Several teams who were in prime position for a playoff spot one month ago struggled greatly in December and now find themselves at home. Four teams–Dallas, Denver, New York Jets and Tampa Bay, lead the way with their late season performances. For Dallas and Denver, this is nothing new. As Doug noted in this post from two summers ago, since 1978, Denver has the largest drop among teams between their winning percentage in the first 12 games of the season, and the last 4 games. Dallas was not too far behind.

I’ve got some thoughts on this that I want to explore further when I have time in the offseason, but it basically has to do with the fact that these teams play more “long” road trips than anyone else. Denver plays virtually every game in a different time zone, and the closest road trip is over 500 miles away, to Kansas City. Dallas, for those that may not realize it, is not in fact located in the Eastern part of the United States. They play three divisional games at a great distance from home, and unless Houston or New Orleans happen to be on the road schedule, also don’t play any teams within 500 miles from home. This year, Dallas’ closest road game was in Saint Louis. Perhaps these constant long trips (while teams from the East get to play at least half their games within a stone’s throw of home) have a cumulative effect that results in reduced performance late in the regular season.

The other issue that I wanted to examine was teams with older starting quarterbacks. We’ve seen the Bucs with Jeff Garcia finish 0-4 after a 9-3 start (though it had as much to do with defensive collapses as Garcia), and also got to witness the complete collapse of Brett Favre with the Jets. This got me thinking, is there evidence that teams with older quarterbacks hit a wall. So I pulled every non-strike season since 1978 in which a quarterback who was age 35 or older threw at least 350 passes for a team, and recorded the team’s record over the first 12 games, and the final 4 games, to see how these teams did down the stretch. Basically, I wanted to see if the Jets/Favre example was random, or if other older quarterbacks had led teams that faltered down the stretch. Without further ado, here are the results, sorted by age of the quarterback:

AGE		No.	F12	L4		DIFF
=========================================================
39+		6	0.542	0.333		-0.208
38		11	0.644	0.466		-0.178
37		12	0.625	0.604		-0.021
36		18	0.588	0.528		-0.060
35		20	0.558	0.638		0.079
=========================================================

Just so you understand what you are looking at here, Teams with a primary starting quarterback who was 38 years of age had a winning percentage of 0.644 over the first twelve games, and a winning percentage of 0.466 in the final four games, for a sizeable drop off of 0.178 compared to the rest of the season.

I had originally prepared this last week, and after I had written it, I saw, from reviewing Brian Burke’s weekly roundup this weekend, that Bill Barnwell over at Football Outsiders wrote on a similar issue, and I encourage you to go check that out as well for a different perspective.

We’ve got a lot of big name quarterbacks on this list, and we can see that for the most part, they played on teams that were winners for most of the season. Here is the complete list of quarterbacks who threw 350+ passes in a season at age 38 or older, accompanied by their team’s record in the first 12 and last 4.

player	team	year	age		W	L		W	L	T
==========================================================================
moon		sea	1997	41		6	6		2	2
testaverde	dal	2004	41		5	7		1	3
favre		nyj	2008	39		8	4		1	3
flutie		sdg	2001	39		5	7		0	4
moon		min	1995	39		6	6		2	2
simms		nyg	1993	39		9	3		2	2
johnson		min	2006	38		5	7		1	3
favre		gnb	2007	38		10	2		3	1
testaverde	nyj	2001	38		7	5		3	1
montana		kan	1994	38		7	5		2	2
moon		min	1994	38		7	5		3	1
tarkenton	min	1978	38		7	5		1	2	1
garcia		tam	2008	38		9	3		0	4
marino		mia	1999	38		8	4		1	3
elway		den	1998	38		12	0		2	2
krieg		chi	1996	38		5	7		2	2
morton		den	1981	38		8	4		2	2
==========================================================================

Now, it is entirely possible that these teams played more difficult schedules or the quarterbacks were injured, or the drop in wins was due to some factor other than a change in the quarterback’s level of play–we are talking about only 17 guys (actually, 13, as Testaverde and Favre appear twice, and Warren Moon three times). Still, I pulled the game by game data, and for the quarterbacks before 1995, used the available games started data, box scores, and overall team passing stats to pull educated guesstimates of statistics for the final month. Twelve of the quarterbacks started all four of their team’s final four games of the regular season; Moon in 1994 and 1997, Johnson in 2006 and Garcia this year played three, and Montana missed two of the final four games. Here are the splits from the early games versus the last four games in terms of completion percentage, yards per attempt, and adjusted yards per attempt, for the twelve that started every game the final month.

	Comp%	YPA	Adj YPA
=====================================
EARLY	61.3	7.16	6.11
LATE	58.3	6.69	5.24
=====================================

We see that, as a group, these twelve quarterbacks showed some decline across the board, some of this certainly could be the result of weather issues. The drop was most dramatic, though, in the adjusted yards per attempt category, primarily because of increases in interception rate. Brett Favre’s collapse rates as the largest, but Tarkenton (back when league wide interception rates were much higher) also threw a whopping 14 interceptions in his final four regular season starts in 1978. Simms, Krieg, Flutie, and the ageless Warren Moon stayed pretty consistent. Marino and Testaverde limped to the finish line. And to bring it full circle, Craig Morton led Denver to an 8-3 start in 1981 and yet they missed the playoffs when they finished 2-3 and Morton was hurt and struggled. Who’d they lose out to? San Diego, on a tiebreaker.

Back when their were only two divisions in the NFL and only the winners of each division met in the championship game, playoff tiebreaker rules were not all that necessary. On the occasions when two teams would tie for a division lead, they would simply push back the date of the league championship game by a week, and have a one-game playoff between the tied teams, and let them settle it on the field. But now, with so many more teams and divisions, vastly different schedules, many more playoff teams, and set schedules for games to occur thanks to television, we can’t simply delay the playoffs and have play-in games to break ties.

The first, and only, case of a tiebreaker deciding a playoff spot prior to the AFL-NFL merger was a notable one. In 1967, the Baltimore Colts entered the final week of the regular season at 11-0-2, trying to become the first team to go undefeated since the 1929 Green Bay Packers. However, they had to travel to Los Angeles to face the 10-1-2 Rams, their Coastal Division rival. The Rams won that game, and the tiebreaker, on net points scored in the two head to head matchups, and advanced to the playoffs. The Colts, despite tying for best record in the league, stayed home.

After the merger, the occasional division tiebreaker came into play, but conference tiebreakers were rare because, until 1975, the playoff seeds were determined by a set rotation for home games, and were not based on record. The tiebreakers became far more common once the league went to a seeding system for the conference playoffs, then added a wildcard game in 1978. Using the official tiebreaker explanations contained in The NFL Record and Fact Book, I have recorded every tiebreaker that has been used to determine either a) a division winner, b) finish within a division for potential wildcard spot, c) seeding within a conference among division winners, or d) seeding within a conference among potential wildcard eligible teams. Every potential tiebreaker was recorded with a couple of things in mind. First, all ties within a division are broken before ties are broken across divisions. Second, if a 3-way tie (or more) can be broken affirmatively, that is, by one team winning the tiebreaker outright in a category, then the remaining teams revert to a new tiebreaker, even if the second team was ranked ahead of the third team in that category. Only if the top two teams tie in a tiebreaker and the third does not is it broken negatively, that is, by kicking out the third team and re-running the tiebreaker again with only the top two teams. Third, if, after a 3-way tiebreaker was decided, there were no more available playoff spots, then I did not further break the tie between the remaining teams. Thus, I am only looking at ties that had a material impact on the playoffs, and not those that merely determined division finish for scheduling purposes the following season.

Inspired by this post, where Doug looked at the question of whether head to head was the right tiebreaker for college football by building a model, I wanted to check the various tiebreakers used in the NFL to see which ones actually appear to be better. To do that, I’m going to use both Simple Rating System regular season ratings of the teams (to see how often the “better” team by SRS wins a certain tiebreaker), as well as actual playoff results.

Prior to this season, 116 two team tiebreakers, 19 three team tiebreakers, and two four team tiebreakers have been used to determine a playoff position since 1970. One of the four team tiebreakers was back in the strike-shortened 1982 season; the other was in 2006 when the Giants got the final wildcard at 8-8 over Green Bay, Carolina and Saint Louis. Here are the current NFL tiebreaking procedures. We can now add a tiebreaker for the AFC East between Miami and New England (decided by conference record), for the AFC West between San Diego and Denver (division record), a tiebreaker between the New York Giants and Carolina for the #1 seed (head to head), and between New England and Baltimore for the final wildcard spot in the AFC (conference record).

I’m going to use the following codes to summarize the various tiebreakers that have been used over the years.

H2H= Head to Head Victory
DIV = Better record in divisional games
CNF= Better record in conference games
OPP= Better record in games against common opponents
NPD= Better Net Point Differential in Divisional Games
HPD= Better Net Point Differential in Head to Head Games
NPC= Better Net Point Differential in Conference Games
NPA= Better Net Point Differential in All Games
SOV= Strength of Victory
COL= Fewer Losses by Common Opponents
PRS= Point Rating System

To be honest, I don’t even know how the last two tiebreakers were determined, and they don’t exist anymore. “COL” was used to determine the NFC Central in 1977 between Minnesota and Chicago, but I can’t figure out how the common opponents could have fewer losses for Minnesota, since they are, well, common opponents. And I have no idea what the Point Rating System was, but it was used in 1975 to place Minnesota as the #1 seed ahead of the Los Angeles Rams, in the first year the league seeded the conference teams.

In Divisional tiebreakers, five three way ties have been used. Four were determined affirmatively by head to head record (that is, one team had either a 3-1 or 4-0 record combined against the other two), and one, the AFC East in 2002, was determined in the negative by kicking out Miami based on division record (2-4 versus 4-2 for both the Jets and Patriots). Here is a breakdown of how often each of the remaining tiebreakers were used to break a two way tie between division opponents.

Two Team Division Tiebreakers
===================================
	H2H	18
	DIV 	14
	CNF	12
	NPD	4
	OPP	2
	HPD	1
	COL	1
===================================

Virtually all the divisional tiebreakers have been decided by either head to head, division record, or conference record, with head to head being the most common.

For conference tiebreakers, we have had the two aforementioned four-team ties, and fourteen three-team tiebreakers, all of which were broken by the conference record tiebreaker (12 affirmatively, and 2 by kicking out the bottom team and reverting to a two team tiebreaker). Here is a breakdown of how often each of the remaining tiebreakers were used to break a two way tie between division opponents.

Two Team Conference Tiebreakers
===================================
	CNF	27
	H2H	24
	OPP	12
	SOV	2
	NPC	1
	NET	1
	PRS	1
===================================

Here, we see that the “better conference record” tiebreaker has actually been used more commonly than head to head, and those two tiebreaker steps have resolved 75% of the conference ties. Only record against common opponents has also figured on more than a rare occasion.

Now, let’s see which tiebreakers do the best job (as that is defined by both SRS ratings and actual head to head results in the playoffs) of selecting the better team to win the tiebreaker. I’m only going to focus on those that have come up on more than a couple of occasions, and am not including any results from this season, as the playoffs are still to come. I’m limiting it to Head to Head, Conference Record, Division Record, and Record against Common Opponents. Before I get to the results, see if you can guess which ones are better or worse.

Okay, time’s up.

This table shows the average SRS rating of the tiebreaker winner, tiebreaker loser, the total number of cases, and then shows how many times, according to SRS, the tiebreaker winner was clearly better (SRS at least 1.51 points higher), roughly equal (SRS rating of both teams within 1.5 points), and clearly worse (SRS at least 1.51 points lower), than the tiebreaker loser.

type	Win	Lose	No.		B	C	W
=======================================================================
H2H	4.73	3.99	40		16	13	11
DIV	2.83	2.88	14		4	4	6
CNF	3.36	3.58	37		14	7	16
OPP	2.94	3.13	14		5	5	4
=======================================================================

It came as a surprise to me, at least until I read Doug’s earlier post and thought about the rationale, but the actual results from the last forty years show that head to head is about as good a tiebreaker as there is. For all these tiebreakers, it’s no surprise that the teams are roughly equal. But the head to head tiebreaker winners are the only ones that are, on average, better than the losers. The remainder appear to be no better than flipping a coin. But that is some theoretical stuff, and whether you accept it or not depends on how you view the SRS as a proxy of who is actually better.

Many times, the tiebreaker loser doesn’t make the playoffs, or never gets a chance to settle the actual question on the field. Occasionally, though, a team that loses out on a tiebreaker gets to later go on the road against the other team. We are dealing with small samples, but here are the results (from the point of view of the home team/tiebreaker winner) when two teams meet in the playoffs that had previously been the subject of a playoff seeding tiebreaker:

TYPE		W	L		PCT
==========================================
H2H		7	3		0.70
NPD		2	1		0.67
CNF		10	7		0.59
DIV		2	3		0.40
OPP		0	2		0.00
==========================================

I’ve added in Net Point Differential in Division Games because three of the four times it was used, the two division opponents went on to meet in the playoffs. Again, Head to Head tiebreaker winners fare better than anyone when playing the tiebreaker opponent in the playoffs. The Common Opponent tiebreaker winner is 0-2 (Pittsburgh losing at home to San Diego in 1982, and Philadelphia losing at home to the Rams in 1989). The Conference Record Tiebreaker teams fare well in the wildcard round, but on the five occasions two teams have faced each other in the Divisional or Championship round and the home team was determined by conference record, the home team has gone 1-4.

So, Doug asked the question “Is head to head the right tiebreaker?” a couple of weeks ago, and at least for the NFL, the answer appears to be “yes”. After that, though, the remaining tiebreakers are suspect. The conference tiebreaker sounds right, particularly since we are seeding for conference playoffs. But the thing is, those teams likely did not play the same conference games. With the current imbalanced conference schedules, the only time this tiebreaker would come into play in a two team tiebreaker is when teams are on opposite sides of the conference slate. This one appears to, somewhat more than half the time, reward the team that played the easier schedule.

The Common Opponent tiebreaker needs to go. The only time it comes into play is, again, when teams are on opposite sides of a conference schedule, have tied on overall conference record, and they will thus have five common games. Consider this, team A plays in a division with Powerhouse, team B plays in a division with Cupcake, and team A played Cupcake in the conference schedule, and team B played Powerhouse. If they perform equally in the other two common games, then Team A is at a disadvantage. Both could lose to Powerhouse (Team A twice and Team B once) and beat Cupcake (Team A once and Team B twice), and Team B wins the common opponents tiebreaker.

What should replace it? I know the league is not going to get into a computer formula situation, but why not use some of the important elements of actually determining the quality of a team–namely strength of schedule or overall point differential. I would think that going with either of these would be better than using something like common opponents (which is unfair to the team that plays the better division opponent) or conference games. The league should also consider moving the strength of victory tiebreaker further up the tiebreaker priority list, ahead of some of the others.

I was tooling around my buddy JC’s sabernomics blog the other day and came across this nifty paragraph in the FAQ, where JC simultaneously apologizes for and absolves himself of some occasional sloppiness in his writing:

I like blogging because it is a good way to post my thoughts quickly. If I proofread my posts as much as I wanted to, I wouldn’t post nearly as often. As George Stigler once said, “If you never miss a plane, you’re spending too much time at the airport.”

The Stigler quote is just a variation of Voltaire’s “the perfect is the enemy of the good,” but it hits home with me because I’m an incorrigible unreasonably-early-at-the-airport guy. I do spend too much time in airports, and I know it.

And then on Sunday I watched Adrian Peterson fumble the football and the announcers say, “he’s trying to do too much.” And he was. And, though I don’t watch enough of him to say this definitively, I’m guessing he often does. He has more fumbles in the last two seasons than any other running back in the league. But that must be tied to his being so damn good. You can’t do more without trying to do more. And it’s mighty tough to find the line between more and too much.

The parallel isn’t exact, but…

missed planes —> fumbles

me —> regular RBs

time spent at airports —> all that yardage Adrian Peterson is gaining that regular RBs are not

What if Adrian Peterson’s life literally depended on him not fumbling at all — not even once — during the 2009 season. How many yards do you think he’d gain? Do you think he’d gain a thousand? I don’t. No way. Remember, one fumble and he’s dead.

If you buy that, then you’re agreeing that Peterson’s willingness to fumble occasionally is worth about 800 or so yards a year at least. And even after you subtract the damage caused by his half-dozen-or-so fumbles, that’s a net gain. The point is, the optimal fumble rate for Peterson (or any other RB) isn’t zero. Zero is such a draconian standard that it can’t possibly be worth what it costs.

And the same is true with offenses in general. If Stigler is a football fan, he’d say, “If you never throw an interception, you’re taking too many sacks, throwing too many balls out of bounds, and getting too many four yard gains on 3rd-and-9.”

So if zero is not the optimal turnover rate, then what is? I don’t know, and I don’t think it’s likely that I or anyone else will come up with a study that will convince anyone. But here is a little evidence of the general idea I’m trying to convey.

We all know that good offenses turn the ball over less than bad offenses, right? Here’s the data on all teams from 2000–2007:

+----------+-------+------+
| points   | teams | to   |
+----------+-------+------+
| 274-     |    50 | 32.5 |
| 275--324 |    72 | 30.8 |
| 325--374 |    59 | 26.9 |
| 375--424 |    46 | 26.8 |
| 425+     |    27 | 24.2 |
+----------+-------+------+

So yes. Bad offenses have more turnovers. But that’s looking at it backwards. If you turn the ball over, you didn’t score on that possession.

But what if we just looked at possessions that did not result in points?

+----------+-------+------+------+--------+
| points   | teams | to   | punt | to_pct |
+----------+-------+------+------+--------+
| 274-     |    50 | 32.5 | 88.8 | 0.269  |
| 275--324 |    72 | 30.8 | 81.7 | 0.274  |
| 325--374 |    59 | 26.9 | 78.2 | 0.255  |
| 375--424 |    46 | 26.8 | 70.3 | 0.276  |
| 425+     |    27 | 24.2 | 61.4 | 0.282  |
+----------+-------+------+------+--------+

The last column is turnovers divided by (turnovers plus punts). This table basically says that good offensive teams and bad offensive teams turn the ball over on roughly the same percentage of their non-scoring possessions. But to the extent that there is a difference, the best offensive teams actually turn the ball over on a greater percentage of their non-scoring possessions than the worst offensive teams do.

Here is another look. I took all teams since the merger and I computed their points scored divided by league average, and also their [turnovers divided by (TOs plus punts)], divided by league average. Just for ease of conversation, let’s call a team’s TO/(TO+punt) its “turnovers per nonscoring possesion.”

So I grouped the teams into three groups: (1) turnovers per non-scoring possession less than 90% of league average, (2) turnovers per non-scoring possession between 90% and 110% of league average, and (3) turnovers per non-scoring possession greater than 110% of league average.

GROUP (1) teams, on average, scored points at 98.9% of league average.
GROUP (2) teams, on average, scored points at 99.6% of league average.
GROUP (3) teams, on average, scored points at 101.7% of league average.

The point is: if you want to be a good offense, it’s just as important — maybe more important — to avoid punts as it is to avoid turnovers.

And maybe part of the reason bad offenses are bad is because they’ve got their turnover tolerance set too low. They’re too willing to punt and not willing enough to risk turnovers. To punt is to spend time at the airport. It won’t give you that punch-in-the-gut feeling that missing a flight will, but it will slowly rob you of some productive and/or enjoyable hours.

The key, of course, is to adjust your turnover tolerance depending on the situation. If you’ve got a multi-million dollar business meeting to get to, or a 4-point lead with a minute left in the game, don’t miss that plane no matter how much time you have to spend eating Cinnabons at the Hudson News. If you’re on vacation and there’s a flight every 45 minutes to your destination, then you can cut it closer.

Certainly teams and players know this. I’m interested in what the optimal turnover rate is on, say, first-and-10 at your own 30 on the opening drive of the game. Do you think teams are generally close to it?

ADDENDUM: Brian Burke had some similar thoughts last week, specifically pertaining to the Redskins.

What would Marvin Harrison’s numbers look like without Peyton Manning? Let’s be clear: I don’t know. You don’t know. I’ll never know. You’ll never know. But what’s the worst thing that could happen if we approach the question as logically as possible and see where it takes us?

Harrison played two seasons before Manning arrived, and they were decent enough seasons for a 1st and 2nd year wide receiver. His third year was Manning’s rookie season, and Harrison played well before suffering an injury that caused him to miss four games. After that, he became Marvin Harrison™.

So how can we say how good Harrison would have been if he just had a regular old QB? For starters, let’s define how good Marvin Harrison has been. Through the 2007 season, he had 1,042 receptions, 13,944 yards and 123 touchdowns. If we award 20 yards per touchdown and five yards per reception, that gives Harrison 21,614 adjusted yards in his career. Whether you love that formula or not, it gives a decent enough impression of how good Harrison has been; only Jerry Rice, Tim Brown and Cris Carter have more adjusted yards (and Harrison moved into second place this year).

The following Colts pass catchers have only played on Manning-led Colts teams: Reggie Wayne, Dallas Clark, Anthony Gonzalez, Joseph Addai and Ben Utecht; other players were so marginal that we don’t care about them (I’m looking at you, Jim Finn, Ben Hartsock et al.). But while Harrison has just about only played with Peyton Manning, some other pass catchers have played with Manning and with another QB. Here’s a full list, with the first column showing games played on the ‘98-’07 Colts and the second listing games played at any other time and place (besides 2008):

Marcus Pollard	       106	 77
Ken Dilger	        63	 93
Jerome Pathon	        46       53
Brandon Stokley	        41	 46
Troy Walters	        52	 40
Edgerrin James	        96	 32
Marshall Faulk	        16	160
Torrance Small	        16	117
Qadry Ismail	        14	110
Terrence Wilkins	41	 13
Lamont Warren	        12	 84
Dominic Rhodes	        71	  9
Tony Simmons	         6	 38
Craig Heyward	         4	145
Jermaine Wiggins	 3	104
Zack Crockett	         2	156
Ricky Proehl	         2	242

For no particularly great reason, I’m going to draw the line between Wilkins and Ismail. Warren and Rhodes were obviously running backs, and Wilkins was really just a returner during his 13-game season with the Rams. Ismail and Small basically played full seasons with the Colts and a bunch of other seasons with other teams.

That gives us nine receivers, running backs and tight ends that played with a Manning-Colts squad and a different squad. Now how the heck do we compare them?

(Continued)

We’ve been keeping an eye on Matt Leinart ever since this blog began. Leinart, along with Young and Jay Cutler, was a first round pick selected in the 2006 draft. There were 70 quarterbacks selected in the first round of the regular NFL draft from 1970 to 2005. With three years in the books for the class of ‘06, I thought this might be a good time to see how those first 70 QBs did after three years. JKL examined Joey Harrington and other young QBs back in September, although he only focused on bad QBs and ignored draft status.

Among the 73 QBs, thirteen made a Pro Bowl (including Young and Cutler) and five of them earned some sort of All Pro honors within their first three seasons. 21 of them were their team’s main starter for all three seasons, fourteen were the main starter for two of the three, another 21 had been the starter for one year and the remaining seventeen were not their teams main starter in any of their first three seasons. Those 70 QBs are listed below, along with the three QBs in the ‘06 draft:

(Continued)

Here are the particulars.

Get those entries in before kickoff of tomorrow’s first game!

EDIT: this contest is now closed. Here is your super-geeky page that tracks the standings.

Here’s a chronological list of every Bowl game. I’ve also listed each team’s Simple Rating System score. The SRS is not a retrodictive system, because it factors in margin of victory and strength of schedule to predict team strength. This particular version of the SRS caps all wins at 40 points and makes all wins of less than seven points equal to seven points. The AvgSRS rating is the average rating of the two teams and the difference column shows the difference in team strength. The favorites, according to the SRS, are in bold:

(Continued)

In case you haven’t noticed, the 2008 Pittsburgh Steelers have one of the greatest defenses in regular season history. Pittsburgh ranks 1st in points allowed, yards allowed, yards per pass allowed, passing yards allowed, yards per rush allowed, and second in rushing yards allowed, sacks, rushing touchdowns allowed and third in passing touchdowns allowed. That’s an incredibly balanced and terrific defense.

Believe it or not, the team that leads the league in points allowed usually isn’t the leader in yards allowed; only four teams in the last 20 years have led the league in both categories — the ‘06 Ravens, the ‘04 Steelers, the ‘02 Bucs and the ‘96 Packers. The ‘85 and ‘86 Bears, the ‘81 Eagles, the ‘79 Bucs, the ‘76 Steelers, the ‘72 Dolphins and the ‘70 Vikings are the only other post-merger teams to pull off this double double.

So what’s the best way to rank the defenses? Let’s run down the major statistics people use to rank the defenses.

(Continued)

Before the season started, we opened up the “38 questions” p-f-r contest. The purpose of this post is to update that with standings, and to make fun of the 21 people who said that the Lions would win more games than the Ravens this year.

(Continued)

EDIT: this contest is now closed. Here is your super-geeky page that tracks the standings.

PRIZE: The usual: honor and glory, and maybe some sponsorship dollars to be named later, if you’re interested.

IS THAT ALL? well, you do get access to the always-updated pool web page that tracks a lot of statistics about the contest, estimates each contestant’s probability of winning, and estimates the final standings after each game finishes. If you like college football and probabilistic geekery based on college football, it’s a lot of fun to follow.

HOW TO ENTER: see below

RULES:

There are 34 bowl games. For each game, one team will be designated the favorite and one team will be designated the underdog. In some games, the underdog will be designated a “longshot.”

Step #1: pick a winner for each game (straight up).

Step #2: group your picks into groups of 1, 2 or 3, subject to the condition that a group of two must have at least one underdog and a group of one must be a longshot.

To get credit for a group, ALL teams in the group must win. Whoever gets the most groups wins. Tiebreaker is whoever picks the most total games correctly (ignoring groups). [NOTE: I think there's a decent chance that the tiebreaker will come into play; that's part of the strategy.]

THE GAMES (listed chronologically):

(Continued)

Pacifist Viking points out (and I’m sure he’s not the only one; he’s just the one I happened to read) that the Cardinals are 5-0 in the division and 3-5 outside the division.

Only two teams since the merger have had a greater discrepancy between their divisional and non-divisional winning percentage. The 1998 Tennessee Oilers were 7-1 in the division and 1-7 outside the division. The 1998 Cowboys were 8-0 in the division and 2-6 in interdivisional games. Somehow, they managed to lose in the playoffs to a divisional opponent, the Arizona Cardinals.

The Cards obviously play in a weak division. It looks at this point like the weakest division in post-merger NFL history. And look who’s second-weakest:

+------------+----------+------+------------------+--------------+
| conference | division | year | Inter_div_record | inter_div_wp |
+------------+----------+------+------------------+--------------+
| NFC        | West     | 2008 | 8-26-0           | 0.235        |
| AFC        | West     | 2008 | 8-24-0           | 0.250        |
| NFC        | Central  | 1984 | 11-28-1          | 0.288        |
| NFC        | West     | 1982 | 7-15-0           | 0.318        |
| NFC        | West     | 1999 | 13-27-0          | 0.325        |
| NFC        | West     | 2004 | 13-27-0          | 0.325        |
| NFC        | West     | 1979 | 13-27-0          | 0.325        |
| AFC        | Central  | 1984 | 13-27-0          | 0.325        |
| NFC        | North    | 2002 | 13-27-0          | 0.325        |
| NFC        | West     | 1976 | 12-25-1          | 0.329        |
| NFC        | North    | 2008 | 11-22-0          | 0.333        |
| NFC        | West     | 1975 | 11-21-0          | 0.344        |
| AFC        | East     | 1989 | 14-26-0          | 0.350        |
| NFC        | Central  | 1979 | 14-26-0          | 0.350        |
| AFC        | West     | 2007 | 14-26-0          | 0.350        |
| AFC        | Central  | 1995 | 14-26-0          | 0.350        |
| NFC        | Central  | 1986 | 14-26-0          | 0.350        |
| NFC        | West     | 2007 | 14-26-0          | 0.350        |
| AFC        | Central  | 1971 | 11-20-1          | 0.359        |
| AFC        | Central  | 1970 | 11-20-1          | 0.359        |
+------------+----------+------+------------------+--------------+

Meanwhile, the NFC South has a chance to match the best interdivisional record since the merger:

+------------+----------+------+------------------+--------------+
| conference | division | year | Inter_div_record | inter_div_wp |
+------------+----------+------+------------------+--------------+
| AFC        | West     | 1984 | 31-9-0           | 0.775        |
| AFC        | Central  | 1975 | 24-8-0           | 0.750        |
| AFC        | South    | 2007 | 30-10-0          | 0.750        |
| NFC        | South    | 2008 | 24-9-0           | 0.727        |
| AFC        | East     | 1999 | 29-11-0          | 0.725        |
| NFC        | East     | 2008 | 24-9-1           | 0.721        |
| NFC        | East     | 2007 | 28-12-0          | 0.700        |
| AFC        | Central  | 1976 | 22-10-0          | 0.688        |
| NFC        | Central  | 1970 | 22-10-0          | 0.688        |
| NFC        | East     | 1982 | 13-6-0           | 0.684        |
| AFC        | West     | 1979 | 27-13-0          | 0.675        |
| NFC        | East     | 1991 | 27-13-0          | 0.675        |
| AFC        | West     | 1995 | 26-14-0          | 0.650        |
| NFC        | West     | 1992 | 26-14-0          | 0.650        |
| NFC        | East     | 1981 | 26-14-0          | 0.650        |
| AFC        | South    | 2008 | 22-12-0          | 0.647        |
| NFC        | South    | 2002 | 25-14-1          | 0.638        |
| NFC        | Central  | 1997 | 25-15-0          | 0.625        |
| AFC        | West     | 1985 | 25-15-0          | 0.625        |
| AFC        | Central  | 1988 | 25-15-0          | 0.625        |
+------------+----------+------+------------------+--------------+

When two NFL teams play each other twice in a season, how often should we expect to see a sweep, and how often should we expect a split?

If every game was a coin flip, we’d have four equally likely possibilities:

Team A wins both
A wins first, B wins second
B wins first, A wins second
B wins both

So 50% of the time it’d be a sweep, and 50% of the time a split.

But games aren’t coin flips. When the 2008 Steelers play the 2008 Bengals, the chances of a sweep would seem to be greater. If we assume the Steelers have an 80% chance of winning each game, then we have:

Steelers win both: .8*.8 = 64% chance
Bengals win both: .2*.2 = 4% chance

So we’ve got a 68% chance of a sweep and a 32% chance of a split. So differences in team strengths have a tendency to cause more sweeps.

But home field advantage should have the reverse effect. Since each team gets one home game, that increases the chances of a split. If the teams are evenly-matched, then assuming the home team wins 60% of the time would yield the following chances of a split:

Home team wins both: .6*.6 = 36% chance.
Road team wins both: .4*.4 = 16% chance.

So it’d be 52% split and only 48% sweep.

So we’ve got two opposing forces at work. And I’ve also assumed independence of the games in the above calculations. Are they really independent, or do we in real life see things like one team having another team’s number? Or do we see the reverse: the losing team in the first game having extra motivation in the second? Do you think that sweeps or splits happen more often in the actual NFL?

It turns out that sweeps are more common. Whether you look at 1970–present, 1978–present, or 2000–present, the percentages are very consistent. About 58% of regular-season two game home-and-home series are sweeps, and 42% are splits.

So now the question is: is this caused by natural variation in team strengths, or is there something else going on?

To find out, I looked at each home-and-home series from 1978 to 2006. I recorded the SRS difference between the two teams and used a logistic regression to translate that into a probability of each team winning, both at home and on the road. Then I used those probabilities to estimate the probability of a sweep. Here’s an example:

In 2000, the Falcons and Saints met twice. According to SRS, the Saints were about 9.4 points better than Atlanta. According to my formula, that translates to roughly a 72% chance of a Saint victory in Atlanta and an 85% chance of a New Orleans win at home. So the sweep probability is .72*.85 + .28*.15, which is about 65%.

Now we look at all 1543 such series and use a calculation like the above to estimate how many sweeps we’d expect to see in total. Answer: if the games are independent, and if my formula for estimating probabilities is OK, we should have expected 911 sweeps in those 1543 series. What we actually saw was 902 sweeps, which is extremely close to the expected value.

So this turns out to be a boring post. Nothing to see here. We see sweeps almost exactly as often as we should expect to see them.

About two and a half years ago, I wrote this:

As you probably know, the participants in the BCS championship game are determined in part by a collection of computer rankings. Those computer rankings are implementing algorithms that “work” because of various mathematical theorems. At some point, I’m going to use this blog to write down everything I know about the topic (which by the way is a drop in the bucket compared to what many other people know; I am not an expert, just a fan) in language that a sufficiently interested and patient non-mathematician can understand.

Since then, I have only written a handful of posts about the mathematics of ranking systems. Here they are:

Simple Ranking System

Another way to derive the simple ranking system

The Maximum Likelihood Method

Some discussion of the technical difficulties involved with the Maximum Likelihood method

Incorporating home-field and/or margin of victory into the Maximum Likelihood Method

I’m going to add another post to this list today by writing about a method that Jeff Sagarin cryptically calls ELO_CHESS. Sagarin’s ELO_CHESS method is one of the six computer ranking systems that figures into the BCS, although as we’ll soon see, we don’t have quite enough information to reproduce his rankings exactly. That’s OK. The point of this post is to understand the theory behind it.

First, a bit of background.

(Continued)

It seems just about everyone has noticed the great seasons that Matt Ryan and Joe Flacco are having.

Best Rookie QB Ever?
Ryan Playing Like No Rookie
Ryan vs. Flacco
Legendary 2008
Learning Curve

With a vast database and a good way to rank the quarterbacks, I think we can answer the question of where Matt Ryan stands in NFL history. Remember, this past summer I ranked every season by every quarterback in NFL history. Using that same methodology, we can compare Ryan (and Flacco) to all rookies that have come before them.

Here’s a quick refresher. I start with a quarterback’s adjusted net yards per attempt ratio. That formula is calculated by looking at a QB’s number of passing yards, adding 10 points for all touchdown passes, subtracting 45 yards for all interceptions, subtracting one yard for every sack yard lost, and then dividing that number by the QB’s total number of sacks plus pass attempts. Regular PFR readers know that I recently decided to up the TD bonus to 20 yards per score, but updating the old QB list with that new metric is a job best saved for the off-season. We’ll stick with 10 yards for all TD passes for now.

After figuring out the ANY/A metric for each QB, you have to compare that number to the ANY/A number for all other quarterbacks in the NFL. Then to figure out “QB value added”, you take the difference between the QB’s ratio and the league average ratio, and multiply it by the number of pass attempts (including sacks) the QB had. Quarterbacks also additional bonus yards for all rushing yards over 4.0 yards per carry. If that sounds a bit complicated, let’s use Dan Marino’s 1983 season as an example.

Marino had 2,210 passing yards, 20 TDs (200 adjusted yards), 6 INTs (-270 adjusted yards) and only 80 sack yards lost (-80 adjusted yards) on 296 passes and 10 sacks (306 total attempts). That means Marino averaged 6.73 adjusted net yards per pass, which is very good now and was even better in 1983. Outside of Marino, NFL QBs in 1983 averaged 4.53 ANY/A, meaning Marino averaged a full 2.20 more adjusted yards per passing play. Multiplied by his 306 passing plays, and Marino gets credit for being 674 yards above average in 1983.

How does that rank all time? Here’s a list of all rookie QBs that added over 250 adjusted yards over average as rookies (note: for years where the NFL season was fewer than 16 games, every QB’s value was prorated as if they played a season length between what they actually did and 16 games; so Charlie Conerly’s season was pro-rated to a 14 game season). Because of the somewhat nebulous nature of how to define a rookie, I’m going to be overinclusive by counting the first season in which a player recorded a stat as his rookie year; however, for guys whose rookie seasons were not their draft year or if the player spent time in another professional football league first, I put two asterisks by their name.

Score				        year    att	pyd	ptd	pint	any/a
674	Dan Marino	        mia	1983	296	2210	20	 6	 6.73
581	Greg Cook	        cin	1969	197	1854	15	11	 5.81
513	Charlie Conerly**	nyg	1948	299	2175	22	13	 6.05
476	Marc Bulger**	        ram	2002	214	1826	14	 6	 7.05
392	Sid Luckman	        chi	1939	 51	 636	 5	 4	 9.92
392	Johnny Unitas**	        clt	1956	198	1498	 9	10	 5.75
392	Ben Roethlisberger	pit	2004	295	2621	17	11	 6.41
384	Norm Van Brocklin	ram	1949	 58	 601	 6	 2	 9.84
367	Jim Kelly**	        buf	1986	480	3593	22	17	 5.20
343	Billy Wade**	        ram	1954	 59	 509	 2	 1	 8.20
323	Pat Haden**	        ram	1976	105	 896	 8	 4	 6.02
320	Johnny Lujack**	        chi	1948	 66	 611	 6	 3	 8.12
319	Joe Namath	        nyj	1965	340	2220	18	15	 5.07
312	Butch Songin**	        nwe	1960	392	2476	22	15	 5.16
298	Jacky Lee	        oti	1960	 77	 842	 5	 6	 8.08
296	Ed Rubbert	        was	1987	 49	 532	 4	 1	10.36
287	Bob Celeri**	        nyy	1951	238	1797	12	15	 5.22
286	Tom Flores**	        rai	1960	252	1738	12	12	 5.23
262	Dick Jamieson**	        nyj	1960	 70	 586	 6	 2	 7.94
256	Aaron Brooks**	        nor	2000	194	1514	 9	 6	 5.93
254	Jeff Garcia**	        sfo	1999	375	2544	11	11	 5.27

Obviously it’s not easy to compare Sid Luckman’s 1939 season to Ed Rubbert’s 1987 season to Jim Kelly’s 1986 year. The above list doesn’t give enough credit to the guys who played the most (a good indicator that the rookie was playing well), so we should drop the baseline to three-quarters of league average, instead of league average. Here’s what I think is the best list for measuring the greatest rookie QB seasons of all time:

Score				        year    att	pyd	ptd	pint	any/a
1020	Dan Marino	        mia	1983	296	2210	20	 6	6.7
 964	Jim Kelly**	        buf	1986	480	3593	22	17	5.2
 916	Charlie Conerly**	nyg	1948	299	2175	22	13	6.1
 814	Ben Roethlisberger	pit	2004	295	2621	17	11	6.4
 801	Greg Cook	        cin	1969	197	1854	15	11	5.8
 777	Butch Songin**	        nwe	1960	392	2476	22	15	5.2
 756	Marc Bulger**	        ram	2002	214	1826	14	 6	7.1
 743	Warren Moon**	        oti	1984	450	3338	12	14	4.9
 722	Jeff Garcia**	        sfo	1999	375	2544	11	11	5.3
 699	Joe Namath	        nyj	1965	340	2220	18	15	5.1
 634	Johnny Unitas**	        clt	1956	198	1498	 9	10	5.7
 620	Charlie Batch	        det	1998	303	2178	11	 6	5.3
 595	Byron Leftwich	        jax	2003	418	2819	14	16	4.9
 586	Tom Flores**	        rai	1960	252	1738	12	12	5.2
 570	Bob Celeri**	        nyy	1951	238	1797	12	15	5.2
 509	Aaron Brooks**	        nor	2000	194	1514	 9	 6	5.9
 482	Dieter Brock**	        ram	1985	365	2658	16	13	4.5
 472	Fran Tarkenton	        min	1961	280	1997	18	17	5.0
 460	Matt Leinart	        crd	2006	377	2547	11	12	4.9
 457	Paul Governali**	byk	1946	192	1293	13	10	5.1
 452	Vince Young	        oti	2006	357	2199	12	13	4.2
 451	Norm Van Brocklin	ram	1949	 58	 601	 6	 2	9.8
 444	Sid Luckman	        chi	1939	 51	 636	 5	 4	9.9
 441	Mark Rypien**	        was	1988	208	1730	18	13	5.5
 438	M.C. Reynolds	        crd	1958	195	1422	11	11	5.3

That list is a pretty solid ordering of the best rookie seasons ever. Conventional wisdom puts Marino and Roethlisberger at the top, along with Greg Cook if the author has heard of him. Depending on whether you want to count Kelly as a rookie in 1986 — he previously had played two excellent seasons in the USFL — he is at the top of the list, as well.

So where do Ryan and Flacco rank? Through 13 games, Ryan has 2940 passing yards, 14 TD (140 adjusted yards), 7 INT (-315 adjusted yards), and 80 sack yards lost (-80 adjusted yards) on 366 pass attempts and 13 sacks. That’s an average of 7.08 adjusted net yards per pass attempt.

Flacco has 2410 passing yards, 13 TD (130), 10 INT (-450), and 208 sack yards lost (-208) on 352 pass attempts and 23 sacks. That’s an average of 5.02 adjusted net yards per pass attempt.

The NFL average for all passers is 5.32 ANY/A through 13 weeks. Last year the average was 5.11 but we can expect the league average to go down as some more December games are played. To calculate a QB’s “value added”, you simply compare each QB to the other QBs in the league. Ryan’s averaged 7.08 ANY/A while all other QBs have averaged 5.27 ANY/A. So Ryan’s been averaging 1.81 more adjusted net yards per pass attempt than other QBs in the NFL, over the course of 379 passing plays. That gives him a value added of 686 yards over average. Flacco’s averaged 5.02 ANY/A while all other NFL QBs have averaged 5.33 ANY/A — that obviously puts Flacco at slightly below the league average. While Flacco’s posted impressive numbers this season, he’s hurt by his high number of sacks and sack yards lost.

If we use the 3/4 of league average mark, Ryan jumps to 1185 yards over replacement and Flacco at 386 yards over replacement (383 passing yards over replacement plus a three yard rushing bonus). Once again, that seems like a more proper way to measure Flacco’s success, as he really has been one of the best rookies we’ve seen in awhile. Unfortunately, he’s playing at the same time as maybe the best rookie we’ve ever seen. As long as Ryan doesn’t implode the rest of the season, he will finish the 2008 season as the greatest rookie QB of all time.

Here are some other historical measures of rookie QB success.

• The NFL began playing Pro Bowls after the 1938 season. Only Vince Young (2006), Dan Marino (1983), Bob Griese (1967), Joe Namath (1965) and Davey O’Brien (1939) made the Pro Bowl as rookies. Ryan’s got some tough competition in the NFC this year — Kurt Warner and Drew Brees both may throw for 5,000 yards while Eli Manning might be leading a 14-2 Giants team. But he’s got a chance to become just the third rookie QB since the merger to make the Pro Bowl.

• Since 1950, only one rookie QB has been named to an All Pro team — Dan Marino. Prior to that year, the only QBs to make an All Pro team based on their accomplishments as quarterbacks were Bob Waterfield (1945), Davey O’Brien (1939) and Sammy Baugh (1937).

• Only three QBs have even been named AP Rookie of the Year, which began being awarded in 1967. Vince Young, Ben Roethlisberger and Dennis Shaw. Only one player has won an NFL MVP award as a rookie — Bob Waterfield on the 1945 Rams.

Of course, no post by me on QBs would be complete without a look at the least valuable rookie seasons of all time. Here’s the list, using the league average as the baseline:

-1243	Bud Schwenk	        crd	1942	295	1360	 6	27	 0.69
- 942	David Carr	        htx	2002	444	2592	 9	15	 3.07
- 838	Stan Heath	        gnb	1949	106	 355	 1	14	-2.50
- 831	Ryan Leaf	        sdg	1998	245	1289	 2	15	 1.85
- 815	Kyle Orton	        chi	2005	368	1869	 9	13	 2.97
- 813	Dan Darragh	        buf	1968	215	 917	 3	14	 1.47
- 785	Jack Trudeau	        clt	1986	417	2225	 8	18	 2.87
- 784	Paul Christman**        crd	1945	219	1147	 5	12	 3.00
- 781	Jeff Komlo	        det	1979	368	2238	11	23	 2.33
- 769	Alex Smith	        sfo	2005	165	 875	 1	11	 1.06
- 764	Andrew Walter**        	rai	2006	276	1677	 3	13	 2.69
- 753	Lamar McHan	        crd	1954	255	1475	 6	22	 2.14
- 723	Randy Hedberg        	tam	1977	 90	 244	 0	10	-3.21
- 719	John McCarthy	        crd	1944	 67	 250	 0	13	-5.00
- 711	Terry Bradshaw	        pit	1970	218	1410	 6	24	 0.61
- 706	Chris Weinke	        car	2001	540	2931	11	19	 3.55
- 701	Tobin Rote	        gnb	1950	224	1231	 7	24	 0.99
- 681	Steve DeBerg**	        sfo	1978	302	1570	 8	22	 1.58
- 620	Richard Todd	        nyj	1976	162	 870	 3	12	 0.66
- 586	Norm Snead	        was	1961	375	2337	11	22	 3.89
- 583	Bruce Gradkowski	tam	2006	328	1661	 9	 9	 3.40
- 567	Joey Harrington	        det	2002	429	2294	12	16	 3.70
- 562	Steve Fuller	        kan	1979	270	1484	 6	14	 2.18
- 555	Tom Sherman	        nwe	1968	226	1199	12	16	 2.65
- 554	Bert Jones	        clt	1973	108	 539	 4	12	-0.71
- 549	Donovan McNabb	        phi	1999	216	 948	 8	 7	 2.09
- 547	Terry Hanratty	        pit	1969	126	 716	 8	13	 0.29
- 544	Gary Huff	        chi	1973	126	 525	 3	 8	 0.10
- 535	Kent Nix	        pit	1967	268	1587	 8	19	 3.03
- 528	Randy Johnson	        atl	1966	295	1795	12	21	 3.29
- 516	Rick Mirer	        sea	1993	486	2833	12	17	 3.66
- 508	Troy Aikman	        dal	1989	293	1749	 9	18	 2.80

Despite having two Big 12 fans writing on this blog (neither of which are me), we’ve been pretty quiet here about Texas, Oklahoma and the rest of college football. At this point, there is very little left to be said that hasn’t been repeated a hundred times. While I don’t speak for Doug or JKL, here’s my very short summary of OU and UT:

• The Texas win over Oklahoma is an important data point when comparing the two teams, but it is not the only data point that matters.

• While we have asked whether H2H is the right tiebreaker, I’ll simply stipulate here that it should be the tiebreaker used whenever possible. If only UT and OU were tied in the Big 12 South, Texas should certainly advance.

• The BCS rankings should not be bound by such rules, of course. It’s perfectly legitimate to rank teams with even records based on other things than head to head. For example, it is reasonable to rank Iowa ahead of Northwestern. Don’t blame the BCS — it is the Big 12 that decided to use the BCS rankings as its tiebreaker.

• That said, I think the Big 12 made the right decision. While many are upset now, what would you think if OU was 2nd, Texas was 8th, and Texas Tech was 10th? Surely the Big 12 would want Oklahoma to advance. It is in the conference’s best interest to advance the team with the highest chance of making the championship game. And while the Pac-10 avoided having Oregon State win the conference, about a week ago I am sure the Rose Bowl was wishing the Pac-10 used BCS rankings to break all tiebreakers, as opposed to head to head (which would have put the Beavers ahead of the Trojans.)

• All that said, OU had a slightly more difficult schedule than Texas and outscored its opponents by slightly more points. As an unbiased observer, I’m perfectly fine with Oklahoma winning the Big 12 South.

I said earlier that almost everything that could be said has been said so far. But there’s a neat little stat I like to use every once in awhile, and it goes like this: you rate each team based on how it did against every opponent, compared to how every other team did against that opponent. Let’s use Penn State as an example. I award three points to the home team, and then use the points differential to rank the games.

PennState	Iowa	        23	24	R	 2	 1
PennState	OhioState	13	 6	R	10	 2
PennState	Illinois	38	24	H	11	 3
PennState	Purdue	        20	 6	R	17	 4
PennState	Indiana	        34	 7	H	24	 5
PennState	Michigan	46	17	H	26	 6
PennState	MichiganState	49	18	H	28	 7
PennState	OregonState	45	14	H	28	 8
PennState	Temple	        45	 3	H	39	 9
PennState	Wisconsin	48	 7	R	44	10
PennState	Syracuse	55	13	R	45	11

PSU's worse game of the year was against Iowa, of course. While Penn State lost by one point, since it was a road game, that counts as a two point "win". Wins and losses are largely irrelevant in this system, though, so don't worry about whether something is classified as a one point win or a two point loss -- it's essentially treated the same way here. So Iowa gets 1 point for playing the toughest game against the Nittany Lions; Ohio State gets 2 points, and all the way down until Syracuse gets 11 points for being the easiest opponent PSU faced.

All 1-AA (Football Championship Subdivision) teams were ignored in this system. Obviously, the fewer points you score the better. For example, check out where Florida ranks against its eleven opponents (the Citadel excluded):

FloridaState	Florida	15	45	H	-33	1
Georgia	        Florida	10	49	N	-39	1
Hawaii	        Florida	10	56	R	-43	1
Kentucky	Florida	 5	63	R	-55	1
LouisianaState	Florida	21	51	R	-27	1
Miami(Florida)	Florida	 3	26	R	-20	1
SouthCarolina	Florida	 6	56	R	-47	1
Tennessee	Florida	 6	30	H	-27	1
Vanderbilt	Florida	14	42	H	-31	1
Arkansas	Florida	 7	38	H	-34	3
Mississippi	Florida	31	30	R	  4	5

Florida put the biggest whooping a team saw all year on nine of its eleven opponents. That’s impressive. Arkansas was only beat worse by Texas and Alabama, while Ole Miss only fared worse against South Carolina, Vanderbilt, Alabama and Wake Forest.

Florida’s average score in this system was 1.5 — they scored 17 ‘points’ in 11 games. That 1.5 rating is the lowest (best) in college football. Not surprisingly, they are followed by USC, and then UT and OU.

These rankings are obviously not the best way to rank college football teams. This is just a fun exercise. As you’ll soon see, teams that play easy schedules have an advantage here — Boise State gave New Mexico State its worst loss of the season, but BSU was competing with the likes of Utah State, Nebraska and San Jose State for that prize. It’s not too impressive to give your opponent its worst loss when those are its other opponents. But Texas gave OU its worse loss of the season, when OU also played Texas Tech, Oklahoma State, Cincinnati and TCU. That’s a lot more impressive.

Still, it’s a quick and neat way to look at things. And thankfully, it’s not the same thing we have heard over and over again for the last three weeks. Here are the full rankings:

1.5	Florida
1.6	SouthernCalifornia
1.9	Texas
1.9	Oklahoma
2.0	BoiseState
2.2	TexasChristian
2.3	PennState
2.7	BallState
2.8	OhioState
3.1	Alabama
3.3	Missouri
3.4	TexasTech
3.5	Utah
3.5	OklahomaState

This is not how you should rank all teams in Division 1 (Football Bowl Subdivision) college football. This is a simple system with some very obvious flaws. This list should not be used as a replacement for other rankings. This system is definitely biased towards teams whose opponents had easy schedules, and this system is slightly biased towards teams with great offenses relative to teams with great defenses.

3.8	Mississippi
4.0	Oregon
4.1	Cincinnati
4.1	GeorgiaTech
4.2	BostonCollege
4.2	Iowa
4.3	California
4.4	Georgia
4.4	FloridaState
4.5	Arizona
4.5	Pittsburgh
4.6	OregonState
4.6	NorthCarolina
4.8	Nebraska
4.8	Tulsa
4.8	Nevada
4.9	BrighamYoung
4.9	Kansas
5.1	Wisconsin
5.1	VirginiaTech
5.2	MichiganState
5.3	Navy
5.3	Rutgers
5.3	WestVirginia
5.3	SouthernMississippi
5.4	AirForce
5.4	SouthFlorida
5.4	Clemson
5.5	LouisianaState
5.5	Illinois
5.6	Rice
5.6	Connecticut
5.6	SouthCarolina
5.6	Miami(Florida)
5.7	Northwestern
5.8	Tennessee
5.8	BowlingGreenState
5.9	Troy
5.9	Houston
5.9	WakeForest
6.0	Minnesota
6.0	Stanford
6.1	EastCarolina
6.1	Vanderbilt
6.1	NorthernIllinois
6.1	WesternMichigan
6.2	Maryland
6.3	Baylor
6.3	Buffalo
6.3	Temple
6.4	CentralMichigan
6.5	NotreDame
6.5	NorthCarolinaState
6.5	Virginia
6.6	ArizonaState
6.6	Auburn
6.6	Purdue
6.7	NewMexico
6.7	LouisianaTech
6.7	Hawaii
6.8	Kentucky
6.9	Akron
7.0	Louisville
7.0	ColoradoState
7.0	Memphis
7.1	Colorado
7.2	FresnoState
7.2	Arkansas
7.3	Louisiana-Lafayette
7.3	FloridaInternational
7.3	Marshall
7.3	Texas-ElPaso
7.4	Kent
7.4	Duke
7.4	ArkansasState
7.5	MiddleTennesseeState
7.5	KansasState
7.5	Toledo
7.6	FloridaAtlantic
7.6	Ohio
7.6	CentralFlorida
7.7	Army
7.8	Michigan
7.8	SanJoseState
7.8	Nevada-LasVegas
7.8	TexasA&M
8.3	MississippiState
8.3	Indiana
8.3	UtahState
8.4	UCLA
8.4	Syracuse
8.5	Alabama-Birmingham
8.5	Louisiana-Monroe
8.6	EasternMichigan
8.8	IowaState
9.2	NewMexicoState
9.3	SouthernMethodist
9.4	Tulane
9.5	SanDiegoState
9.5	Miami(Ohio)
9.6	WesternKentucky
9.7	Washington
9.7	Wyoming
10.3	NorthTexas
10.7	WashingtonState
10.7	Idaho

In case you’re curious about Alabama’s low ranking, here’s the full schedule report:

ArkansasState	Alabama	 0	35	R	-32	1
Auburn	        Alabama	 0	36	R	-33	1
Clemson	        Alabama	10	34	N	-24	1
Arkansas  	Alabama	14	49	H	-38	2
Georgia	        Alabama	30	41	H	-14	2
Tennessee	Alabama	 9	29	H	-23	2
WesternKentucky	Alabama	 7	41	R	-31	2
Mississippi	Alabama	20	24	R	- 1	3
LouisianaState	Alabama	21	27	H	- 9	4
MississippiStateAlabama	 7	32	R	-22	4
Kentucky	Alabama	14	17	R	  0	7
Tulane	        Alabama	 6	20	R	-11	8

Here’s Texas’s report:

FloridaAtlantic	Texas	10	52	R	-39	1
Missouri	Texas	31	56	R	-22	1
Oklahoma	Texas	35	45	N	-10	1
Rice	        Texas	10	52	R	-39	1
Arkansas	Texas	10	52	R	-39	1
Colorado	Texas	14	38	H	-27	2
Kansas	        Texas	 7	35	H	-31	2
TexasA&M	Texas	 9	49	R	-37	2
Texas-ElPaso	Texas	13	42	H	-32	2
OklahomaState	Texas	24	28	R	- 1	3
TexasTech	Texas	39	33	H	  3	3
Baylor	        Texas	21	45	R	-21	4

Texas Tech played a tougher game against OU and an equally tough game against Nebraska (I randomly sorted the teams for tiebreakers, so Texas should really get a 2.5 instead of a 3 for that game. For Texas’ actual score, I gave them a 2.5 (and that’s why I put them above Oklahoma) but for all other teams I used a random tiebreaker). Baylor was beat worse by Oklahoma, Wake Forest and Oklahoma State. Oklahoma State was beat worse by OU and Texas Tech. UTEP was beat worse by Tulsa; A&M by Oklahoma; Colorado by Missouri; and Kansas by Texas Tech.

Here’s Oklahoma:

Baylor	        Oklahoma	17	49	H	-35	1
Cincinnati	Oklahoma	26	52	R	-23	1
TexasA&M	Oklahoma	28	66	H	-41	1
TexasChristian	Oklahoma	10	35	R	-22	1
TexasTech	Oklahoma	21	65	R	-41	1
Nebraska	Oklahoma	28	62	R	-31	2
OklahomaState	Oklahoma	41	61	H	-23	2
Washington	Oklahoma	14	55	H	-44	2
Kansas	        Oklahoma	31	45	R	-11	3
Texas	        Oklahoma	45	35	N	 10	3
KansasState	Oklahoma	35	58	H	-26	4

Kansas State was rocked harder by Texas Tech, Nebraska and Kansas. UT fared worse against Texas Tech and Oklahoma State. Washington was beat by more by the Trojans; Oklahoma State by Texas Tech; and Nebraska by Missouri.

By far the best college football blogger there is — maybe the best anything blogger there is — is Dr. Saturday over at Yahoo. He used to be known as Sunday Morning QB, and I’ve referenced him a few times here at this blog.

In an effort to make sense of the resumes of the three teams involved in the Big XII South logjam (yes, there are in fact three teams in the tie!), he cooked up a stat last weekend called Time of KO, which is the time on the clock at the instant the winning team scored the points that would surpass the losing team’s total for the entire game.

In Thursday’s Titans-Lions game, for instance, Detroit scored 10 points in the game. Tennessee scored its second TD at the 8:30 mark of the first quarter, so that’s the time of KO. The Colts 10-6 win over the Browns yesterday would be classified as a fourth quarter KO. The Colts scored a touchdown to take a 10-6 lead at the 9:55 mark of the fourth, to be precise.

The knockout analogy is a bit strained. You can only recognize it in hindsight here, whereas in boxing it’s pretty obvious at the time. Still, I think it’s a fairly nifty way to put wins into broad general categories of impressiveness. In some situations, it misses the mark. It is possible, for instance, to register a first quarter KO in a game that was tight right down to the very end. It can also be manipulated to some extent by garbage TDs at the end of the game. If you’re looking for the ultimate power index, this ain’t it. But it’s easy to understand, it’s more informative (I think) than simple point margin, and it’s fun.

Here are the number of first quarter, second quarter, third quarter, and fourth quarter KOs for every team in post-merger NFL history with 14 or more wins. As is typical, I’ll let you draw your own conclusions and share them in the comments:

(Continued)

It’s been awhile since I utilized the Insane Ideas tag. This is one of those things that has zero chance of being even listened to, much less accepted, by any audience of football fans outside this board. I’m not sure it has much of a chance even here, but I’ll give it a shot.

Ever since I can remember, the use of head-to-head as the first tiebreaker seemed a little funny to me. Not wrong, necessarily, just funny. And that’s what I’m going to discuss here in this post.

Just to avoid unnecessary extraneous issues, let’s fix the conversation on the cleanest possible case. Namely, we’ll assume the conference in question plays a full round robin and no other games (that count in the standings). PAC 10 football is an example of this scheme, and a timely one, as they might be heading toward a potential USC / Oregon State tie at the top this season.

At least on a theoretical level, any tiebreaking rule involving only wins and losses — including head-to-head — has a flip-side that argues against itself. After all, these teams have the same record. That necessarily means that if one of them had a better record in a certain subset of games — the one where the two played each other, for instance — then the other one must necessarily have had a better record in the other subset. If Oregon State was 1-0 and USC 0-1 in games between Oregon State and USC, then that means that USC must have had a better record than Oregon State, against the exact same teams, in the other games. Why favor one set of games over the other?

I think the answer is that, to most people, it just seems morally right. The fact that USC beat Stanford and Oregon State didn’t isn’t as important as the fact that Oregon State beat USC. In other words, most people think that Oregon State is more deserving of the title than USC is. And I really can’t argue with that.

But what I’m interested in is a somewhat more objective question: does the head-to-head tiebreaker do a good job of crowning the better team? Most people would agree that, if a tie should come to pass in the PAC 10 this year, the better team will not be the one wearing the crown.

I think most people are fine with that, and I’m not saying I’m not. I’m just wondering: is that typical? When two teams end up tied at the end of the year, which team is more likely to be the better team: the team that won the one-game subset of games between the two teams? Or the team that did better in the eight-game slate against the rest of the conference? The team that beat USC but lost to Stanford? Or the team that lost to Oregon State but beat Stanford?

Using real-life data is probably out of the question, as this kind of situation doesn’t happen often enough to get a good sample, so I ran a simulation. I simulated a gajillion PAC 10 seasons using rules similar to (but not exactly the same as) those established in the Ten Thousand Seasons post from way back. Remember, in a simulation, we can be omniscient. In each of these gajillion fake seasons, unlike with real football seasons, we know who the better team is. The better team doesn’t always win, of course (if it did, we wouldn’t need tiebreakers at all), but at the end of the season, we can say definitively whether or not the best team won.

So every time there was a two-way tie for first, I gave the crown to the team with the head-to-head win, but I noted whether or not that team was the better team.

What do you think happened?

A. The conference champ (i.e. the head-to-head winner) was the better team more often than not

B. The conference champ was the worse team more often than not

C. The results were not statistically distinguishable from a 50/50 split.

ANSWER EDIT…

Before running the study, my thoughts were exactly the same as those spelled out by commenter “Dead Cat Bounce”:

I’m guessing (B). The other subset is bigger, 8 games, than the head-to-head subset of 1 game. Statistically, it would be tougher for the “worse” team to maintain their advantage over a longer series than just springing an upset in a single game.

[NOTE: he actually wrote (A), but I am pretty sure, based on the rest of his comment, that he meant (B). If not, I apologize.]

While the one game is direct evidence and the other eight are indirect, the other eight are more evidence and the one is less. A lot less. That’s why I thought, like DCB, that (B) would be the answer.

But it wasn’t. The actual answer is some combination of (A) and (C). In particular, the technically correct answer is (A): the head-to-head winner was in fact the better team more often than not and the margin was statistically significant. But (C) might be the more practically correct answer. Though statistically significant because of huge sample size, the margin was about 51/49 or 52/48 or something that, for practical purposes, is essentially 50/50.

So where did my and DCB’s reasoning go wrong?

Here’s where: if all you know is that one team did better than another team in an N-game season, then it definitely matters how big N is. But if you know, as we do in this case, that one team was EXACTLY ONE GAME better than another team in an N-game season, I don’t think it matters much how big N is. In this example, while USC did better than Oregon State in the bigger 8-game part of the schedule, we know by virtue of the fact that the teams are tied overall, that they really didn’t do better in the 8-game part of the schedule. They only did better in one of those games, the Stanford game in this case. So it’s not an 8-vs-1 situation. It’s really a 1-vs-1 situation with the other seven necessarily canceling out.

Much has been made of the fact that the Steelers-Chargers game last Sunday was the first 11-10 game ever. Let’s call this a singular game as there has been but one. Just how unique is a singular game? It turns out, not very.
(Continued)

Yesterday I looked at what my algorithm thought were the ten top NCAA programs in terms of producing NFL talent. Today, we’ll look at some other teams of interest.

Just missing the top 10 is my alma mater, a school which has had moderate success at various times, but isn’t thought of as a football power:

Arizona St.

QB   Jake Plummer          76
RB   John Henry Johnson    57
RB   Gerald Riggs          54
RC   Charley Taylor        82
RC   Jerry Smith           71
RC   John Jefferson        54
RC   Larry Walton          42
T    Marvel Smith          45
T    Levi Jones            34
G    Randall McDaniel     103
G    David Dixon           51
C    Grey Ruegamer         10

DT   Curley Culp           92
DT   Dan Saleaumua         62
DE   Jim Jeffcoat          59
DE   Al Harris             47
ILB  Bob Breunig           65
ILB  Derek Smith           61
OLB  Larry Gordon          51
OLB  Derrick Rodgers       43
CB   Mike Haynes          108
CB   Eric Allen            98
S    Darren Woodson        78
S    David Fulcher         52

Whether you want to call him an end or a linebacker, Terrell Suggs will join the team within a year or two.

This team would likely have a quarterback controversy involving Plummer and Danny White, which is an interesting QB pair. I wasn’t quite old enough to understand exactly what public opinion on White was (or maybe I’m too old to remember), but it seems to me like he was something of a Jake Plummer of his day. He had a fair amount of success, but it was always intermixed with disappointing inconsistency and a perceived failure to fulfill his promise.

Moving on…

(Continued)

First a plug: if you’re an NBA fan, you should check out what Neil Paine (with assists from Justin Kubatko) has been doing with the basketball-reference.com blog. It’s good hoops-related reading in the same spirit as this blog.

Neil recently did a couple of posts on which colleges have produced the most NBA talent. Inspired by that, and my NFL all-franchise team posts (AFC - NFC) from this summer, I’ve decided to create a “team” from the talent produced by each college, and rank them.

The rules are essentially the same as those for the NFL all-franchise team posts, but with a few college-related extras added:

1. If you haven’t read about my Approximate Value (AV) method for rating players, you should read about it right here.

2. A player is only eligible to play for his final college, if he attended more than one. For example, Troy Aikman can be on the UCLA team, but not the Oklahoma team.

3. Keep in mind that these lists are ordered by NFL production. Archie Griffin was one of the best college football players of all time, but he loses his spot on the Ohio State team to the merely-very-good-in-college Robert Smith, who had a much more illustrious NFL career.

4. The AV systems gives a player a score for each player season. To combine these into a career number, I take 100% of the player’s best season, plus 95% of his second-best season, plus 90% of his third-best season, and so on.

5. I’m only comfortable (for now) applying the AV methodology to seasons 1950 and later. Players who debuted before 1950, however, are included if their post-1950 seasons alone merit inclusion. In this case, they have a ‘+’ after their AV score to remind you that their career AV is (probably) higher than the number shown.

6. To avoid 4-3/3-4/5-2 issues, I gave each defense 12 players, including two DT/NTs, two DEs, two OLBs, and two ILB/MLBs. I have also now lumped all safeties together instead of distinguishing between free and