Make sure that you haven't repeated any teams!
Honor system 1 ballot per voter per week (I don't want to collect email addresses to limit it on my end)
A wild and wacky week last week. One with ballot implications for sure
Yeah it’s just a part of how my poll is set up. There can be some goofyness sometimes, but it usually works out by the end of the season.
I use the Pre-Season AP rankings as a starting point, purely to give the early weeks some kind of structure. Those rankings are applied as a diminishing factor from Week 0 through Week 6. After Week 6 my rankings are only based on W/L record and opponent’s rankings.
Through Week 10, Georgia has the weakest strength of schedule of all Power 5 teams, with their opponents ranking 82 on average. For comparison, the average ranking of JMU’s opponents is 74. So JMU’s wins are worth more at the moment. Those SOS rankings are re-calculated each week, so they’re going to change over the last weeks of the season. If UGA wins out, they’ll be fine.
This also explains the K-State/Troy discrepancy. Yes K-State beat Troy earlier in the season, but my spreadsheet doesn’t really care who you beat, it only cares about what your opponent is ranked and whether you won or lost. Teams don’t get a head-to-head boost against previous opponents.
I like this ranking system in part because it takes all the human emotion out of it. Teams are ranked only based on the results of their games and how strong their opponents are ranked, so their name or conference or my personal feelings of where they should be ranked don’t factor into it. It also looks at their entire body of work evenly, so early games aren’t weighted differently than late-season games, which is different from how AP and most human polls do it.
Of course, one of the obvious drawbacks of this system is highlighted by the UGA scenario: the best team in the country could be cursed with a bad schedule. For that I rely on the other eye test based polls in c/cfb to balance me out. I bring the robotic analysis, I’ll let others handle the vibes.
Thanks for the in-depth reasoning. As I mentioned in a previous week, I’m totally cool with this approach (and my poll also disregards head-to-head, but it is always an interesting argument against our power-rating-style computer polls).
Any plans to adjust the methodology with realignment kicking into high gear next year? I foresee lots of strong teams stuck playing each other, resulting in everyone picking up losses and looking worse on paper than in reality (as already happens with the PAC every year). I, for one, plan on restructuring my algorithm since it was designed based on the assumption that the 5 “power” conferences (+ND and BYU) were relatively equal.
Totally agree…we’ve chatted about this before, but a lot of the fun of voting in this poll for me is seeing how everyone does it differently. We’re trying to answer the question of who is the “best” team, but that word means something a little bit different to everybody.
Yeah I try to add improvements every year based on stuff I see during the season and I’ve been thinking of ways to re-tool it for next season, especially with the weirdness of realignment coming and some of the odd results I’ve gotten this season. I haven’t figured out exactly what I want to do yet, because I don’t want to add too much of my own biases into the formula. For example, I don’t want to give an artificial boost to teams just based on conference affiliations, because if we really do get an excellent team in a lower conference I don’t want my spreadsheet pushing them down just because of that. This season I’ve been running a few modified spreadsheets on the side just to play with, but I haven’t landed on anything I really like yet. Thankfully the offseason is long and I have plenty of data to tinker with to keep dialing it in.
I’m curious about yours, how much does it factor in stats throughout the season? I think you mentioned before that pre-season expectations are factored out at this point, do you keep it updated with performance stats every week?
It’s mostly just margin of victory/defeat. I’ve tinkered with adding more advanced stats, but then it feels like I’m geeking out too much and the tail is wagging the dog. At the end of the day, your team just needs to score more than the other team and if you can do that consistently, especially against good competition, you should be rewarded.
One criterion I really would like to tinker with is home field advantage. Right now it’s just a +4 point modifier for everyone, but there are definitely some environments where it’s worth more and some where it’s less. I haven’t found good public data on that (I’m sure all the bookies have that down pat though) so it’ll take a fair bit of statistical analysis to ascertain what those values are. Maybe next off-season.
Yeah it’s just a part of how my poll is set up. There can be some goofyness sometimes, but it usually works out by the end of the season.
I use the Pre-Season AP rankings as a starting point, purely to give the early weeks some kind of structure. Those rankings are applied as a diminishing factor from Week 0 through Week 6. After Week 6 my rankings are only based on W/L record and opponent’s rankings.
Through Week 10, Georgia has the weakest strength of schedule of all Power 5 teams, with their opponents ranking 82 on average. For comparison, the average ranking of JMU’s opponents is 74. So JMU’s wins are worth more at the moment. Those SOS rankings are re-calculated each week, so they’re going to change over the last weeks of the season. If UGA wins out, they’ll be fine.
This also explains the K-State/Troy discrepancy. Yes K-State beat Troy earlier in the season, but my spreadsheet doesn’t really care who you beat, it only cares about what your opponent is ranked and whether you won or lost. Teams don’t get a head-to-head boost against previous opponents.
I like this ranking system in part because it takes all the human emotion out of it. Teams are ranked only based on the results of their games and how strong their opponents are ranked, so their name or conference or my personal feelings of where they should be ranked don’t factor into it. It also looks at their entire body of work evenly, so early games aren’t weighted differently than late-season games, which is different from how AP and most human polls do it.
Of course, one of the obvious drawbacks of this system is highlighted by the UGA scenario: the best team in the country could be cursed with a bad schedule. For that I rely on the other eye test based polls in c/cfb to balance me out. I bring the robotic analysis, I’ll let others handle the vibes.
Thanks for the in-depth reasoning. As I mentioned in a previous week, I’m totally cool with this approach (and my poll also disregards head-to-head, but it is always an interesting argument against our power-rating-style computer polls).
Any plans to adjust the methodology with realignment kicking into high gear next year? I foresee lots of strong teams stuck playing each other, resulting in everyone picking up losses and looking worse on paper than in reality (as already happens with the PAC every year). I, for one, plan on restructuring my algorithm since it was designed based on the assumption that the 5 “power” conferences (+ND and BYU) were relatively equal.
Totally agree…we’ve chatted about this before, but a lot of the fun of voting in this poll for me is seeing how everyone does it differently. We’re trying to answer the question of who is the “best” team, but that word means something a little bit different to everybody.
Yeah I try to add improvements every year based on stuff I see during the season and I’ve been thinking of ways to re-tool it for next season, especially with the weirdness of realignment coming and some of the odd results I’ve gotten this season. I haven’t figured out exactly what I want to do yet, because I don’t want to add too much of my own biases into the formula. For example, I don’t want to give an artificial boost to teams just based on conference affiliations, because if we really do get an excellent team in a lower conference I don’t want my spreadsheet pushing them down just because of that. This season I’ve been running a few modified spreadsheets on the side just to play with, but I haven’t landed on anything I really like yet. Thankfully the offseason is long and I have plenty of data to tinker with to keep dialing it in.
I’m curious about yours, how much does it factor in stats throughout the season? I think you mentioned before that pre-season expectations are factored out at this point, do you keep it updated with performance stats every week?
It’s mostly just margin of victory/defeat. I’ve tinkered with adding more advanced stats, but then it feels like I’m geeking out too much and the tail is wagging the dog. At the end of the day, your team just needs to score more than the other team and if you can do that consistently, especially against good competition, you should be rewarded.
One criterion I really would like to tinker with is home field advantage. Right now it’s just a +4 point modifier for everyone, but there are definitely some environments where it’s worth more and some where it’s less. I haven’t found good public data on that (I’m sure all the bookies have that down pat though) so it’ll take a fair bit of statistical analysis to ascertain what those values are. Maybe next off-season.