Thursday, November 19, 2009

Does the Big XII North Champion Have Any Chance?

The Pac10 is shaping up for a great finale with a potential six-way tiebreaker.  The SEC is set for Florida-Alabama.  The Big East is bearing down on Cincinnati-Pittsburgh for the conference title.  Then, there's the Big XII where Texas will play the best of limited options from the Big XII North.  As of today, the only teams still in the running are Nebraska (7-3, 4-2) and Kansas State (6-5, 4-3).  Clearly, anything is possible in a one-game series, as Darren Sproles, Tavita Prichard, Ian Johnson, and Armanti Edwards can attest.  Under what circumstances is it possible for either of these teams to beat Texas?

Nebraska (7-3)

Pros:  Nebraska has one of the most dominant defenses in college football.  In fact, when adjusted to an average opponent, Nebraska commands the #1 scoring defense in the country at 7.1 PPH, giving them the best possible opportunity to slow down Texas's #3 ranked offense.  They are extremely good at taking the ball away, averaging 1.9 takeaways/100 plays.  They should be able to disrupt Texas's offense and force them into playing an ugly game, reducing the gap that Nebraska's mediocre 14.8 PPH offense needs to overcome.

Cons:  Unfortunately, Nebraska is tied with Texas for the #1 defense in the country.  Their offense has struggled so mightily against quality opponents, that we actually expect them to get shutout before considering Texas turnovers and penalties.  Nebraska's offense turns the ball over 1.8 times per 100 plays, consistently killing their offensive momentum.

RBA Prediction:  Texas 24 - Nebraska 3.

Best Case Scenario:  If Nebraska can play flawlessly on offense by eliminating their turnovers and penalties, they can make up about a field goal in this game, leaving them at an 18-point disadvantage.  They need to have a career day from RB Roy Helu, Jr. to reduce Texas's offensive play count while the Nebraska defense gets after QB Colt McCoy.

Kansas State (6-5)

Pros:  The Wildcats have improved quite a bit under Bill Snyder and have shown to have a consistent offense, demonstrating 12.8-14.7 PPH across all ranges of competition.  They haven't played anybody with Texas's strength, but there's no current indication that they will have trouble scoring some points against Texas.  They are also very opportunistic on defense, averaging 2.3 takeaways per 100 plays.

Cons:  Aside from their ability to generate turnovers, Kansas State's defense is terrible.  In fact, the slope on their defensive statistics is the highest of all teams in FBS, implying that they have the worst defensive slip against quality competition.  RBA suggests that Texas will be able to put up a touchdown every 10 plays, and Texas plays fast.  Texas should score at will and empty the benches early in the third quarter.

RBA Prediction:  Texas 77 - Kansas State 20

Best Case Scenario:  I think RBA is a little overzealous by picking Texas to hang 77 on Kansas State because Mack Brown would call off the dogs long before that to avoid injuries for the BCS Championship.  Kansas State needs Texas to melt down offensively, gift wrap a few turnovers, and forget how to play rush defense.  I know that anything is possible, but this particular upset would rank up there with Stanford-Southern Cal in 2007.  Honestly, the best thing that can happen to Kansas State is losing to Nebraska this Saturday.

Wednesday, November 18, 2009

Week 12: Wednesday Predictions

From here on out we'll be posting our predictions for the day's games at noon, Pacific Time. My predictions will be in blue and Eddie's will be in yellow. The numbers next to each team are their rankings in our respective systems. In the future we plan to merge the tables a bit to highlight where we're in agreement and we think each other are off-base.

With that, here are tonight's games.

Home Visitors Odds
(92) Ball St.24 (53) Central Michigan31 75.0
(109) Miami-OH27 (85) Buffalo31 76.1

Home Visitors Odds
(114) Ball St.17 (45) Central Michigan38 87.3
(111) Miami-OH20 (96) Buffalo21 58.1

Tuesday, November 17, 2009

TFG Gets a Schedule

We're going to do things slightly differently starting this week.  Each day where there's a game we'll be posting our picks that morning.  In addition, we're establishing a regular schedule from now until the end of the season.
  • Monday: New rankings and commentary on our rankings
  • Tuesday: A look at the BCS picture at non-"Big Six" teams
  • Wednesday: Discussing the ACC and Big East
  • Thursday: The Big 12 and Big 10
  • Friday: The SEC and Pac-10
This will be in addition to whatever commentary or subjects happen to inspire Eddie or me at the time.  Mark your calendars!

Introducing Regression-Based Analysis

I'd like to thank Justin for the opportunity to share my competing algorithm with the TFG readers. I have taken a different path deliberately to explore different ways of predicting college football games. I'm a big fan of the tempo-free statistics, but I thought of several components that Justin's algorithm implicitly neglects. In particular, the TFG method focuses solely on teams' average performance rather than their variances. My regression-based algorithm (RBA) explicitly models a team's offensive and defensive performance as a function of opponent strength and places a premium on turnovers and penalties, which can turn the tide of a single-game contest. (See Arkansas-Florida 2009, where Arkansas forced four turnovers to get within striking distance and got absolutely boned by penalties late in the 4th.)
I agree with Justin that all statistics need to be scaled by the number of plays to generate a tempo-free statistic. Unless otherwise specified, all the statistics in this description may be computed as points per 100 plays. The last step of the algorithm will be to scale the tempo-free point total by the expected number of plays.
Estimating Offensive and Defensive Efficiency
My algorithm begins by computing offensive and defensive statistics as a function of opponent strength. For simplicity, I will only discuss the offensive efficiency because the same computations are used to compute the defensive efficiency. I approximate the offensive efficiency using least-squares approximation because it minimizes mean-square error. We model the offensive efficiency as the linear equation Y = m*X + b, where X is a random variable representing the opponent's strength and Y is the offensive efficiency. We use least squares to estimate the parameters (m,b).
The following figure illustrates Auburn's offensive and defensive efficiency. The horizontal axis indicates opponent strength. The vertical axis shows efficiency per 100 plays. Each cross represents the measured offensive efficiency in a game. Each circle represents measured defensive efficiency. These blue and red lines show the least-squares approximation of the offensive and defensive efficiencies, respectively. From this plot, we observe the expected behavior. Auburn's offense scores fewer points against stronger opponents and their defense allows more points against stronger opponents.
Let's contrast this performance with Florida. Florida's performance is relatively constant across opponents of all strengths, implying that they impose their offensive and defensive will against all their opponents.
Impact of Turnovers, Takeaways, and Penalties Turnovers and penalties represent behavior synonymous with sloppy play. Conversely, takeaways indicate exceptional play. In general, we expect the former to reduce an opponent's efficiency and the latter to improve an opponent's efficiency. For simplicity, we will restrict our discussion to turnovers, as the same analysis is used for takeaways and penalties. To determine the impact of an event on a team's score, we compute the difference between the team's score and its mean score. For example, if a team scores 21 points in a game but averages 27, we indicate an impact of -6 points. I started out simply using the points per turnover as a metric: points/turnover * turnovers/play * plays = points/play. However, this approach is full of mathematical FAIL. Actually, we observe an increase in efficiency when a team has only one turnover because even great teams tend to commit at least one turnover a game. To accommodate this behavior, we model efficiency as a joint probability distribution of turnovers and points. The following figure illustrates the points per 100 plays a team loses for committing a given number of turnovers.
The solid line indicates the mean number of points. We assume a Gaussian distribution for a given number of turnovers. The error bars indicate one standard deviation from the mean (68% of outcomes lie within this range).An interesting observation is that turnovers are not a good estimator when few turnovers are committed. This makes sense because if we have few turnovers, the game's outcome must have been determined from other sources. We also observe a general downward trend as the number of turnovers increases because, as expected, turnovers negatively impact a team's performance. A small rise occurs at seven turnovers because few samples in this range.
Generating Final Scores The final score may be computed by summing all the random variables in the RBA: score = ((Po + Pd)/2 + Ph + Pto + Pta + Pn) * N where Po = offensive points/play, Pd = opponent's defensive points/play, Ph = home field advantage (or away field penalty), Pto = turnover derating, Pta = takeaway bonus, Pn = penalty derating, and N = number of plays.
Due to the central limit theorem, the summation of a series of random variables converges to a Gaussian distribution. Furthermore, the summation a*X1+b*X2 of two Gaussian variables N(m1, v1) and N(m2, v2) is a Gaussian N(a*m1+b*m2, a^2*v1+b^2*v2). From these two computations, we may compute a closed form solution for the final efficiency. For simplicity, we only use the mean for N instead of its probability distribution. (Frankly, the math requires over 14 pages of dense mathematics to explain.)
Conclusion
That's how the RBA algorithm works. I'm still relatively new to the practice of teaching my computer to pick college football games, so we can expect a few tweaks here and there throughout the remainder of the season and during the offseason. Your suggestions are welcome!

A Second Point of View

The Tempo-Free Gridiron is getting a second author to help provide another take on college football.  Eddie is a diehard college football junkie with an encyclopedic knowledge of the game.  He's the complete opposite of myself in that regard, as I have no allegiance to any particular team or conference.  I do this out of pure curiosity for the statistics and predictions with no regard to the names on the jersey (or boxscore).  As you'll see, Eddie starts with the exact same data set that I use, but his system attempts to account for more factors out of the gate.  This should be an interesting contrast in styles: blank slate versus encyclopedia, and simple versus comprehensive.

In his introductory post Eddie will give you the under-the-hood look at his methodology along with his current rankings and predictions for upcoming games.  Over the next week we hope to do per-conference breakdowns and a look at the BCS picture, along with some compare-and-contrast posts that highlight differences in our systems and their results.

Welcome, Eddie.

Sunday, November 15, 2009

Week 12: Full Rankings

Biggest jumps: Nevada (+22, 67 to 45); Rutgers (+16, 42 to 26); Stanford (+8, 32 to 24); South Carolina (+8, 55 to 47); UCLA (+8, 64 to 56).

Biggest drops: South Florida (-16, 48 to 64); Texas A&M (-10, 51 to 61); Arkansas State (-10, 89 to 99).

Full rankings after the jump.

Week 12: Top 25

Another week, another set of new rankings.  I'm making a point of getting these in before the BCS polls come out.

Rank
+/-
Team
WinPct
SoS
Off Pts
Def Pts
Pace
001
+1
Texas
0.9562
0.5365
29.4
9.4
85.9
002
+1
TCU
0.9544
0.4982
25.4
8.2
84.2
003
-2
Florida
0.9533
0.6201
28.8
9.4
78.3
004
+1
Alabama
0.9479
0.5916
24.4
8.3
80.7
005
-1
Ohio St.
0.9305
0.5342
22.0
8.4
82.7
006
--
Oklahoma
0.9261
0.6759
22.7
8.9
90.8
007
--
Boise St.
0.9001
0.3725
26.6
11.8
84.9
008
+2
Virginia Tech
0.8830
0.6152
24.2
11.4
82.0
009
--
Penn State
0.8756
0.4976
19.9
9.6
82.4
010
+1
Oregon
0.8542
0.5957
23.0
11.9
91.7
011
-3
USC
0.8439
0.5816
20.8
11.1
85.7
012
+1
Nebraska
0.8380
0.5400
19.9
10.8
82.1
013
-1
LSU
0.8366
0.5457
20.0
10.9
81.7
014
--
Clemson
0.8249
0.6111
20.3
11.4
84.8
015
+2
Iowa
0.8138
0.5708
20.9
12.1
81.9
016
-1
Texas Tech
0.8064
0.5848
25.5
15.1
90.2
017
-1
Cincinnati
0.7994
0.4951
22.2
13.3
86.0
018
+7
Mississippi
0.7982
0.5939
21.2
12.8
83.1
019
--
Pittsburgh
0.7853
0.5121
22.9
14.2
82.5
020
+1
Oklahoma St.
0.7732
0.6038
23.9
15.2
83.3
021
-1
Utah
0.7696
0.4345
21.7
13.9
85.7
022
NA
Georgia Tech
0.7561
0.6176
27.9
18.3
79.5
023
-5
BYU
0.7426
0.3827
22.4
15.2
83.2
024
NA
Stanford
0.7375
0.5451
25.4
17.3
83.0
025
-1
Oregon St.
0.7354
0.6061
20.8
14.2
87.5

New entries: Georgia Tech, Stanford.

Dropped out: Miami-FL, Tennessee.

The ordering at the top shuffled around slightly as Texas and TCU leapfrog Florida after the Gators put up a lackluster performance against then-(55) South Carolina.  Winning percentage-wise, however, the top three teams are separated by less than three one-thousandths of a point.  TCU's dismantling of previous (20) Utah helped them maintain their power ranking but also improved their lackluster strength of schedule.  Alabama leapfrogged Ohio State after the Buckeyes barely snuck by Iowa.  Oklahoma and Boise State continue to drift down in winning percentage but hold their positions by lack of any solid competitors stepping up to claim the spots.

USC drops a staggering five one-hundreths of a point in winning percentage but only drops three places.  One more game like the one against Stanford and the Trojans will be in the same rarified air as Pittsburgh and Oklahoma State.  Oregon finally leapfrogs USC as the top team in the Pac-10, but Stanford is on a tear.  Over the last three weeks Stanford has climbed an NCAA-best 28 spots in the rankings.  Their nearest competition for "Most Improved" are UNC and Nevada, each climbing 18 spots in the same stretch.  Two more solid games to close out the year and Stanford could easily be a Top 15 team.

Georgia Tech clinched the ACC Coastal Title by pounding a hapless Duke team in Durham.  They appear to be on track for a rematch with Clemson, a team which they barely beat at home at the start of the season.  Were they to meet on a neutral field right now, Clemson would be a 60-40 favorite.  Mississippi claims the "Most Improved Within the Top 25" label for this week, jumping seven spots to 18th after defeating Tennessee at home.

Week 11: Summary

This past week was a slightly better one for my system as it went 44-12, going 3-4 in games where the favorite only had a 50-60% chance of winning.  This brings the four-week total to 166-53, or 75.8% accurate.  This week I'll be examining avenues to improve that further.  But first, the Week 11 recap ....

Games To Watch

(23) Miami-FL at (38) North Carolina.
I said: If North Carolina can keep Miami's offense in check they should be able to get their points.
Result: Miami outplayed UNC, racking up 438 yards to UNC's 326.  However, Miami also threw four INTs and UNC never trailed in this one, pulling off the 33-24 upset in Chapel Hill.

(29) Auburn at (37) Georgia.
I said: The law of "common opponents" says Auburn wins this one in a walk, but my numbers say this is anyone's game.
Result: With nine minutes left this one was knotted at 24, but Georgia got an INT in Auburn territory and drove 39 yards for the game-winning touchdown.  Auburn had a chance to tie it, getting to UGA's 22 yard line before sacks and penalties gave them a 4th-and-23 (they didn't make it).  Prediction: 31-24 UGA.  Result: 31-24 UGA.  You're welcome.

(26) Arizona at (49) California.
I said: Both play at a similar pace, so this is going to come down to one of two things: can Cal dig up some extra defense or can their offense find an extra gear to outscore Arizona?
Result: At the two-minute mark Cal was up 18-16 and Arizona was within field goal range.  What followed was one of those freak plays that no statistics can predict.  Arizona quarterback Nick Foles had his own pass deflected back to him; instead of batting it to the ground he caught the ball and tried to make another (illegal) pass.  Instead of 4th-and-3 at the Cal 25, Arizona faced a 4th-and-17 at the Cal 39.  The conversion failed and on the next play Cal RB Shane Vereen ran 61 yards to put Cal up 24-16.  Cal got the defense they needed and some lucky breaks on the offense to pull off the upset.

Coin Toss Game of the Week

(15) Texas Tech at (21) Oklahoma State.
I said: If [Oklahoma State] can slow down TTU's offense they'll have a shot.  Otherwise they risk getting completely blown out.
Result: The Cowboys held TTU to their lowest point total of the year, including TTU's game at Texas.  That's no small feat, and OSU was rewarded with a win.

Wednesday, November 11, 2009

Pac-10: Crowded at the Top

For once the Pac-10 isn't all about just USC this year.  Currently four Pac-10 teams are in the BCS Top 25, and three of those are in the TFG Top 25 (Arizona peeks in at number 26).  With only a few games left, here are the current Pac-10 standings:

Team
Wins
Losses
Oregon
5
1
Arizona
4
1
Stanford
5
2
USC
4
2
Oregon State
4
2

With a single loss to USC, Oregon sits in the driver's seat a half game ahead of Arizona.  However, there's still a lot of football to play and we could be headed for a very interesting situation.  Let's take a look at each team's remaining schedule and their likely end-of-season standing.

Oregon

The Ducks have had an up-and-down season, getting pounded by Boise State to start the year, absolutely thrashing USC at home, and getting steamrolled by Stanford.  They currently sit at #11 in the TFG rankings with three games to play.

Home
Visitors
Odds
(11) Oregon
31
(57) Arizona St.
24
87.6
(26) Arizona
27
(11) Oregon
31
70.4
(11) Oregon
31
(24) Oregon St.
27
68.6

As it stands, Oregon is the favorite in each of their remaining games and very well could end the season at 8-1 in the Pac-10.  There's a 42.3% chance of this happening.  However, there's also a 43.1% chance that Oregon will win two of their remaining three games.  That would place them at 7-2 in the Pac-10, with the most likely two losses being to Stanford (in the books) and Oregon State.

Arizona

The Wildcats sit a half-game back of Oregon, but have a long road to the end of the year.

Home
Visitors
Odds
(49) California
27
(26) Arizona
31
61.3
(26) Arizona
27
(11) Oregon
31
70.4
(57) Arizona St.
24
(26) Arizona
31
71.2
(8) USC
31
(26) Arizona
24
77.6

Three of their last four games are on the road, and their lone home game is against Oregon.  They have only a 23.8% chance of going 4-0 to finish 8-1 in-conference.  As mentioned previously, it's likely that Oregon will beat Arizona, and even more likely that USC will beat Arizona for the last game of the season.  This would place Arizona at 6-3 in-conference and likely out of contention for a BCS bid.

Stanford

Stanford shocked Oregon last week, nearly doubling up the Ducks in their visit to the Bay Area.  That win propelled the Cardinal into third in the Pac-10, a half-game ahead of USC and a half-game behind Arizona.  However, the road gets difficult from here.

Home
Visitors
Odds
(8) USC
31
(32) Stanford
27
83.6
(32) Stanford
31
(49) California
27
60.9
(32) Stanford
31
(39) Notre Dame
28
54.7

Stanford will get a chance this weekend to prove if they're The Real Deal(tm) by going to USC to play the Trojans.  The odds don't look good, though, as they've got less than a 1-in-6 shot of pulling off the upset.  To be fair, last week they only had a 1-in-5 shot of defeating Oregon yet managed to pull off that upset in stunning fashion.  However at the time Oregon was still on a high from their USC victory and Stanford may have caught them napping.  I doubt they'll get lucky twice.  Even assuming a victory at USC -- which I won't assume here -- they'll have to defeat Cal in order to finish up the year at 7-2; current odds on that combo are 10.0%.  A more likely outcome is a 6-3 finish with a win over Cal and a loss at USC (50.9%) or even two losses (32.7%).

USC

Home
Visitors
Odds
(8) USC
31
(32) Stanford
27
83.6
(8) USC
31
(64) UCLA
17
91.1
(8) USC
31
(26) Arizona
24
77.6

The general consensus is that this year's USC is not the USC of recent history.  Sure, they've been in the top 10 for most of the year and have quality wins over (4) Ohio State and (24) Oregon State.  But there's that pesky "2" in the loss column, coming to lowly (69) Washington and (11) Oregon.  Of all the teams in the Pac-10, though, USC has the easiest road to the end of the season.  Three games at home, with a 59.1% chance of winning out and only a 34.4% chance of going 2-1.  Odds are USC recovers and finishes 7-2 in-conference.

Oregon State

Oregon State is currently tied for fourth with USC, although is technically behind them since USC holds the head-to-head tiebreaker.  OSU has a relatively easy road to the end of the season ... with one exception.

Home
Visitors
Odds
(24) Oregon St.
38
(69) Washington
24
82.4
(117) Washington St.
17
(24) Oregon St.
38
97.5
(11) Oregon
31
(24) Oregon St.
27
68.6

Wins are likely against both UW and WSU (80.3%), leading to a season finale against Oregon.  This is where it gets interesting.  OSU is the underdog in this game, especially since it's on the road, but if they can pull off the upset -- as discussed in the Oregon section -- then OSU will win out and go 7-2 for the year.  At which point the Pac-10 standings will look like this:

Team
Wins
Losses
Oregon
7
2
USC
7
2
Oregon State
7
2
Arizona
6
3
Stanford
6
3

Arizona will hold the tiebreaker over Stanford thanks to the Wildcats' 43-38 victory over the Cardinal a few weeks ago. But the OSU-USC-Oregon situation will be more difficult to resolve.
  • OSU will have defeated Oregon but lost to USC
  • USC will have defeated OSU but lost to Oregon
  • Oregon will have defeated USC but lost to OSU
If I'm reading the Pac-10 tiebreaker rules correctly, then it will come down to the result of each team's game against fourth-place Arizona.  Arizona defeated OSU but -- in our scenario -- will have lost to Oregon and USC.  This eliminates OSU from the picture, and we repeat the tiebreaker scenario with only USC and Oregon, which would result in a Rose Bowl bid for the Ducks.

Is this is going to happen, though, it'll happen one game at a time.  Namely, this weekend
  • Oregon must defeat Arizona State
  • USC must defeat Stanford
  • Oregon State must defeat Washington
The current odds of all three happening are 60.3%.  We'll find out on Saturday evening if we're one step closer to breaking out the Pac-10 tiebreaker rulebook.

Week 11: Predictions

Last week was a bad one for the system because it didn't break 70% after approaching 90% three weeks ago.  Hopefully the tweaks from this past week will help.  With that I bring you to

Games To Watch

(23) Miami-FL at (38) North Carolina. The numbers give an edge to "The U" but not by much: 57-43.  UNC is 2-1 in their recent games, including a narrow loss to (35) Florida State and a solid win on the road at (10) Virginia Tech.  Miami, on the other hand, is 2-1 with a home loss to (14) Clemson and a skin-of-their-teeth win to (60) Wake Forest.  Their 52-17 demolition of UVa on the road is a bit deceptive, as UVa has been in a tailspin since mid-October.  If North Carolina can keep Miami's offense in check they should be able to get their points.

(29) Auburn at (37) Georgia.  Auburn is 3-3 since the start of October, with wins over I-AA Furman, a squeaker over (22) Tennessee, and a good win over (25) Mississippi.  Georgia is 2-3 in that same stretch with wins over I-AA Tennessee Tech and (94) Vanderbilt.  Their losses, though, have been to (12) LSU, and on the road at (22) Tennessee and (1) Florida. The law of "common opponents" says Auburn wins this one in a walk, but my numbers say this is anyone's game.

(26) Arizona at (49) California. This is one that could have important implications for the Pac-10, an issue that I'll cover in a later post.  Arizona and Cal have similar defensive ratings (15.4 and 16.1) so it'll come down to offensive production.  Arizona's offense has been much more efficient, rating 22.1 to Cal's 18.2, good for about 6 points per game.  Both play at a similar pace, so this is going to come down to one of two things: can Cal dig up some extra defense or can their offense find an extra gear to outscore Arizona?

Coin Toss Game of the Week

(15) Texas Tech at (21) Oklahoma State. This one falls under the category of both "Games to Watch" and "Coin Toss Game of the Week".  This is another 50.2-49.8 game with Texas Tech taking the slight edge.  Oklahoma State hasn't fared well against high-powered offenses, with TTU tipping the scales at an efficiency of 28.0, good for third in the nation behind Texas (29.4) and Florida (30.7).  The TTU defense is another issue; think of TTU as the Texas offense combined with Vanderbilt's defense.  The OSU defense is good but not excellent, so much of their hope rides on being able to get a boost from their fans at home.  If they can slow down TTU's offense they'll have a shot.  Otherwise they risk getting completely blown out.  The CTGotW games have been anything but close, so we'll see if I junxed this one.

Full predictions after the jump.  The odds say I get 42 of these 57 right for a 73.4% accuracy rate.  I'm already 1-1 so far, as Ohio's FG with 1:25 left in the game ruined my perfect week.