Welcome to the first 2007 installment of By the Numbers! Let's start with a refresher course on what these ratings are and are not. More details can be found at http://www.roanoke.edu/staff/minton/bynumbers.html. The basic form of the ratings comes from just looking at each team's points scored minus points against. A team that is on average losing by 6 points per game should have a rating that is 6 points less than the average of their opponents' ratings. A team that wins by 8 points per game should have a rating 8 points higher than their opponents'. Therefore, the difference in any two teams' ratings should be a point spread giving how much better one team is than the other. Also, strength of schedule is built in: if you win by 8 points per game, you add 8 to your opponents' rating. The better your opponents are, the higher your rating will be. The computer is only used to sort out the numbers for this system. This gives 40% of a team's rating. The other 60% comes from looking at wins and losses. In essence, this is pretending that every game is 20-0, and then applying the same logic as before.
That is all the ratings include. The system does not know when the games were played or which team was at home or anything else except schedule, wins and losses, and points for and against. I've always been amazed that this limited data produces anything worthwhile. One problem that I've always worried about is what to do when a Michigan plays an Appalachian State (if you're the announcer, just scream as loud as you can). The issue is that I only want to rate the 1-A teams. (I know, it's now Bowl Series teams, but I don't like the abbreviation.) In the past, I've usually counted all of the 1-AA teams as the same team. That is, when Michigan plays App State and Illinois plays Western Illinois, the computer thinks they both played the same team. This year, that became a simplification that I wasn't willing to make. Parity not only means that many ACC teams are happy to win 7 games, but also that 1-AA teams can play with 1-A teams. And play they have, recording several wins.
So this year, I'll track 240 teams instead of 120. For this week, I ran the ratings using all 240 teams and under the old assumption of all 1-AA teams being the same. The results are very interesting, especially if you like to argue about which conference is the best. LSU is #1 either way, but in the all-240 version they lead #2 South Florida (no Bull!) by half a point while in the 1-A-only version they lead #2 Arizona State by a full 7 points. Much to my surprise, the model with more teams has a much smaller spread. The other main difference is that the 1-A-only version likes the Pac 10, with Arizona State #2, Oregon #3 and Cal #5. These teams drop to #6, #9 and #11, respectively, in the all-240 version. (The city-wide collapse of LA sports teams didn't help.) Meanwhile, the ACC gets a huge boost. Boston College, for example, gets credit for playing a very good UMass team and jumps from #15 to #3. You'll see similar jumps for Hokies, Hoos and a host of ACC teams. Michigan, of course, is helped by getting credit for playing a good App State team, rising from #54 to #33. Oklahoma drops from #10 to #32 -- I have no clue why there is such a difference, but the strength of schedule is awful after games against #178 Utah State and #205 North Texas.
Which version is right? Neither. Which is better? I'll let you decide. I will say that the all-240 is fun if for no other reason than seeing Notre Dame at #110. Do I really think that Elon and the Fighting Blue Hens of Delaware could beat the Irish? Don't be silly. But Duke .... While I'm picking on teams, try to find Louisville in the list. That, to me, is a more suprising collapse than Notre Dame.
No really big games this week, but there are several games that will help decide who's in the top half and who's in the bottom half of conference standings. The ratings say that the UVa - UConn game is a battle of top 20 teams. There are several other games that go against common beliefs. It's early in the year, so it will be interesting to find out who's right. The following picks are straight from the ratings with 4 points added for home field.
Virginia Tech over Duke 10 (that's 8 more than VCU won by in basketball -- ah, the memories)
Virginia over Connecticut by 4 (both beat Pitt and Duke; who knows?)
Wake Forest over Florida State by 3 (an upset? depends on which new QB plays better)
South Carolina over North Carolina by 5 (another win for the Smelley gang?)
Illinois over Iowa by 13 (Juice tries to steal another Big 10 win for the Illini)
Michigan over Purdue by 2 (not App State, but another important test at the Big House)
Missouri over Oklahoma by 3 (maybe the computer should have respected OU sooner)
Texas A&M over Texas Tech by 2 (first real competition for Graham Harrell's flying circus)
LSU over Kentucky by 4 (big game #2 for the Cats, hoping for a different "Bluegrass Miracle")
San Jose State over Hawaii by 2 (an upset could refract the Rainbows' unbeaten dream)
The far right-hand column shows the ranking of each 1-A team using the old rating system.
|55||N Dakota St||39.30|
|98||New Mexico St||34.47||82|
|108||San Diego St||33.64||78|
|117||Wm & Mary||31.93|
|166||S Dakota St||26.85|