Football Rating Discussion
Friday, October 17, mid-morning
I haven't had much to say lately because I believe the ratings speak for themselves. It's very interesting just to watch the teams meander up and down the ratings all year, finally settling in to a reasonable position--which is what they're doing now. And as always, remember that these ratings are for amusement purposes only--essentially the same purpose as the non-region games that teams play!
Number of undefeated teams going into Week 8:
6A - 4; 5A - 4; 4A - 8; 3A - 6; 2A - 6; 1A - 6; Total number - 34
Saturday, October 11, 11 p.m.
The Washington County-Georgiana score has generated a bit of controversy. Honestly, I'm not sure who won the game; I received an email stating that WCHS won 42-0; the Associated Press has reported Georgiana winning 17-0; and message boards have posted both. I am waiting for my weekly email from the Birmingham News for confirmation. Either way, the score will remain with a Tuesday game date to denote that I am not confident I have a correct score.
Tuesday, October 7, 11 p.m.
The Disclaimer has moved front and center on the page. Apparently there was some controversy at the AHSAA generated by the site; for the sake of clarity (and CYA) I'll reiterate that this site is not affiliated with the AHSAA and is not official in any way. While yes, I do trust my ratings and I trust that the information presented here is absolutely correct, that does not mean my information takes precedence over official media releases by the Association. All results here are unofficial and should be treated as such.
If anyone thinks this site is unnecessary/useless/too controversial/etc., you should let me know. Until I hear otherwise, the ratings go on!
Saturday, October 4, 2:20 p.m.
No change to the ratings formula. Forget what I said below.
Undefeated teams remaining after Week 6:
6A - 8; 5A - 5; 4A - 10; 3A - 7; 2A - 7; 1A - 6. Total 43 (of 369 teams)
Friday, October 3, early morning
I'm considering something I shouldn't be considering at mid-season. A change in the formula.
Given all the griping about how some teams are doing well but rated way too low (some of which are legitimate complaints), there are ways to mitigate the "penalty" for playing weak opponents. By raising the value of those games (it's at 0.05 now), teams can have more input from those games toward their ratings. That is, if you play a horrible team, the penalty for that game will be a little less (right now you only get 5% "credit" for playing that game). I may change that number to 0.25, 0.33, or 0.50, depending on how the weekend's games go. For the record, I'm closely watching Cullman; Hamilton; West Morgan; Winfield; and others. Teams, beware! I think this small change will bring much more peace and happiness to those who pay attention to the ratings.
In other news, I got clearance today from Mr. John Wobus to reproduce his computer rankings on this site. I look forward to creating those pages and posting yet a second computer poll very soon. He uses a different formula, but I'm sure you'll like it. (Yeah, right.)
Enjoy the weekend's games! Since it is my hometown after all, I'll likely make the drive over and attend the Deshler/Hamilton game on Friday night.
Tuesday, Sept. 30, 10 p.m.
An important thing to remember is that wins that hurt and losses that help a team's rating have almost no impact on the results (they are worth 0.05 of a real game). What does this mean? It means that when you beat an 0-5 team, your rating doesn't go down as much as it probably should. Or when you lose to a 5-0 team, the "reward" for playing such a high-quality team is reduced. So even though I preach that strength of schedule is important, that effect is mitigated in some situations.
This makes good sense. The reward for losing to great teams is diminished--so each team's rating is probably closer to it's true value (whatever that value really is). As an example, three of Hamilton's five wins are against very weak teams and those wins hurt Hamilton's rating--thus the results of those games count only 5% as much as the other two games, which the computer feels are teams that are comparable in strength.
I think one of the other things happening to a lot of the northwest Alabama teams is what I call "schedule feedback"--everyone plays similar opponents for their non-region games. As a result, it becomes tougher to rate the teams compared to all other teams in the state, and the teams get bunched together with similar ratings that really don't mean much compared to other teams outside the region. I believe this is what's happening to teams with very good records but low ratings (3A - Lauderdale County, Rogers, Winfield, West Morgan; 4A - Hamilton, East Limestone). No, I'm not saying that all these teams are playing each other--they are all 4-1 or 5-0 so that is impossible. Instead, they are all playing the same poor teams and thus they all suffer. That's one theory why they are all rated (too) low. Enjoy!
Sunday, Sept. 28, noon
Well, the first negative feedback has come in. Someone has stated that "the ratings are a joke and speak for itself [sic]". I hardly think they are a joke, as the computer doesn't lie. But the best example of a perceived wrong is Hamilton, a 5-0 team that is rated 41st in the class.
I do not believe Hamilton is 41st. They are, by their 5-0 record, a legitimate top 10 contender. However, their poor performance against a struggling Lamar County (3A) team (Hamilton won 22-16) definitely showed that the team has flaws. Hamilton has played two weak 3A teams and only one team with a winning record. As I have often said, strength of schedule is extremely important to computer ratings. I can't say it enough. As of the end of Week 5, Region 8 is the weakest region in Class 4A--I look forward to seeing if the computer vindicates itself come playoff time.
In Class A, Loachapoka (3-2, rated #6) is being heavily rewarded for losses to Wadley (A, #2) and Clay County (A, #4). One loss to a poor team would put them back toward the middle of the class, where a 3-2 team probably belongs.
Note that the 2A teams that were in the top 20 last week are now gone...as expected! Lineville (2A) is now rated 22nd overall in the state, and with two solid wins over mediocre 4A teams, I can see why. They are playing at the 4A level and being rewarded for it.
Next week, almost all the games will be regional play, as will be the case for the next four weeks. Watch the playoff picture take shape, and enjoy the ratings!
Sunday, Sept. 21, 11 a.m.
All games from Week 4 are in. I'm still not pleased with how some of the 2A and 3A teams are at the top of the statewide ratings, but the non-region games next week should fix that.
I should note that after all my griping in previous weeks, after four weeks I consider these ratings as "reasonably" representative of the ratings of our state's football teams. I'm not a math major (only a minor!), but I know enough statistics and enough about football to now be confident in my work.
This week I want to spend a second talking about the lowest-rated undefeated teams. Example: 4A Hamilton (4-0) is rated 31 in their class. Bullock County (0-4!) is rated higher than Hamilton because Bullock has played four undefeated teams--and Hamilton has yet to play a team with a winning record. That (lack of) strength of schedule hurts you. I'm not knocking the Aggies, but they have lots to prove before they merit a top 10 computer rating. Daphne and Opelika, the two low-rated 6A teams that are both 4-0, suffer the same fate--their opponents are a combined 4-10 and 2-14, respectively. I know that scheduling strong opponents is more difficult at the high school level, but time and time again we've seen teams with great records in the regular season that have played weak schedules get demolished in the playoffs (many, many examples from all sports). Moral of the story--play quality non-region/non-area/non-conference opponents!!
Sunday, Sept. 14, late evening
All games from Week 3 are in. I suppose regional play encourages teams to report their scores more regularly! Anyway, here is some food for thought. The Colley system (link at the bottom on the front page) is more conducive to generating these stats since it has a much stronger emphasis on winning than classification in the early part of the season:
Number of Undefeated Teams Remaining
6A - 14; 5A - 14; 4A - 17; 3A - 13; 2A - 13; 1A - 12 (ASD doesn't count). Total number = 83!
That's a pretty healthy number of teams. I know there will be many that take on their first loss on Friday (my alma mater, Hackleburg, plays Courtland and will likely be one of those losing!).
So, why are the lower-classification teams sneaking up into the top ratings of the state? Well, as I have discussed earlier, the teams are "linking" themselves and each class is beginning to sort itself out in terms of ratings (for example, I believe Reeltown and Courtland are two of the top three 1A teams). However, now the whole state has to "sort itself out," so to speak. What I mean is that we need the coming week of region games so that everyone plays more games against teams of their caliber. The reason the classes get sorted out in the end is that in each class, they play a majority of similarly-ranked opponents (just like the Division I-A colleges, I-AA's, etc.). Thus, the 6A teams will mostly rise to the top, and the 1A teams will move to the lower end of the ratings. Those exceptional teams that are worthy of high ratings, however, will remain there. Who knows--Pisgah may be one of the top teams in the entire state. We'll see.
The reason Fyffe and Pisgah are right there together in the statewide top 10 is simple--they have both done well so far, and have played each other! One of the two--if not both--should disappear from the top 10 next week.
Saturday, Sept. 13, noon
Hello again. The ratings begin to take much better shape after this week of games, but there are still kinks to be ironed out (yes, this IS normal). There is still some "infiltration" of the statewide top 20 by a couple of 2A teams--namely, Pisgah and Fyffe--that probably don't deserve to be there. Their ratings have affected each other--both are doing well so far this year, and they have played one another. After another game, however, they should fall back down to around, say, 50th or 60th, which for exceptional 2A teams is probably not unreasonable.
Looking at the 6A ASWA poll and comparing it to the ratings, you might say, "Where's Prattville?" Well, two of their opponents are winless so far. That seriously hurts your early-season ratings. The team will have to "prove" to the computer that its 3-0 record is not just a byproduct of playing weak teams. John Carroll's opponents are 6-3 so far this year, and that is likely why they are rated so highly. Just a little food for thought.
Note again that the teams are not yet completely "linked" together--next week's region play will help, and the non-region games of week 5 will have a significant stabilizing impact on the ratings. Stay tuned!
Tuesday, Sept. 9, 11 p.m.
So, do I think Hackleburg is one of the top 10 teams in the state because the ratings say so? NO, NO, NO!
Computer ratings are unrepresentative after two weeks, just like human polls still are. Example: AUBURN. Ratings aren't useful until all the teams are "linked" together by playing one another. There are, in effect, probably 50 or 60 "groups" of teams that aren't tied together right now--they have only played each other. Remember how ESPN will occasionally say "Boise State deserved the national championship because they beat Idaho who beat Oregon State who beat California who beat so-and-so etc etc."? As crazy as stuff like that sounds, the computer systems need that because you must tie all the teams together for the system to work. Sorry to sound like a broken record, but there are really people that are convinced I think Fairhope (5A, 0-2) is #2 in Class 5A right now. Couldn't be farther from the truth.
Give the ratings another couple weeks to sort themselves out, just like the teams do. I promise. :-)
Sunday, Sept. 7, 11 a.m.
Hello again. Just a quick note to point out that the ratings are very unstable right now, because the teams aren't really "linked" well at all. That is, team A has played team B, who has played team C, who has played D, but they haven't played each other--which is what is important in computer ratings. Linking the teams to each other is crucial to having a good system...and that only starts to happen after 4 or 5 games. The next two weeks of region games will go a long way toward stabilizing the ratings. Don't be surprised if teams drop 100 points or add 100 points (or more!) to their rating to put them back where they should be!
Meanwhile, the list of unreported games adds up. I'll keep looking for these, but there is nothing I can do about it. Help me out if you can. Thanks!
Saturday, Sept. 6, 10 a.m.
Good morning everyone. It is my hope to have this page updated by 10 a.m. on every Saturday...since I know there are die-hards who just love seeing how badly the computers can mangle their team's real strength!
There is an interesting thing happening at the top of the ratings this week:
3=Class 5A "=Class 6A W L - - 1 3 St. Paul's 2 0 1291 " Daphne 2 0 1291 3 " Alma Bryant 1 1 1191 3 Fairhope 0 2 1191
4 p.m. Note: These numbers differ from what shows up on the "official" rating page because I've added scores since 10 a.m. The principle still applies, however.
Why, oh why, would Fairhope be rated the #3 team in the state??? If you know their schedule, it's simple: their two losses have come to St. Paul's and Daphne. Since both of these teams are undefeated, and have played only 5A and 6A teams, they get a fantastic rating...one that, in the next couple weeks, will slowly go down and be more representative.
This is a good example of what computer rating systems do early in the season. If I lose only to the best teams, I look like a strong loser to the computer (even if I lose 50-0). Remember, as in college football, strength-of-schedule does mean something. Of course, in real-life AHSAA play, it doesn't...but in the computer world, it does. Regardless of whether Fairhope beats Shaw (1-1, rating 691, rated #83 in the state) next week, their rating will go down to a somewhat more representative value.
Also...you can probably still see that there was a "division-by-class" of the original ratings. That is, I started the teams in each class out with different point values (to see what those values were, look at the numbers in the alignment list; Class A started with 106, etc.). I promise that importance will be negligible by the end of the year. When I tested last year's games (entering them week by week), the original rating was almost gone by week 5. A team rates highly by beating good teams, not losing to high-rated teams. The "fake" effect of class separation will be gone in another few weeks.