CFBTN vs. Vegas week 5

The chart below compares an average Vegas line (as of Thursday) to the CFBTN prediciton. CFBTN likes the home team more in games to the left of the line, and vice versa. The three games this week where the model and the house disagree by more than a touchdown are highlighted in red. 

Smart 3rd Down Efficiency

3rd down efficiency measures how often a team converts on third down. The problem with this metric is that it is influenced more by how a team does on 1st and 2nd down to set up 3rd down than how they actually do on 3rd down. Smart 3rd down efficiency adjusts the results for the distance needed to gain a 1st down. A positive number means the team converts more 3rd downs than would have the average team in those situations. A negative value means the opposite, that the average team would have been better in those same situations.

Offensively, the top team in the nation on 3rd down has been Purdue. The Boilermakers have converted 21% more 3rd downs than average. Oregon St has had the nation's best 3rd down defense, allowing 19% fewer conversions.

Previewing Tulane Watch 2012, track meets in the Big 12, and the chaos that is college football

Matt Forte no longer plays for Tulane.
I have the stats to prove it.

Tulane Watch 2012: I’ve decided to track Tulane all season as the Green Wave try to become the worst offense in the history of college football. Right now they are last in points, yards, rushing yards (10!!) per game, and rushing yards per attempt. The big outlier is their stellar 60% completion percentage. This week they get ULM, who has not allowed fewer than 31 points all season. The model projects Tulane to 30 rushing yards, 1.1 yards per attempt, and 18 points. In other words, this is a real opportunity for Tulane to boost their stats.

What a difference a year makes. In 2011, Wisconsin beat Nebraska by 31. Heading into this weekend, Nebraska is favored by 14 (-13 in Vegas). Houston ran up 73 against Rice last year, but now the model likes the Owls by 1 (+4.5 in Vegas). Finally, Akron got a 32 point whooping from Miami (OH) in 2011 but now the model is giving them a 2 point advantage (+5.5 in Vegas). And if you haven’t heard, Arkansas lost to ULM and Rutgers.

CFBTN 101: Notre's Dame title game odds

Notre Dame just keeps winning games. If they win eight more they'll earn the privilege of losing 38-7 to Alabama in the MNC title game. Right now the model gives Notre Dame an 3.4% of reaching that point.

Introducing the EP3+ prototype

EP3, or effective points per possession, evaluates drives based on how many points they produce, not just on the scoreboard, but also by affecting field position to make it harder or easier for the opponent to score. EP3+ adjusts that evaluation to recognize that not all opponents are created equal. While I've been reporting EP3s all season, this is the first ever publication of the EP3+, and it is really just an EP3+ prototype. In the final version drives will be weighted by leverage - in other words, the more impact the drive can have on whether or not the team loses or wins, the more importance it will be given when calculating EP3+s. While the current version accurately reflects how well teams have played, a version that gives less weight to drives late in blowouts will be more predictive.

Alabama comes out on top in the EP3+ as it does in any power ranking worth its salt. Texas comes in #2, and while the rank is not hugely shocking, the reason for the ranking is: to this point, the Longhorns have had the nation's 2nd most effective offense. Texas Tech is 5th primarily due to its 2.79 EP3 average on offense (compared to a national average of .27).

16 stats: What's wrong with Arkansas?

So, what's wrong with Arkansas? Short answer: Petrino crashed his bike with a girl on it and got fired, but his schematic legacy of not protecting the quarterback remained which led to Tyler Wilson getting a concussion and, inexplicably, Arkansas doesn't have an adequate backup despite their pass-heavy, let-the-quarterback-get-slaughtered approach to offense.

Long answer: pretty much everything.

For a closer, more statistical, less tabloid/seedy look at the Arkansas transformation from 2011 to 2012, we turn to the 16 Stats that Matter. This offers a proximate determinates approach to analysis. That is, it won't tell us what a team can do to fix their problems, but it can tell us where a team should be looking.

CFBTN 101: Forecasting bowl eligibility

Arkansas has lost three straight games. That last happened in 2008, Petrino's first season, when the Razorbacks lost to Alabama, Texas, and Florida. In 2012, Arkansas has lost to Alabama again, but now they've also lost to Louisiana-Monroe and Rutgers. 2008 is also significant because it is also the last season Arkansas was not bowl eligible. Arkansas gets Texas A&M this weekend and a chance to match its 2004 stretch of four straight losses.

"Team: Summary" offers a quick overview of the expected outcomes - number of wins and losses, going undefeated or being bowl eligible, division and conference titles - for all 124 teams. The model is now giving Arkansas an 8% chance of being bowl eligible this season and projecting them to 3.8 wins this season.

Who's College Football's Kevin Bacon?

Last week we looked at how few games connect some conferences (Six Degrees of College Football). This makes it virtually impossible to evaluate teams across those conferences, particularly when we are picking two teams to play for the MNC.

Because there are so few games stitching the nation together, those games and the teams playing in those games critically shape our perception of the national landscape. For example, there are only two 1st and 2nd order connections between the SEC and Pac-12. This amounts to six total games, only two of which actually involve both SEC and Pac-12 teams. If Oregon and LSU each have one loss at the end of the season and we are trying to decide which one gets to lose to Alabama in the MNC game, those six games, and especially the two 1st order linkages, will most strongly influence our opinion of the two conferences and the teams from those conferences. (Looking back to last season, Alabama ultimately got a second shot at LSU because LSU whooped up on Oregon in the first week.) If Washington plays below its ability against LSU and then over its head in conference games, Washington will make it significantly harder for Oregon to play for a national championship.

So, who's our Kevin Bacon?

Blackwell's EPA, the bottom 50

[These results cover the first three weeks of the season.]

Some readers might consider this post vicious, but I do it only to prove a point. In the table below are the 50 worst quarterbacks in college football by EPA (expected points added). My point is this: if you want to know why a team is struggling, look no further than quarterback play. So yeah, it is a little vicious.

That UMass is bad is not surprising to anyone, but Houston's incompetence on a football field this season has been overwhelming. This is our first season calculating EPAs and NEPAs, so I can't compare Piland's numbers to Keenum's from a year ago, but I would guess that he did better than a -33.5 EPA.

Poor Brandon Allen stepped in for an injured Tyler Wilson and cost his team a half point every time he didn't hand the ball off to a back (Knile Davis has a -.2 EPA/PP, so Arkansas didn't really have any better options). James Franklin has battled injury while his team has been outscored by 42 in two games against SEC opponents. (Berkstresser has not been much, or any, better).
[click "Read More" to see the table]

Blackwell's EPA, week 3 QBs

Expected points added (EPA) is not the only metric we should use when evaluating players. That being said, Doege's 66.9 NEPA is mind blowing and demands that the Tech QB be included in the Heisman conversation. Through three games, he has thrown for 858 yards and 12 TDs while averaging 8.5 yards/attempt. After a warm-up game against Northwestern St, he has completed more than 77% of his passes and is averaging a hair under 10 yards/attempt in two of the four best single-game EPA performances by a QB this season.

But even Doege's performance is outdone by what Pachall has been doing for TCU on a per play basis. Pachall has thrown for 841 yards and 8 TDs, completed 76% of his passes and is averaging just under 12 yards/attempt. His .83/.89 EPA/NPA/PP fall short of De'Anthony Thomas' numbers (see RB EPAs in the last post), but it leads all other QBs (but Minnesota's Shortell and teammate Trevone Boykin, who has attempted just 9 passes) by a significant margin.

Blackwell's EPA, week 3 RBs

Through the first three weeks, De'Anthony Thomas leads all running backs by expected points added (and by a healthy margin). Thomas has helped his team score 33.4 more points and outscore their opponents by 36.3 more points than would have the typical running back given his opportunities. That comes out to 11 and 12 points per game. More remarkable, though, is his production per play. Thomas adds almost a point and half every time he is involved in a play. Teammate Kenjon Barner just misses the top 50 at #51. Despite significantly more touches, Barner's EPA is only 6.2.

Tulsa has two backs in the top 10 with almost identical per play production. Alabama has only one back in the top 50, and it's not Eddie Lacy. Instead, it is freshman TJ Yeldon at #16 (and I just barely missed him on my fantasy team). And where, you might be wondering, are Heisman hopefuls Marcus Lattimore? Montee Ball? Knile Davis? Eddie Lacy? All four are in the negative. Montee Ball has cost his team a full touchdown with his lack of production and Knile Davis is moving quickly towards two touchdowns worth of ineffectiveness.

Week 4 EP3s

Nevada scored 69 points in 11 qualifying possessions against Hawaii. USF and Ball St scored touchdowns in 7 of 13 qualifying possessions. On the other end, Arizona was shutout by Oregon despite having more than a dozen opportunities.

Arkansas, Arizona still aren't good, Oregon still is, and how Directional Michigan was almost like a unicorn

Four interceptions and a sub-.500 completion percentage, combined with fewer than 3 yards per carry, is not going to beat Oregon. Three interceptions and a sub-.500 completion percentage, combined with 3.5 yards per carry will get you one failed two-point conversion attempt from beating Boise State. Turns out Oregon is still Oregon and Arizona is still Arizona, and BYU and Boise State have devolved into hideous, offense-less monstrosities.

I said the first team to 35 would win the Baylor-ULM game, and I thought it would be Baylor. I was wrong. Baylor was first to 34, ULM beat them to 35, and then Baylor won. I think Briles did that just to mess with me. Then, Arkansas loses by 52 to Alabama, it loses to Rutgers with Tyler Wilson, and Auburn almost beats LSU, and we still have no idea just how good ULM really is. That reminds me, Rutgers beat Arkansas. And in other news, Rutgers beat Arkansas.

Quick note: How good is Alabama?

If the season ended now, this Alabama team would be the 4th most dominant team by power rating in the modern era of college football, behind 2001 Miami, 1995 Nebraska and 1974 Oklahoma. I was out of the country in 2001 and wasn't alive in 1974, but I remember the '95 Nebraska team well. I remember being sincerely concerned for the welfare of their opponents.

Previewing Bowden Ball, Montee Ball, Tulane running the ball, and words that rhyme with Waco

Good thing its a road game for Baylor because
nothing good rhymes with Waco; I looked it up.

1) ULM gets one more shot at an AQ team in Baylor today. This game has surprisingly big national championship implications, not because either team is a contender but because it will influence how the contenders in the Big 12 and SEC are viewed after beating them (or beating teams that played them). There is only one first order connection (A plays B) between the Big 12 and and SEC (Ole Miss vs. Texas) and nine second order connections (A and B play C). Two, more than 1/5th of those 2nd order connections, go through ULM. Two more go through SMU. The model actually likes ULM by 1 and Vegas likes Baylor by  a touchdown or so. I say first team to 35 wins, and I'm pretty confident the first team to 35 will be Baylor.

2) You can tell a lot about the prominence of a public university and its football team in that state by googling the state name. Search "Alabama" and the university comes up second and news for the football team fourth. Search "Florida" and the university comes up on the second page and the football team several pages back. Search "New Mexico" and you get three results. Search "New Mexico versus New Mexico State". Really, go, do it. I'll wait. Poor, poor New Mexico. This year they are coming in 119 and 112 in the power rankings. Someone will lose and someone will win, but who really cares?

This really happened, and this guy played
running back at West Virginia (Kegs N' Eggs)

3) In successive weeks Akron has scored 14, 32, and 66. At this rate, Akron will score over 100 in their next game. If they do, Tennessee will be looking for a new coach on Monday. (If they work fast they can catch Terry before he leaves town). The model recognizes that Akron is not a good team, but it does see that they are averaging 378 passing yards per game and that Tennessee has allowed almost 25 points per game against NC State, Georgia State and Florida. Tennessee will win by a lot, but Akron and the end zone will not be strangers.

4) Ole Miss plays Tulane this weekend in the most important game left on their schedule. Why? Because it is the last game this season the Rebels should win. The model has them as 1 point favorites against Auburn right now, but 11 point favorites against Tulane this weekend. Tulane is bad in every way and Ole Miss is 123rd nationally with a 73% completion percentage against (and that was against offensive powerhouses UTEP, Central Arkansas, and Texas). Predictably, the model sees the two teams gaining a lot of yards through the air, and since Tulane is averaging 8 rushing yards per game, the Green Wave better hope there will be a lot of passing yards. (That's not a typo, by the way. They are averaging 8, eight, 7+1 yards per game and 0.3 yards per carry. At that rate, it would take them 333 plays to get from end zone to end zone.)

5) Western Kentucky is a 4 point favorites against Southern Miss in Vegas, and the model likes them by another touchdown. Southern Miss is completing 44% of their passes and getting 3.1 yards per carry. And Western Kentucky just beat an SEC team (which tells you more about the bottom of the SEC than Western Kentucky). The bad news for the Hilltoppers is that this game will be a preview of what happens when a major program steals your head coach.

6) In three games against UNI, Oregon St and Utah St, Wisconsin has a +4 point differential (and they're ranked?). The Wisconsin running game has been abysmal, but UTEP allows 5.6 yards per carry and has the 98th rated run defense in the FBS. If Montee Ball can't get on track this weekend, he won't all season, and it will be a very long, long year in Madison. Win or lose, if Wisconsin gets less than 4 yards per carry this weekend then I'll predict now they won't be bowl eligible.

StatView: Arizona vs. Oregon

In statview, I begin by ignoring everything I know, think I know, or just imagined I know about two teams. I then look at their statistical profiles and, from that, try to predict what the game might look like and some keys to watch for.

Looking at Arizona and Oregon, the first thing that jumps off the page is that this game will have a lot of statistics. Combined, they are averaging over 1,200 yards and 100 points per game. They do this in part by having the 4th and 5th highest tempo offenses in the country. Since they average about 15 possessions per game against lower tempo opposition, together they could have upwards of 20 possessions per team. If they are able to average 3 points per possession (currently at 3.3 for Arizona and 3.5 for Oregon), a Vegas total of 77.5 seems low. That being said, the model predicts 78 points with about 32 total possessions and 2.5 points per possession.

Oregon has clearly been the better of the two teams. Oregon has averaged over 50 points per game in their last 4 meetings (to Arizona's 36). But it is harder to get a read on this season because Oregon has not yet been seriously challenged. It is difficult to know how its 6.2 rush yards per attempt and 329 rush yards per game will hold up against stiffer competition, but its opponent-adjusted rush rating of 8.7 puts them 10th nationally right now.

Arizona, on the other hand, has been tested. They rebounded from a less than awe inspiring win against Toledo by thumping Oklahoma St. On the surface, Arizona and Oregon look very similar, which, in combination with the new head coach, has led many to see Arizona as a poor man's Oregon. Arizona has completed 70.1% of its passes for 11.4 yards per completion, while Oregon has completed 70.7% for 11.5 yards. The lead backs for the two teams are averaging 5.8 yards per carry; Carey and Barner are virtually identical across the board in the individual stats. (Arizona does not have anyone that can match De'Anthony Thomas' 17.5 yards per carry, though.) But Arizona throws the ball 48% of the time to Oregon's 32%, gets 60% of its yards on passing plays, to Oregon's 45%, and Arizona averages one full play more per possession. Instead of Oregon's approach, Arizona has accumulated statistics this season by playing fast, avoiding turnovers and penalties (especially compared to Oregon), and sustaining drives.

On defense, the keys for Arizona will be to contain De'Anthony Thomas when he gets the ball (he is projected to 100 yards on 7 touches) and turn Oregon over. There is no reason to believe, statistically, that Arizona can consistently stop Oregon from moving the ball, so they will need to force them to move the ball all the way down the field and hope for turnovers and penalties to kill drives.

On offense, Arizona will need to maintain its 11.4 yards per completion. Oregon is allowing only 8.6 yards, a key to its 4th ranked opponent-adjusted pass defense. Arizona needs to stay close on the scoreboard and on schedule in down and distance to keep the whole playbook available. 

The model likes Oregon by only 12, 45-33 and gives the Ducks an 82% chance of winning. It projects only 815 yards or so of offense between the two teams, so 78 yards will only be possible if turnovers or special teams create some short fields and the kickers stay off the field. I would normally say the stats favor Oregon if the game becomes a shootout, but Arizona more than held its own against another high noon veteran in Oklahoma State two weeks ago.

CFBTN 101: Pick All, Georgia vs. Alabama

CFBTN has predictions for every game in the country, this week and every other week. In addition to weekly games the model also predicts team records, bowl eligibility, conference championships, and the MNC. "Team: Pick All" is unique because it offers predictions for events that won't every happen. "Pick All" matches up all 124 teams against every other team at a neutral site and predicts the final score and win odds.

One prediction you will find in "Pick All", and a game that may actually happen, is Georgia versus Alabama. The model gives the two teams solid odds of winning their respective divisions, in part because they are not scheduled to play each other during the regular season. While the model likes Georgia, it likes Alabama a whole lot more. It favors the Tide by 15, 32-17, and gives them an 86% chance of winning. Find all "Pick All" picks at Team: Pick All.

Six degrees of college football

With 124 teams and 12 scheduled games, FBS college football teams play less than a tenth of the national competition. This problem is compounded by scheduling incest - 2/3 to 3/4 of almost every team's schedule is filled with the same handful of conference opponents. As a result, we depend heavily on inter-conference play and the transitive property* to evaluate teams. But with only 3 or 4 out-of-conference games for most teams, it can often take multiple games and multiple teams to connect any two teams, and the random variation that affects every game is multiplied many times over.

The chart below shows the number of 1st order (the teams play each other) and 2nd order (the teams play the same 3rd team) connections between conferences. Some conferences are paired with several 1st order connections: Big 10 and MAC, Sun Belt and SEC, WAC and MWC, for example. The number of 2nd order connections are more difficult to guess; for example, there are a large number of 2nd order connections between the ACC and MAC largely because they both play a lot of games against the Big East.

Looking at conferences with legitimate national title contenders, we quickly see where the scarcity of inter-conference connections can become a serious problem for selecting teams to play for the MNC. The ACC and Big 12 have 3 and 4, the Big 12 and Pac 12 have 1 and 3, and the SEC and Pac 12 have 2 and 2.

* A is 10 points better than B and B is 20 points better than C so A is a lot better than C, too.

CFBTN 101: P-Top, -Mid and -Bot

Technically, P-Top, -Mid and -Bot are three measures of how hard it would be to go undefeated, win half your games or win at least one game in each conference. An equally valid way of interpreting these measures, though, is that they tell us how good the conference is at the top, from top to bottom, and at the bottom.

The SEC is leading the nation in P-Top, but as in 2012, the Big 12 beats out the SEC in P-Mid and P-Bot. Any unbiased observer (if there is such a thing) would agree: while the SEC has the nation's elite teams (and a lot of national championship trophies to show for it), the Big 12 has only one uncompetitive team to the SEC's two. The Pac 12, Big 10 and ACC are 3rd, 4th and 5th in that order, with the Pac 12 scoring relatively low at the bottom. The other notable outlier is the WAC, who is 7th by P-Top and 11th in the other two, suggesting that Utah St and Louisiana Tech (and possibly San Jose St) are competitive at a much higher level than the rest of the conference. Go to Conf: Rank for complete results.

Blackwell's RB EPA rankings, week 2

Yesterday we revealed quarterback EPA rankings. Today we unveil running backs.

In two games, Cody Getz has 348 yards rushing and 6 TDs. UCLA's Johnathan Franklin had 431 rushing yards and 3 TDs at that point. EPAs for Kenny Hilliard (14.1), T.J. Yeldon (12.3) and De'Anthony Thomas (11.4) are especially impressive because they are doing it in loaded backfields. At the other end, overrated Big 10 backs Montee Ball and Le'Veon Bell have a combined EPA of 2.5 (10% of Cody Getz), and Montee Ball has a negative NEPA.

 

CFBTN 101: First down on 3 carries

Two running backs average 5 yards per carry. One gets exactly 5 yards per carry. The second gets 25 yards every 5th carry and 0 on the rest. The first running back could carry his team to a touchdown on every possession while the second would never score unless they were gifted with great field position.

First down on 3 carries measures the probability that a back would get 10 yards on 3 carries (and keep a drive going) based on an estimated distribution of their yards per carry. Of qualified backs (at least 36 carries) Johnathan Franklin is still leading the pack. FIU's Kedrick Rhodes is a surprise #2, with only 4.9 yards per carry, but a real life example of the situation described above. Rhodes has a 92% of getting at least 2 yards when he gets the ball but only a 7.4% chance of getting 10 yards. Instead, he consistently gets around 4 or 5. For more First down on 3 carries results go to Individual: Stats above, click on Qual under Runners, and sort by First down on 3 carries.

Blackwell's QB EPA Rankings, week 2

In football, individual players produce in a lot of ways. They rush and pass and catch and tackle and sack and fumble and get first downs and score touchdowns and kick field goals, etc. But as with teams, all that really matters is points. Expected Points Added looks at the points a team can expect to score before a play occurs and the points after the and subtracts the difference to see how points the play, and the players involved in the played, contributed. 

Each week, Brent Blackwell pours through thousands of plays to calculate EPAs and NEPAs* for hundreds of college football quarterbacks and running backs. Throughout the season, we will be ranking these players by their EPA contribution. As the season progresses, we hope to also include rankings for EPA per play and conference rankings. These rankings include all games through week 2.

A Texas Tech quarterback tops the rankings. This is most notable because Tubberville is now the man in charge there and not Leach (who does not have a quarterback in the top 100 after getting manhandled by his alma mater in week 1). Tech is scoring over 50 points per game, Doege has thrown for over 300 yards in the last two and has a QB rating over 200 in both games. The going will begin to get tougher this week, though, when he faces an Iowa St team that has a 48% completion percentage against.

Arizona St's Taylor Kelly, not to be confused with Kelly Taylor, is second in EPA (but significantly down the list in NEPA). He is fourth nationally in yards per attempt and led the Sun Devils to 108 points in their first two games.

A number of big names fall in the 5-11 range: Tyler Bray, Tyler Wilson (who will probably benefit from missing the Alabama game), Matt Barkley, Collin Klein, Braxton Miller. David Ash for Texas at 14 is a bit of an eye opener - quarterback play has helped hold the Longhorns back since Colt left. Zach Mettenberger and Aaron Murray fall just behind Missouri's second quarterback, Corbin Berkstresser, at 61st and 62nd. 

*NEPA is net expected points added. It includes both the points the team can expect to score on this possession and also the points their opponent can expect to score on their next possession.

Week 3 EP3s

Beginning next week I will be reporting opponent adjusted effective points per possession statistics, but for one last week we just get the raw numbers. Notable results: the Arkansas offense came in second to last, just ahead of Lamar. Houston and SMU finished 127th and 128th offensively; while UCLA and Texas A&M obviously played a role there, this probably says more about the state of Houston and SMU. Michigan St was an embarrassing 123rd against Notre Dame and Utah St should have won Saturday despite coming in at 120th (because Wisconsin had the week's worst EP3 in a winning effort at -1.71).

Texas's 2.67 EP3 against Ole Miss is just bizarre. Last year Texas couldn't have managed a 2.67 without an opponent on the field. On the other hand, Baylor managed only -.02 against Sam Houston St. Fresno St had an EP3 over 2 against Colorado. Poor, poor Colorado.

Week 3 recap

Yeah! We're still not good at football, but the football gods like us
more than Utah St!

1) I picked Utah St to win this weekend because I believed a) they were as good as Wisconsin and b) the football gods owed them one. I was right. I was wrong in thinking that the football gods cared.

2) I reported that Nick Saban would pick a score from 2011 and challenge his team to match that score against Arkansas. I still have not heard which game he picked, but we can narrow it down to two: a) the Ole Miss game and they forgot to let Arkansas score or b) the Arkansas game but they decided to match the total, not the actual score. Either way, Arkansas did not beat 'Bama 100-3.

3) We also learned that coaching matters in football; see Houston, Southern Miss, the aforementioned Arkansas and, at the next level, the Saints of New Orleans. I would throw USC in the mix.

3) We learned that ULM's defense is really letting that team down. Giving up 31 points in consecutive weeks to offenses that look like something we were running in 8th grade? Seriously, Auburn and Arkansas have looked absolutely inept on offense since the first week. Every team in the SEC now must become serious ULM fans to keep these games from too dramatically bringing down the conference ranking.

4) We learned that those pundits that lost their pundit cards after the Syracuse game for saying USC has the nation's best starting 22 are not getting them back. Ever. San Jose St came closer to beating Stanford that USC. But before learning that USC is not legit, I learned that Florida St is, and even if they weren't they'd still go undefeated (for more on this, search Hokies, Panthers, and Penguins).

5) I was actually disappointed because a UCLA running back failed to top 200 yards rushing in a game (I predicted here that he'd be closer to 500 yards). It is just amazing how quickly expectations can change for a  team and a program.

My theory: UNC was intimidated by the size of Wright's arms
until they remembered you don't run with your arms. (AP)

6) You can't judge a game by the first 17 minutes. Louisville was up 29-0 on North Carolina and won by 5. Texas A&M was scoreless against SMU, and then scored touchdowns on 7 of their next 8 possessions. Speaking of A&M, we learned that losing by 3 in their season opener with a freshman QB might end up being a positive mark on an otherwise very good resume . . . and Tennessee is still 21st century Tennessee (they allowed 336 yards rushing against Florida to A&M's 142).

7) Mississippi St beat Troy by 6, Ole Miss gave up 66 to Texas. I've already discussed Auburn and Arkansas. Who's afraid of the SEC West? If you took out Alabama and LSU, who's afraid of the SEC?

Yards Adjusted Completion Percentage

One "advanced stat" I developed a few years ago is the yards adjusted completion percentage (YACP). I began with the assumption that it is harder to complete longer passes. To reflect that I adjust each QB's completion percentage using an average completion percentage by pass distance to calculate the YACP. Two weeks into the season the list is headed by Lousville's Bridgewater in on top and Tyler Bray comes in at number 2. Brett Smith, Everett Golson and Zach Maynard also rank in the top 10 despite completing fewer than 70% of their passes overall. You can find the complete list here (click on yards adjusted to sort the list).

Week 3 picks and other miscellania

Nick Saban will be able to pick the score this weekend against Arkansas. My sources tell me he will literally pick a score from last season at random and challenge his team to match that score (to keep them from looking past this game to Florida Atlantic). This should give Arkansas a 1 in 13 chance of winning, but I'm not sure even Nick Saban could coax 9 points out of Arkansas on Saturday.

Yeah, this isn't going to help

Utah St will beat Wisconsin this weekend. The Badgers are 14 point favorites in Vegas and the model likes them by 12 (33-21), but karma says this one is going to the Aggies. After blowing leads and losing close games all season last year, Utah St wins a game in overtime against in-state rival Utah? Something's rotten in the state of Wisconsin (and most of the Big 10) and Utah St will pile on the dirt.

I painted their uniforms red to give you a visual preview of Saturday.
That's the kind of production value you find at CFBTN and nowhere else

Houston is allowing 247 rush yards/game and UCLA is averaging 1,000,000 rush yards/game. UCLA is averaging 7.4 yards per attempt on the ground while Houston is allowing 5.2. UCLA is scoring 43 points/game, Houston is allowing 43 points/game. Do you see where this is going? Johnathan Franklin might reach 1,000 yards rushing in just three games. That's probably a record, or will be until Shoelace racks up 1,500 yards this weekend against the Minutemen. 

Two schools will be leaving a Bowling Green this weekend to play in-state rivals. There is a 9% chance that both teams will win this weekend. Twice since the 2009 season have these two schools won in the same weekend, both in the last half of 2011. Someone, somewhere will find that noteworthy 

You might be thinking that Pitt is one of the worst teams in the country, but by the transitive property we know they are significantly better than Valparaiso; Valpo lost to Youngstown St by 59 last weekend. That's something, right? (PS, Arizona St is scoring 54 points per game; more on that later.)

Florida St is second nationally in points/possession and points/possession allowed. The Seminoles jumped up in points/possession when they were able to rack up some early points against Savannah St and then call the game before they had to stop scoring. I they manage 4.8 points/possession against Wake I'll be really impressed (it won't happen). Seven of the top 9 teams nationally by points/possession are in the Big 12. The other team is Arizona St, who is completing 80% of their passes and getting almost 15 yards per completion. I don't think they'll be able to keep that up all season. I definitely know Pitt couldn't keep that up all season . . . or do it at any point in the season.

Schedule and title odds: Charting the best route to the MNC

The three keys to maximizing your probability of playing for the MNC are 1) be really, really good, 2) play enough tough games to set your team apart from the competition, and 3) don't play a schedule that is so tough you have little chance of getting through without a few blemishes on the resume. The better your team, the tougher your schedule can be and, therefore, the bigger your scheduling window between too soft and too tough.

To get a better idea of what ideal scheduling windows look like I ran two teams, one really, really good team (Alabama, power rank #1) and one really good team (Georgia, #5) through 20 other team's schedules 1,000 times (40,000 total simulated seasons) to see how many times they ended up in one of the two top spots in the national rankings.

Looking at the table, UD (undefeated) and BE (bowl eligible) are two ways of measuring the strength of a schedule. A team's UD will be relatively high if they face one or two elite opponents. Ole Miss, Auburn, Mississippi St and Texas A&M have the top 4 UD schedules nationally. A team's BE is relatively high if it plays a number of solid opponents. Kansas, Texas A&M (two FCS opponents means A&M must win 7), Iowa St, Oklahoma and Baylor take up the top 5 spots. (Teams with 9 conference games tend to score high on BE.)

This is what Georgia fans see when they look at FSU's schedule

PR is the relatively probability Alabama or Georgia would play for the MNC in that other schedule versus their own schedule. For example, Georgia would be 3.94 times more likely to play for the title in Cincinnati's schedule than in their own. They would be almost half as likely in Ole Miss's schedule (though I assume this is not seriously capping Ole Miss's national title aspirations). If Cincinnati and Ole Miss were as good as Georgia (and they are not), Cincinnati would be more than 7x more likely to play for the MNC than Ole Miss. That's a big deal.

Of the leading contenders in the list (Alabama, USC, Oklahoma, Florida St, Georgia), Florida St has by far the best path to the MNC. Both Alabama and Georgia would significantly improve their chances in Florida St's schedule. Georgia would also be significantly better off in USC's or Oklahoma's schedule (or even Michigan St's schedule). So, the argument that Georgia is ranked high because they have a relatively nice path this season is a bunch of bunk (even if they do have a pretty nice schedule compared to the rest of the SEC).

In general, a team's title chances are highest when the BE is high relatively to the UD. Both Georgia's and Alabama's MNC game odds were negatively correlated with UD - the higher the UD the less favorable the schedule - and Alabama had a strong positive correlation with BE/UD, so that the higher the BE relative to the UD the better Alabama's title game chances.

Quest for perfection, 2012

In 2008, the BYU football team took on the mantra "Quest for Perfection". TCU ended that quest emphatically, beating BYU 32-7. Instead, rival Utah would find perfection that season (shockingly, since they weren't even questing after it).

In 2012, BYU is still perfect after two games (and Utah is not). The model is giving the Cougars a 1% chance of finishing the season unbeaten, but if they can stay clean over the next week (games at Utah and at Boise St), perfection may be quest-after-able once again.

The model gives 9 teams a better than 3% chance of going undefeated through the regular season and four teams a better than 10% chance. Alabama leads the pack, followed by Georgia and Ohio; these teams have already cleared legitimate hurdles. The next group includes USC, Oklahoma, Florida St and Louisiana Tech - all clear favorites in their conferences by my estimation. They are followed by LSU and Oregon, both very good teams but a step behind the conference leaders.

Statistically, we could have a max of 10 undefeated teams this season (e.g., Alabama and LSU cannot both go undefeated), down from 18 before the season kicked off. More realistically, using 1,000 season simulations, we should expect 1.48 undefeated teams through the regular season.

 

Week 2 EP3s

The EP3 (EPPP or Effective Points per Possession) is a way of measuring offensive and defensive performance that considers the effect of a possession on field position as well as points on the scoreboard. I still need another week or two to adjust EP3 for the strength of the opponent and, with that, use the EP3 to calculate power rankings, but for now we can look at raw offensive and defensive performances. (Michigan and Denard Robinson can attest to the importance of the opponent on a team's offensive performance.)

The top offensive performances of the week came from Texas Tech, who proved that they are still really good at plastering bad teams and that Texas St. is still a bad team, and Florida St, who proved nothing about themselves but did help to show that the mercy rule is still in effect in college football.

At the other end, SFA taught us that 466 yards won't help if you turn the ball over 10 times. TEN TIMES. New Mexico, Western Kentucky, Duke, UConn, Army, and even Washington were in the just-punt-it-away zone, where the offense probably would have been better getting off a surprise punt than running real offensive plays and risking a turnover.

Tracking Attrition, 2011 to 2012

Good programs develop players, bad programs don't. Attrition from year to year can set a team back. Losing those players before they have used up their eligibility can keep a program from reaching its potential. Below, I've listed for each team the make-up of the team by class and the percent of players by class in 2011 that are no longer on the roster in 2012 (ignoring seniors, of course).

A spike among freshman and, to a lesser extent, sophomores is often a sign of a coaching change: see Akron, Arizona St, Arkansas, Florida Atlantic, Kansas, Ohio St, Pitt, Southern Miss. Alabama sent half its team to the NFL.

[I am using official team rosters, and some teams have not yet updated their rosters from 2011. I have left these teams blank.]

What we learned, week 2

1) We are underrating a) Bobby Petrino, b) Tyler Wilson and/or c) ULM. We also learned that if you never bother to protect your quarterback he will eventually break his collarbone (or get a concussion, or something).

2) We were underrating a) Russell Wilson, b) the Oregon St defense, and/or c) the epidemic that is Randy Edsallitis.

3) We were overrating a) the Oklahoma St defense, b) the Auburn offense, and c) football in the state of Pennsylvania.

4) We learned just how far the Nebraska defense has fallen and just how bad of a coach Neuheisel really was.

5) We did not learn that the SEC is fallible, because losses to Northwestern and ULM will be quickly written off and forgotten. Instead, people that didn't watch the game will point to a 3 point win by Florida and miss the fact that A&M was the more physical team in that game (and not the team faking injuries because they couldn't keep the tempo).

6) We learned that Pac-12 membership has not elevated Colorado and Utah to new heights.

7) Anyone that said USC has the best starting 22 in the country before the season started needs to have their  pundit card revoked. They will get it back if and when USC scores 13x as many points as Washington (41/3=13.67).

8) Either the UConn defense is really good or we are overrating Tennessee a bit.

Week 1 EPAs, RB and QB

The computer is crunching week 2 numbers as we speak, but in the meantime we take a quick look back at week 1. Brent has finished grinding through 10,000 rows of play by play data to bring us running back and quarterback EPAs. The EPA and NEPA methodologies are still works in progress, but when perfected they will arguably represent the best possible reflection of a player's performance.

In short, Brent attaches a point value to each play based on the change in the situation - down, distance, field position, possession - using the expected points for the team before and after the play.

At running back, the week 1 champion was Air Force's Cody Getz, who rushed for 218 yards and 3 TDs. Taylor Martinez led the pack at quarterback. 

Tooting my own horn and week 2 picks

Pictures are worth a thousand words, especially when  they show
W&M shutting out Maryland with 10 minutes left in the game

First, I'd like to point out that I said before the season that Alabama was easily the nation's best team and I did not have a kind word for Maryland. For week 1 I highlighted Idaho/Eastern Washington as my FBS/FCS match-up to watch and also said that Wisconsin/UNI could be interesting. I also noted, before the game, that Savannah St is terrible and would lose by 100. Maybe there's something to this voodoo witch-doctoring after all.

This week two games rise above the rest. First, Missouri will take on Georgia and, after watching SC and Florida in week 1, this could determine the SEC East. Overall, a less than stellar performance from the Bulldogs was more than offset by a 13.6 NEPA**, 1.2 NEPA/play performance from Todd Gurley - who is again proving that ironic names always work in football. To put the 1.2 NEPA/play in perspective, the entire team had an offensive NEPA/play of about .3 and they scored 45 points and won by 22. We learned little about Missouri except that they look really good when the other team refuses to keep the ball or tackle. Georgia is giving points and my model is calling it a pick'em. Given that I picked Georgia to win the East from the beginning, I'll go dogs over cats.

The only question remaining is if Johnny Football will
produce the right photo id when he claims his Heisman

The second game has another SEC East team playing a new SEC opponent. Texas A&M will get a taste of the SEC while Florida will experience Kyle Field. Muschamp has named Driskel the starter at QB and the stats agree - Driskel's 3.2 NEPA was not overwhelming but far better than Brissett's -.3. A&M is favored in this one and the model likes the Aggies by a full touchdown. Personally, I think Christine Michael will rush for 150 yards (ironic name can't go wrong), Trey Williams will bust a couple jaw-dropping 70 yard touchdown runs, and Manziel will literally make our heads explode, but I might be a bit too pessimistic about new coach Kevin Sumlin's offense.

** NEPA - net expected points added.
I'd like to thank Brent Blackwell for his work on the NEPA statistics.

Effective Points/Possession week 1

As the season goes on, I will be reporting the Adjusted Effective Points/Possession. For now, with only one game under our belts, I will just be reporting the raw EPPP. Effective Points/Possession is the average point value of a possession; this is not just the number of points scored but the value of the field position gained or lost. This is a new stat for me, but it is quickly becoming one of my favorites.

The top performance by offensive EPPP in week 1 came from Air Force on their way to beating Idaho St 49-21. Air Force scored 7 touchdowns, punted twice, missed a field goal, and ended the game. More impressive when you consider the opponent, Nebraska scored 7 touchdowns, missed 2 field goals, punted once and ended the game. Youngstown St comes in at #4; that's all I've got to say about that.

At the other end North Carolina turned Elon over 4 times and held them under 20 yards on all but two possessions. Missouri scored a couple defensive touchdowns. Mike Leach managed only -2.2 EPPP against his alma mater in his debut with Washington St. And Stanford, a point/possession leader the past few seasons, rung up -.5 EPPP against San Jose St.

Curious stat of the week, week 1

On Nov. 26, 2011, with under a minute left in the game, down 42-14, Auburn's Clint Moseley fumbled the ball at the Alabama 32. Turns out that was a big deal. Why? Because no other offense would be able to finish a drive on the other side of the Alabama 49 yard line for another 80 minutes of game play. LSU stormed through the 2011 season like Napeleon through Europe only to falter at midfield in the BCS title game. Instead of using scorched earth and light cavalry raids like the Russians, Alabama stopped LSU by pounding them mercilessly into the ground. Eight months later, Michigan would set out for the land of milk, honey and offensive points, but would wander, not for 40 years, but for more than 20 minutes in the wilderness on their own side of the field.

Between 2007 and 2011, two other teams, Boise St and USC (against New Mexico St and Washington St, respectively), prevented opponents from ending drives on their side of the field. Three times in that period Ohio St. kept teams from ending a drive past their 38 yard line. Eleven teams have kept opponents out of their side of the field for at least 3 quarters before finally submitting in the 4th. TCU has done that 3 times.

Fifteen teams found the end zone in every game between 2007 and 2011. One of those is Texas Tech, whose former coach was just locked out of the end zone for an entire game for the first time since 2006.

By my count, 8 times since 2007 has a game between two FBS opponents ended without either team getting into the end zone. Can you name them?

Reverse engineering Vegas and the NFL

I don't gamble, but I love Vegas. Sports gambling creates a semi-liquid information aggregator. Like the stock market, people make gambling decisions based on the information they have. The more confident they are in that information, the more they will bet. The price goes up or down to reflect the general direction of all this accumulated knowledge. As long as people are rational (which, of course, is not always the case in sports*) odds in Vegas, like prices in the stock market, should reflect all available information (see efficient market hypothesis). So, even though people are not forced to disclose their information, the consequence of that information is made public.

And I can use that.

I started with NFL win total over/unders. Using those numbers and team schedules, I estimated team power rankings (T), strength of schedule (O) and game-by-game probabilities. With those game-by-game probabilites, I calculated playoff (P=Playoff, WC=wild card) and division title probabilities. In short, I used 32 numbers generated by a sports gambling market to map out the NFL season.

[The power rankings are scaled to point margins, so to estimate an average margin of victory subtract O from T and add or subtract 3 or 3.5 points for home field advantage.]

* On that note, these specific win totals add up to 262, which is remarkable since there are only 256 games . I assume that, because most fans are irrationally optimistic about their own team and they are also more likely to place a bet on their team, win totals are impossibly optimistic as well.

What I learned this weekend, week 1

This weekend I learned the flu is a beast. They say it can feel like you've been hit by a truck. I've never been hit by a truck, but I have been hit by a car, and this illness has been much worse. (Attempting to block All-Pro DT Ty Warren every day in practice for years is also worse than getting hit by car, in case you were wondering.)

Since I am responsible for all daily maintenance of the site, and since I've only recently risen from the dead, things might be a bit delayed for the next few days.

I also learned that reports that the Big 10 and SEC East are catching up with the SEC West were much exaggerated. I learned that Urban Meyer will be dominating the Big 10 faster than I previously believed.