My nitpick: The rankings

[John Comey]

I am starting to chart the top 25 each week for my historical dynasty, and I'm noticing some things with the rankings that disturb me. Oklahoma dropped from #1 to #6 in the polls, despite drubbing South Florida 44-10. Michigan dropped from 4 to 5, despite beating #11 Notre Dame 34-3. Nebraska moved from 5 to 1 after beating Missouri 41-23, while Auburn moved up to #2 (from #7) after beating unranked Arkansas 26-9. The big mover was Texas A&M, who moved up to #3...from #17...after beating then-#16 K-State 20-18. Miami, ranked #9 the previous week, fell out of the poll altogether after losing to Va Tech 24-10 (Va Tech was ranked 25th). I know this is a very difficult area of the game to code...but there is no stability, which is one thing you find, at least somewhat, in the real polls. There *is* too much fluctuation, at the very least. No team jumps up fourteen spots after beating a team ranked one spot ahead of them by two. It just does not happen. I am writing this to see if it's my league that's messed up, or if this is something that may need looking at in the future. Thanks, John

[Roger Kaputnik]

The problem is that the game makes too much of strength of schedule. It is an important factor but it should be toned down a little bit. Top 25 teams play weak teams all of the time and don't go down in the rankings IRL.

[John Comey]

I can agree with that sentiment. Since this is going to get read, Michigan just hung up 68 on #11 Minnesota. They gave up 42. Either the Big Ten suddenly found a wealth of offense, or scoring is really high (my belief). -John

[Roger Kaputnik]

Well, the players each team gets are somewhat random and every game is different so I wouldn’t be too concerned over that. Maybe the problem is that the defenses suck.

[srudoff]

I have dropped from #1 in the rankings to #5 when I had a bye week around week 6 The teams that moved ahead me did not have any impressive wins that week

[ThrowTheBall]

rankings is probably one of the hardest parts of the game to code. Pre and during season polls dont really matter that much anyway, itt ends to usually be pretty close at the end when it counts.

[John Comey]

If you read, I specifically stated that it is one of the toughest parts of the game to code. Arlie and I have had conversations on this. But it still seems a bit out of whack, even given those conversations. I have other things I have to do first, but I'm going to come back to this and research this more when I have time. -John

[Old Boiler]

I played around with my own polling system for TCY, since theirs also exhibited unrealistic behavior. One thing i did was build in an "incumbency" factor, so that a team was unlikely to lose positions even if their rating fell. The closer a team was to the top, the stronger this effect was (very hard to dislodge a #1 team unless the #2 looked overwhelmingly better). Does BBCF use anything like that? On a similar topic, it would be kinda cute if the pointspread prediction was recorded in the boxscore or gamelog for future reference; as far as i can tell, that disappears forever once the game is played. This, of course, is the most miniscule of insignificant issues . . . .

[Jeremy Weimer]

This is something that definitely needs to be worked on and fixed. I'm playing a solo game with Michigan and I was the #1 ranked team to start the season. I proceeded to go 2-0 and then have a bye week. After the bye week I dropped down to #3 which makes no sense. Then week 4 I played #22 Michigan State (2-0) and beat them and after the win I dropped to #4 in the rankings. I should still be #1, there's no reason for me to have dropped at all.

[Miral]

[QUOTE=ThrowTheBall]rankings is probably one of the hardest parts of the game to code. [/QUOTE] This is true, and its because you are trying to follow the irrational logic of pollsters. There are several guys who have a seat on the polls who have said this year that as long as USC keeps winning, they should be ranked #1 regardless of how well they play, simply because they won the last two championships. Computer code only pays attention to this season, and doesn't weight a week 10 loss any heavier than a week 1 loss... unfortunately the real men who decide the final rankings do. That's my rant for the poll situation, I say the program is far more accurate than the AP.;)

[Arlie Rahn]

I think the problem is here is trying to "fairly" rank teams. There is some weight given to current poll ranking, but that really doesn't become a factor until midseason. So, in weeks 3-6, you may see some movement based on margin of victory, SOS, and "big wins" (wins against the top 25). In week 4 in BBCF, you may see a team that blows out a nice team move from 4 to 2 - even if the teams at 2 and 3 had a BYE or won against lesser opponents. This probably wouldn't happen in real life as pollsters hate to penalize a team that hasn't lost (even though some would argue the preseason polls do exactly that). In the end, I will keep trying to tweak this. As of right now, I think the polls from midseason on are pretty solid - it's just some of the fluctuation in the early weeks that bothers some people. I am hesitant to put more emphasis on prior weeks early in the season as many of the ranked teams in week 1 are often shown to be overrated by midseason. So, I'd rather have the polls be "fair" in weeks 8 and 9 than remove the fluctuation early and end up carrying some undeserving teams in the top 25. Still, I'll see if I can find a way to tone down some of the early variance.

[Roger Kaputnik]

I think Arlie is right on this one. There is some weird variability in the rankings in the early weeks but the right teams usually end up in the right places later in the year.