neutz (Buffalo) vs. Morgan State (+17.5 EPA)

Buffalo had just played somewhat well against Georgia but a lack of talent led to a multiple score loss.  The following week they took out their frustrations on lowly FCS Morgan State.  Neutz had 8 catches for 154 and 4 TD.  The Junior finished the year with 65 catches for 1015 and 11 TD.

#9: Kenny Stills (Oklahoma) vs. West Virginia (+17.6 EPA)

One of the best games of the year, and if you love individual performances that produce video game stats, this game was for you.  In this game, Oklahoma's Landry Jones posted our #2 performance of the season, and he spread the ball around pretty well.  Three OU receivers - Jalen Saunders, Justin Brown, and Sterling Shepard - all had more yards than Stills.  Stills, on the other hand, had the TDs - 4 of them, on 10 catches for 91 yards.

#8: Tobais Palmer (NC State) vs. Clemson (+18.1 EPA)

While Tajh Boyd was having a historic game for Clemson, NC State's Mike Glennon managed to keep the Wolfpack in the game by hooking up with Palmer for big play after big play.  Palmer had 7 catches for 219 yards and 3 TD:

The scary part:  The EPA listed above doesn't include special teams contributions.  Palmer had 8 returns for another 277 yards that day.

#7: Allen Robinson (Penn St) vs. Indiana (+18.2 EPA)

It's nice to just talk about Penn State football.  The sophomore Robinson exploded onto the scene in 2012, leading the Big Ten in catches (77), receiving yards (1046), and TD catches (11).  This was Robinson's best game - 10 catches for 197 and 3 TD.  The following short highlight from the game shows only two of his catches, but they both let you know why he's a force to be reckoned with.  On a screen pass, he makes multiple defenders miss and takes it to the house.  Then, against decent man coverage, he adjusts to the ball and makes an impressive catch in the endzone.  Enjoy him while you can, Happy Valley - the NFL could very well be calling in a year.

#6: Cobi Hamilton (Arkansas) vs. Rutgers (+19.1 EPA)

In the loss to Rutgers, Tyler Wilson completed 20 passes for 419 yards and 3 TD.  To receivers not named Cobi Hamilton, he completed 10 passes for 116 yards and 0 TD.  If you're not good at math, I'll tell you that leaves the following line for Hamilton: 10 catches for 303 and 3 TD.  Very impressive, seeing how Rutgers had a bit of a pulse this year on defense and all.  His 303 receiving yards set a school record, and more importantly, an SEC record.

Hamilton finished the season with 90 catches for 1335 yards, but only 5 TD, 3 of which came against the Scarlet Knights.

#5: Cody Hoffman (BYU) vs. New Mexico State (+22.4 EPA)

12 catches, 182 yards, and 5 TDs. 

Watch this video, and you may realize that you're probably one of the first 20,000 humans to have seen his performance.  Hoffman finished the year with 100 catches, 1248 yards, and 11 TD.

#4: Terrance Williams (Baylor) vs. West Virginia (+23.1 EPA)

What can I say about this game that hasn't already been said?  You can watch the game in the QB performances post.  Williams had a ridiculous 17 catches for 314 and 2 TD.  The consensus All-American led the country in receiving yards this year with 1832.

#3: Marqise Lee (USC) vs. Arizona (+23.4 EPA)

USC had a year to forget, but if they need some solace, they can take it in the highlights of Lee, one of the best receivers in Trojan history.  Behold his best game:

16 catches for 345 and 2 TD was enough to crack our top 3, but it wasn't enough to beat Arizona.  In this game, Lee set the Pac-12's single game receiving record.  The All-American would go on to lead the country in catches with 118, and had over 2600 all-purpose yards on the year.

#2: Tavon Austin (West Virginia) vs. Oklahoma (+25.6 EPA)

I'll admit, I wasn't quite sure whether to include Austin in the RB countdown or leave him here.  I decided to leave him here as he is, primarily, a WR.  These aren't exclusively the top receiving performances, nor was the RB list limited to rushing contributions alone.  And with 114 catches on the year, Austin was pretty clearly a receiver.  On this day, however, he gashed Oklahoma on the ground, rushing the ball 21 times for 344 yards and 2 TD.  He also had 4 catches for 82 yards.  And 8 kick returns for 146, and those aren't even factored into the EPA.  His 568 all-purpose yards set a Big XII record, and Austin's 344 rushing yards set a school record.  Not bad for a WR.

Maybe the most amazing fact about our #2 WR performance of the season?  It wasn't even the best WR performance on his team, which leads us to...

#1: Stedman Bailey (West Virginia) vs. Baylor (+26.0 EPA)

(Baylor game starts at 2:18)

Geno Smith had, by far, the best QB performance of the year.  Somebody had to catch those passes.  Tavon Austin had more catches, but nobody touched Bailey in terms of big plays.  13 catches, 303 yards, and 5 TD - only one-fifth of his amazing season total.


Brent Blackwell compiles the EPA rankings for cfbtn.com.  Follow Brent on Twitter by mashing the pretty button below. Follow @brentblackwell

The One Man Show (OMS) for Quarterbacks

Finally, I found an advanced quarterbacking metric that Manziel does not win. He finishes 3rd.

Football is played in drives and downs, but individual player statistics are collected as totals per game and averages per play. The problem is that performance in the latter does not equal success in the former. For example, a player that gets 4 yards on every play will score twice as often as a player that scores a long touchdown once every five plays and gets 0 yards on the other four, but the second player will have the higher yards per play (and will get on Sports Center).

The One Man Show (OMS) is my attempt to measure a player's ability to keep the drive alive. Here's the situation: using actual plays from the past season for each player, I simulate drives. If the player can get a first down on three plays they continue the drive. The field is infinitely long, so the drive could go on forever. The drive can also end on a turnover. Each player is given 10,000 plays and they average about 1,700 drives (I would like to get that up to 100,000 plays but with my current computing speed that would take most of a day; if you would like to see 100,000 simulated plays, make a donation and help me get a new computer . . . seriously, do it). These are based on real game plays and are not adjusted for the strength of the opposition, and the list is limited to quarterbacks with at least 200 plays.

The Yards/P is average yards per possession or per drive. This is the average number of yards the player was able to gain in simulations without getting to 4th down. The Cowboys' Walsh averaged almost 70 yards, followed by David Fales, The Johnny and Aaron Murray. Fales led the nation in completion percentage and Manziel and Murray were no slouches in that department and were at the top nationally in explosive plays - big plays move the ball down the field in a hurray.

At the other end, the mighty quadrumvirate of Colorado, Hawaii, Southern Miss and UMass are represented. Their quaterbacks averaged fewer than 22 yards and around four and a half plays per drive (remember, you get three just for showing up).

The second column, with similar results, is yards of field position per possession. In this case, the team punts on fourth down, unless they have turned the ball over, and the total change in field position is reported. The third column is the average points scored assuming the team started on their own 20 yards line. Again, a quarterback that performs well in the first two performs well here as well; Walsh average six times as many points as Hawaii's Schroeder.