![[The Sports Report Letterhead]](tsr_letr.gif)
The Sports Report Ranking System(s)
Rankings and Statistics in General
That depends on what you're looking for. Not all ranking systems are the same. There is always a bias toward something.
Herman Matthews, who publishes his rankings in the Scripps Howard newspapers, is probably the best in the business. His rankings are good at just about everything. They rank well top-to-bottom, and they're highly predictive. The fact that other systems are just now beginning to match his accuracy, even though he has been using the same methodology since the mid-1960s, says something.
For predictive ability only, it's hard to go wrong with Jeff Sagarin, who publishes his rankings in USA Today. His rankings are slightly more predictive than Matthews.
The best for ranking teams from highly-disparate leagues - especially for ranking high school teams from different classes - is the late David Rothman's FACT system. This ranking also has the advantage of being publicly-available.
Unfortunately, both Matthews and Rothman opted to drop out of the BCS, leaving Sagarin as the unquestioned best still in the BCS.
Kenneth Massey's system is fairly generic and should appeal to statisticians (although it's odd that there are several systems similar to his, all written by Virginia Tech people over the past 20-odd years).
Richard Billingley's system has been included in the BCS largely because he was already compiling historical statistics for the NCAA. As a ranking system, it's not remarkable in any way.
The other rankings are pretty ordinary.
It doesn't. Someone else can make that call.
Yes. The objection is on three fronts.
First, all systems are biased. Even Herman Matthews's and Jeff Sagarin's. And also TSRRS. So the thought of ranking systems being "objective" simply isn't valid. Each system values what its creator told it to value.
Second, when you look at how the BCS awards points, you'll see that the opinion of Kenneth Massey (via his ranking system) is worth roughly 10 times the opinion of Bobby Bowden. Kenneth is a good guy, and his system is pretty good at what it's supposed to do, but, frankly, Bobby Bowden's opinion should always be worth more than Kenneth Massey's.
Third, all of the computer ranking systems are proprietary. Where do the numbers come from? Are they independently verifiable? Several years ago, it was discovered that the BCS rankings were not actually calculated the way the BCS intended them to be, due to Cal-Davis being accidentally included as a Division I-AA team. This was discovered via independent verification. Without independent verification, there's no sure way to know whether the presented numbers are valid.
Yes, but not for the reasons you think.
Schedule strength means different things to different teams. Not all "average" schedules are average for the teams playing them.
Let's go through an example - you may find this interesting. There are some numbers in it, but they're not too bad. Be warned, though: the first four paragraphs are background. Nevertheless, if you make it through those, the fifth paragraph will be worth it.
Take three teams: one that would win 90% of its games against an average opponent (e.g., Florida), one that would win half - 50% - of its games against an average opponent (e.g., Kansas), and one that would win 10% of its games against an average opponent (Buffalo). So far, so good.
What happens if an average team - one just like Kansas - plays this schedule? It loses 90% of the time to Florida (remember - we said Florida would win 90% of its games against an average opponent, and Kansas is an average team in this example). It has a 50-50 chance against Kansas (which is "just like" it). And it wins 90% of the time against Buffalo (which wins only 10% of its games against an average opponent, and Kansas is an average opponent in this example). Therefore, over time Kansas would win 50% of its games against this schedule (90% plus 50% plus 10%, all divided by 3, is 50%).
Notice that if Kansas plays only teams just like itself, it also wins 50% of its games. Okay, that part was obvious, wasn't it?
Now let's have a team just like Florida play an entire schedule of "Kansas" teams. Again, because we defined "team just like Florida" this way, it would win 90% of its games against the Kansases of this world. That's pretty straightforward, right?
Now here's the interesting part. What if that "team just like Florida" plays that schedule we started with? Well, when it plays Florida, it splits that game because it's the same strength team. When it plays Kansas, it wins 90%. And when it plays Buffalo, it wins just about all of them. Yet even if it wins every Buffalo game, the team still wins only 80% of all of these games over time (50% plus 90% plus 100%, all divided by 3, is 80%).
The net result is that the diverse schedule we started with, while averaging out to .500, is much tougher for Florida to play than the schedule of Kansases (which also averages out to .500). Yet Kansas shouldn't care which schedule it plays. And on the bottom end, in case you're wondering, Buffalo's win-loss record is better off playing the diverse schedule we started with.
Schedules are in the eyes of the beholder.
Maybe. Or maybe not.
The "three T's" are turf, time, and temperature.
The theory is that "home-field advantage" (HFA) is, for the most part, a statistical illusion caused by having incomplete data. But when you take the three T's into account, most of the HFA goes away.
Take the easy one: turf. All surfaces are not the same, and the difference between grass and artificial turf is severe. It starts with the shoes, which are cleats on grass but tennis shoes on artificial turf. Option teams are particularly susceptible to turf differences. So if you play Nebraska, be sure to play the Huskers on grass. Watch the videotape of the 2000 Nebraska-Oklahoma game. Watch Bobby Newcombe fall several times when cutting on the grass. On artificial turf, he might score a couple of touchdowns on the same plays. Except he wouldn't have the same plays, because Oklahoma plays Nebraska differently on artificial turf.
The second "T" is time. In particular, there is a difference between day games and night games. There are several causes. Part is vision - some people see better in the day, others see (relatively) better at night. Part is internal time clocks - the same thing that makes some of us "morning people" and others "night people." There is also a psychological component. Put them together, and you find that the bench players are often better than the starters - once they get in the game. But in the other environment they play their way back to the bench.
The third "T" is temperature. The Minnesota Vikings were famous for crushing all the warm-weather teams in the playoffs. Michigan and Nebraska do likewise in college. But when they meet a similar-climate team, the advantage goes away.
That's the "three T's." Roughly, you get 7 points for turf, 1 point for each hour of average-starting-time difference, and 1 point for every 10 wind-chill-humi-temp degrees.
With all of the "three T's," keep in mind that starters are determined based on performances in previous games. So by the time a team sees the effect of the 3 T's in action - whether credited to the 3 T's or not - it's often too late to do anything about it.
Does home field advantage exist? Until Oklahoma put down grass, the Nebraska-Oklahoma game actually had a slightly-negative HFA over the previous 30 years. Moreover, the 3 T's predicted that their rivalry should have been even. Michigan always lost the Rose Bowl until recently, even with substantially better teams than the Pac-10 representative. But the 3 T's predicted that too, and it also predicted that the Wolverines would close the gap substantially once they put down grass, which they did. The 3 T's says that the Clemson-Georgia and Auburn-Georgia rivalries should be turf-neutral, and they are. And the 3 T's says that Nebraska and Oklahoma lost 17 points when they played Florida State and Miami in the Orange Bowl in the 1980s - no wonder the Huskers and Sooners did so poorly. And yes, Nebraska would get those 17 points back when Miami plays in Lincoln, which might help explain why the Hurricanes have rejected all overtures for a home-and-home series. This scenario was repeated again in the 2002 BCS championship game. Miami was perhaps 10 points better than Nebraska, but the 3 T's gave Miami an additional 16 points, meaning Miami should have won by 26. Nebraska played well, losing by only 23, and nobody knew it.
The "three T's" is real. It exists. But is it all of HFA? Perhaps not. Crowd noise may play a factor. Familiar surroundings may play a role as well. But the "three T's" is most of it.
No, point spreads are not additive. While extreme differences are tougher to come by these days, they're still around a little.
The classic example was in the mid-1980s, when Brigham Young and Louisiana State were both so strong. BYU was a tremendous offensive team that would win 48-28. LSU was a defensive powerhouse that would beat those same types of teams 17-3. But had BYU and LSU played each other, the game would have been fairly even.
This also illustrates a big problem with "additive" ranking systems. Many computerized ranking systems severely miscalculated Alabama's strength in 1992. They looked at point spreads and saw an Alabama team that couldn't blow anybody out. What they failed to see was an Alabama team that needed only to kick a field goal each quarter in order to win every game. As a result, the "additive" systems bombed in predicting a Miami win in the Sugar Bowl, while the "percentage" systems generally predicted an Alabama rout and were later vindicated.
Updated 3-October-2007.
Copyright © 2007 by The Sports Report. All rights reserved.
Return to The Sports Report Home Page.