Don't stress it, by week 8 all the kinks in the power rankings tend to be worked out.
rellascout
offline
offline
Originally posted by RoseBowlChamps
Don't stress it, by week 8 all the kinks in the power rankings tend to be worked out.
Yup it is just looking at raw data points..... there are not enough data points yet to be meaningful.
Don't stress it, by week 8 all the kinks in the power rankings tend to be worked out.
Yup it is just looking at raw data points..... there are not enough data points yet to be meaningful.
Alex Riggs
offline
offline
Exactly, this is used off a spreadsheet that figures in who you play, who they played, and how much was scored. Kinda BCS like. It will take several games to figure it all out. But like my team the Mudbugs we lost one and won the next. Yet we dropped a ranking because of the fact that the one team we beat lost 2 straight, so really we havnt proved anything
Kiber
offline
offline
Well - the Gate won 2 games by a combined 410 to 9. But I'm still fine with our fairly low ranking. We played weak teams and merely did what we should have done. No reason to be much higher for simply doing what was expected. Plus - it's pre-season, so it doesn't matter. I still like looking at the data though - after all, isn't that why we play this game.
fujicrow
offline
offline
If it's just calculating raw data then there's a problem somewhere because you have 0-2 teams that lost by landslides ahead of 1-1 teams. The rankings are so random I can't really see the accuracy in the system. Even if it is just 2 games into the pre-season, if your system is getting them this mixed up there's a problem with it. I'm sure it might look better as the season progresses as the good teams will win more games by more points against tougher teams, but an accurate system would handle the rankings properly from day 1. Sure bad teams might be ranked ahead of good early on, but you should be able to look at it and see by the numbers why that happened. These rankings are just too random for me to trust the accuracy of your system.
UCFKnights555
offline
offline
Originally posted by fujicrow
If it's just calculating raw data then there's a problem somewhere because you have 0-2 teams that lost by landslides ahead of 1-1 teams. The rankings are so random I can't really see the accuracy in the system. Even if it is just 2 games into the pre-season, if your system is getting them this mixed up there's a problem with it. I'm sure it might look better as the season progresses as the good teams will win more games by more points against tougher teams, but an accurate system would handle the rankings properly from day 1. Sure bad teams might be ranked ahead of good early on, but you should be able to look at it and see by the numbers why that happened. These rankings are just too random for me to trust the accuracy of your system.
whats the point of having power rankings if bad teams are ahead of good teams. In statistics you learn that you can't get a dependable statistic without a decent sample. Just because he needs more data, doesn't mean his system doesn't work. By your argument I could poll 5 people and it should be just as dependable as polling 5,000. I'd rather wait for good power rankings, then just seeing a bunch of teams in a random order, just for the sake of having power rankings
If it's just calculating raw data then there's a problem somewhere because you have 0-2 teams that lost by landslides ahead of 1-1 teams. The rankings are so random I can't really see the accuracy in the system. Even if it is just 2 games into the pre-season, if your system is getting them this mixed up there's a problem with it. I'm sure it might look better as the season progresses as the good teams will win more games by more points against tougher teams, but an accurate system would handle the rankings properly from day 1. Sure bad teams might be ranked ahead of good early on, but you should be able to look at it and see by the numbers why that happened. These rankings are just too random for me to trust the accuracy of your system.
whats the point of having power rankings if bad teams are ahead of good teams. In statistics you learn that you can't get a dependable statistic without a decent sample. Just because he needs more data, doesn't mean his system doesn't work. By your argument I could poll 5 people and it should be just as dependable as polling 5,000. I'd rather wait for good power rankings, then just seeing a bunch of teams in a random order, just for the sake of having power rankings
Kiber
offline
offline
Originally posted by Alex Riggs
Rank … Strength of Record … Quality of Play … Team
1 ....................... 1 ........................ 0 ................. Madison Street Militia (2/0)
2 ....................... 1 ........................ 0 ................. Baby Giants (1/1)
3 ....................... 1 ........................ 0 ................. Fredericksburg Cavaliers (2/0)
4 ....................... 1 ........................ 0 ................. Minnesota Wolverines (2/0)
5 ....................... 1 ........................ 0 ................. NBA JAM (1/1)
6 ....................... 1 ........................ 0 ................. Orlando Knights (1/1)
7 ....................... 0 ........................ 0 ................. Cincinnati Tigers (2/0)
8 ....................... 0 ........................ 0 ................. Fort Sam Houston Rangers (2/0)
9 ....................... 0 ........................ 0 ................. Radio City Irrelevants (0/2)
10 ...................... 0 ........................ 0 ................. Colorado Springs Flyers (0/2)
11 ...................... 0 ........................ 0 ................. Ethiopian Doughnut Chasers (2/0)
12 ...................... 0 ........................ 0 ................. Bunghole Liquors (0/2)
13 ...................... 0 ........................ 0 ................. Glasgow Mongolians (0/2)
14 ...................... 0 ........................ 0 ................. Chicago Lazer Squirrels (0/2)
15 ...................... 0 ........................ 0 ................. Burnie Bruisers (0/2)
16 ...................... 0 ........................ 0 ................. KC Gorillas Farm Team (0/2)
17 ...................... 0 ........................ 0 ................. West Lawn Bulldogs (2/0)
18 ...................... 0 ........................ 0 ................. Malibu Mercenaries (0/2)
19 ...................... 0 ........................ 0 ................. Pensacola Wolfpac (2/0)
20 ...................... 0 ........................ 0 ................. Pittstown Pirate Hookers (0/2)
21 ...................... 0 ........................ 0 ................. The Jigg City Blitz (2/0)
22 ...................... 0 ........................ 0 ................. Steamboat Springs Mountaineers (2/0)
23 ...................... 0 ........................ 0 ................. San Francisco Gold Rush (2/0)
24 ...................... 0 ........................ 0 ................. St. James's Gate (2/0)
25 ...................... 0 ........................ 0 ................. Lake Superior Squall (2/0)
26 ...................... 0 ........................ 0 ................. Springfield Isotopes (0/2)
27 ...................... -1 ........................ 0 ................. Lafayette Mudbugs (1/1)
28 ...................... -1 ........................ 0 ................. Cleveland Greyhounds (1/1)
29 ...................... -1 ........................ 0 ................. San Diego Cyberballers (0/2)
30 ...................... -1 ........................ 0 ................. Charleston Swamp Foxes (0/2)
31 ...................... -1 ........................ 0 ................. Richmond Rams (1/1)
32 ...................... -1 ........................ 0 ................. Sin City Revelation (0/2)
OK - I just added in the win/loss record for each team (I hope it's OK to edit his post this way - if not, I apologize). This does NOT mean that I think the rankings are screwed up - I have no idea what the calcs are that he used. And I absolutely think that any rankings won't become valid until we get some more data to play with. I just wanted to toss out the win/loss record to give people some more datapoints to play with.
As you can see, there's a big batch of 2/0 teams sitting around 21-25. If they can keep up their winning ways, they can hope to move up quickly, I would imagine. Meanwhile - of the teams ranked 9th through 16th, only 1 has any wins at all. So . . . . . .
Rank … Strength of Record … Quality of Play … Team
1 ....................... 1 ........................ 0 ................. Madison Street Militia (2/0)
2 ....................... 1 ........................ 0 ................. Baby Giants (1/1)
3 ....................... 1 ........................ 0 ................. Fredericksburg Cavaliers (2/0)
4 ....................... 1 ........................ 0 ................. Minnesota Wolverines (2/0)
5 ....................... 1 ........................ 0 ................. NBA JAM (1/1)
6 ....................... 1 ........................ 0 ................. Orlando Knights (1/1)
7 ....................... 0 ........................ 0 ................. Cincinnati Tigers (2/0)
8 ....................... 0 ........................ 0 ................. Fort Sam Houston Rangers (2/0)
9 ....................... 0 ........................ 0 ................. Radio City Irrelevants (0/2)
10 ...................... 0 ........................ 0 ................. Colorado Springs Flyers (0/2)
11 ...................... 0 ........................ 0 ................. Ethiopian Doughnut Chasers (2/0)
12 ...................... 0 ........................ 0 ................. Bunghole Liquors (0/2)
13 ...................... 0 ........................ 0 ................. Glasgow Mongolians (0/2)
14 ...................... 0 ........................ 0 ................. Chicago Lazer Squirrels (0/2)
15 ...................... 0 ........................ 0 ................. Burnie Bruisers (0/2)
16 ...................... 0 ........................ 0 ................. KC Gorillas Farm Team (0/2)
17 ...................... 0 ........................ 0 ................. West Lawn Bulldogs (2/0)
18 ...................... 0 ........................ 0 ................. Malibu Mercenaries (0/2)
19 ...................... 0 ........................ 0 ................. Pensacola Wolfpac (2/0)
20 ...................... 0 ........................ 0 ................. Pittstown Pirate Hookers (0/2)
21 ...................... 0 ........................ 0 ................. The Jigg City Blitz (2/0)
22 ...................... 0 ........................ 0 ................. Steamboat Springs Mountaineers (2/0)
23 ...................... 0 ........................ 0 ................. San Francisco Gold Rush (2/0)
24 ...................... 0 ........................ 0 ................. St. James's Gate (2/0)
25 ...................... 0 ........................ 0 ................. Lake Superior Squall (2/0)
26 ...................... 0 ........................ 0 ................. Springfield Isotopes (0/2)
27 ...................... -1 ........................ 0 ................. Lafayette Mudbugs (1/1)
28 ...................... -1 ........................ 0 ................. Cleveland Greyhounds (1/1)
29 ...................... -1 ........................ 0 ................. San Diego Cyberballers (0/2)
30 ...................... -1 ........................ 0 ................. Charleston Swamp Foxes (0/2)
31 ...................... -1 ........................ 0 ................. Richmond Rams (1/1)
32 ...................... -1 ........................ 0 ................. Sin City Revelation (0/2)
OK - I just added in the win/loss record for each team (I hope it's OK to edit his post this way - if not, I apologize). This does NOT mean that I think the rankings are screwed up - I have no idea what the calcs are that he used. And I absolutely think that any rankings won't become valid until we get some more data to play with. I just wanted to toss out the win/loss record to give people some more datapoints to play with.
As you can see, there's a big batch of 2/0 teams sitting around 21-25. If they can keep up their winning ways, they can hope to move up quickly, I would imagine. Meanwhile - of the teams ranked 9th through 16th, only 1 has any wins at all. So . . . . . .
Last edited Apr 21, 2009 09:11:57
sushil33t
offline
offline
larger sample = lower error because you are able to model the error better based on the variance from each data point. These power rankings, while providing some fun for us before the season begins, bears no relevance on what will actually happen. You shouldn't get worked up about it at the very least.
fujicrow
offline
offline
I'm not getting worked up over it, just making the point that the formula obviously has flaws. Of course as the season wears on it looks better, mainly because the W/L record will matter more because there's "more data" in that regard. But to argue that because it makes sense once you get more than halfway through the season is kind of silly. A good formula, regardless of what you're calculating, should be relevant from start to finish. I'm not sure exactly what formula is being used now, but you shouldn't have a situation where certain criteria is more or less important within the formula depending on when in the season it's being calculated. If that's the case there'd truly only be one calculation where the formula would reach its peak in accuracy, in that case you might as well just be making the rankings up off the top of your head.
I'm not saying you shouldn't use a statistical formula in attempting to rank teams, just that you should try to perfect it rather than just running with any old formula you've found somewhere. You could do this simply by changing the importance each critera has in deciding the final results. As it is now it seems, at least to me, that W/L and "quality of play" (not really sure what that represents) seems to each be worth 50% of the teams final ranking "score".
I've seen people use BCS rankings as reference, but this is nothing similar really. BCS rankings mostly use a combination of human and computer polls, w/l record, and strength of schedule in it's formula. This is really just 1 computer generated poll with a rather suspect formula. First we need to determine what "quality of play" really means, because I think we can all agree that it would be difficult for a formula to determine "quality" of anything as it's a pretty subjective concept. If one team shuts out a 0-15 team and another beats a 15-0 team by 1, which one had better "quality of play"? I'm assuming you'd have to involve strength of schedule into the formula, but to what degree? That's the sort of thing that typically gets sorted out in human created polls better than computer generated ones because a formula, no matter how complex, can't be subjective.
In my opinion, you'd be best off with a formula using W/L and strength of schedule, which is much easier attained, with a ratio probably around 75/25 in favor of W/L. That would probably be the simplest way to do it and more than likely more accurate as well. This is "just a game", but if you're going to go through the trouble of doing something, you should at least take pride in doing it right.
I'm not saying you shouldn't use a statistical formula in attempting to rank teams, just that you should try to perfect it rather than just running with any old formula you've found somewhere. You could do this simply by changing the importance each critera has in deciding the final results. As it is now it seems, at least to me, that W/L and "quality of play" (not really sure what that represents) seems to each be worth 50% of the teams final ranking "score".
I've seen people use BCS rankings as reference, but this is nothing similar really. BCS rankings mostly use a combination of human and computer polls, w/l record, and strength of schedule in it's formula. This is really just 1 computer generated poll with a rather suspect formula. First we need to determine what "quality of play" really means, because I think we can all agree that it would be difficult for a formula to determine "quality" of anything as it's a pretty subjective concept. If one team shuts out a 0-15 team and another beats a 15-0 team by 1, which one had better "quality of play"? I'm assuming you'd have to involve strength of schedule into the formula, but to what degree? That's the sort of thing that typically gets sorted out in human created polls better than computer generated ones because a formula, no matter how complex, can't be subjective.
In my opinion, you'd be best off with a formula using W/L and strength of schedule, which is much easier attained, with a ratio probably around 75/25 in favor of W/L. That would probably be the simplest way to do it and more than likely more accurate as well. This is "just a game", but if you're going to go through the trouble of doing something, you should at least take pride in doing it right.
fujicrow
offline
offline
If you wanted to be more complex, go with 70/20/10 with W/L-Strength of schedule-Margin of victory. This would avoid the nasty ties early on in the season and wouldn't be that much more difficult to calculate.
Alex Riggs
offline
offline
The way it works is it ranks you first by strength of schedule. Then by quality of play. Quality of play is based on the scores of each game compared to what the other team scored on each game. Right now theres just not enough data for the quality of play section. Im going to throw a couple weeks worth of games and simulate results to be sure it looks more accurate and post something later tonight. That way you feel better
fujicrow
offline
offline
See, that's the problem with the formula then. You're not really factoring actual wins and losses except for what part they play into "quality of play". At least now I understand how winless teams can be ranked ahead of teams with perfect records thus far. In any sport the greatest determining factor for determining success is whether you win or lose your games. Some teams win pretty, some teams win ugly. Some teams have difficult schedules, some teams have cake schedule. At the end of the day though, it's whether the team is winning or not that determines whether they're having a successfull year, strength of schedule and other factors simply help to rank teams as there will be teams with near equal records.
Playing in a 16 team division and playing 16 games per season, most teams are going to have very similar strength of schedule and since W/L record and all other factors are all combined by appearances in equal ratios into "quality of play", this gives an abnormally large weight to scoring margins. If a team has a horrible record but ends with a favorable "quality of play" score, they could edge out teams with much better records simply because they happened to have one weaker opponent than the other.
Furthermore, I don't think the formula you're using is accurately ranking strength of schedule even. My team played against your #8 and #9 teams who are both undefeated but have 0 for strength of schedule, whereas your number one team played against the #2 and #30 and the number two team played against the #1 and #28. It seems they've achieved the #1 and #2 spots in this ranking mostly due to the fact that they played each other. Both teams played one team with -1 strength of schedule and the one against eachother, so that begs the question, if their +1 came from playing each other, how did they get flagged as being a "tough" game when neither had played anyone else but -1 teams? The Cyberballers played against one team with +1 strength of schedule and another with -1, and finished themselves with a -1 strength of schedule. You'd think it'd at the very least even out. Just seems a little suspect to me is all, I can't look at it from any angle to where it actually makes sense, and I usually go by the rule that if a formula can't be solved then it's bunk.
Playing in a 16 team division and playing 16 games per season, most teams are going to have very similar strength of schedule and since W/L record and all other factors are all combined by appearances in equal ratios into "quality of play", this gives an abnormally large weight to scoring margins. If a team has a horrible record but ends with a favorable "quality of play" score, they could edge out teams with much better records simply because they happened to have one weaker opponent than the other.
Furthermore, I don't think the formula you're using is accurately ranking strength of schedule even. My team played against your #8 and #9 teams who are both undefeated but have 0 for strength of schedule, whereas your number one team played against the #2 and #30 and the number two team played against the #1 and #28. It seems they've achieved the #1 and #2 spots in this ranking mostly due to the fact that they played each other. Both teams played one team with -1 strength of schedule and the one against eachother, so that begs the question, if their +1 came from playing each other, how did they get flagged as being a "tough" game when neither had played anyone else but -1 teams? The Cyberballers played against one team with +1 strength of schedule and another with -1, and finished themselves with a -1 strength of schedule. You'd think it'd at the very least even out. Just seems a little suspect to me is all, I can't look at it from any angle to where it actually makes sense, and I usually go by the rule that if a formula can't be solved then it's bunk.
Last edited Apr 21, 2009 19:29:51
fujicrow
offline
offline
I'm not trying to be argumentative, your system just confounds me no end. It seems to me that it could be done more accurate, and simpler so that there's no need to "add stats" to the equation, it'd be able to calculate statistics from day 1. Just seems like whoever created that spreadsheet made it alot more convoluted that it needs to be, to the detriment of accuracy and for some people, sanity.
Dpride59
offline
offline
Originally posted by fujicrow
I'm not trying to be argumentative, your system just confounds me no end. It seems to me that it could be done more accurate, and simpler so that there's no need to "add stats" to the equation, it'd be able to calculate statistics from day 1. Just seems like whoever created that spreadsheet made it alot more convoluted that it needs to be, to the detriment of accuracy and for some people, sanity.
He is using the most widely accepted Glb spreadsheet. 2 preseason games in it is not exact but it doesn't take team builds into affect. A true rankign would take a human vote into effect. San Diego lost to two of the better teams in this conference so they should be ranked in the top 8 still. According to you, they shouldn't be in the top 20. I Think it is just fine, and will even itself out as the season progresses. NBA JAM ran 100% blitz last game. The preseason is for testing things. I can promise you the next two games will be much more reflective of the style I plan on running. I wouldn't worry about wins/losses/ranking atm. The regular season is what its all about. Get your popcorn ready!
I'm not trying to be argumentative, your system just confounds me no end. It seems to me that it could be done more accurate, and simpler so that there's no need to "add stats" to the equation, it'd be able to calculate statistics from day 1. Just seems like whoever created that spreadsheet made it alot more convoluted that it needs to be, to the detriment of accuracy and for some people, sanity.
He is using the most widely accepted Glb spreadsheet. 2 preseason games in it is not exact but it doesn't take team builds into affect. A true rankign would take a human vote into effect. San Diego lost to two of the better teams in this conference so they should be ranked in the top 8 still. According to you, they shouldn't be in the top 20. I Think it is just fine, and will even itself out as the season progresses. NBA JAM ran 100% blitz last game. The preseason is for testing things. I can promise you the next two games will be much more reflective of the style I plan on running. I wouldn't worry about wins/losses/ranking atm. The regular season is what its all about. Get your popcorn ready!
You are not logged in. Please log in if you want to post a reply.
