View Full Version : Yet another computer rankings set
Huckleberry
11-17-2008, 10:04 PM
The Ratings (http://www.barkingcarnival.com/wp-content/uploads/2008/11/txhsfb.htm)
Second Round Preview (http://www.barkingcarnival.com/huckleberry/high-school-playoffs-second-round-preview)
In the time-honored tradition of internet message boards, I'm sure there will be questions, both polite and otherwise. :)
I'll do my best to answer them.
The Original
11-17-2008, 10:17 PM
Pearland #59 in state...
Does that even make any sense?? :confused:
Pearland #59 in state...
Does that even make any sense?? :confused:
#45 in 5A.
Fleeman93
11-17-2008, 10:21 PM
Pearland #59 in state...
Does that even make any sense?? :confused:
Sounds about right by this poll.
:D
Saint210
11-17-2008, 10:22 PM
Bleh! My school is like #185.
We will prove everyone wrong :D
LP-79
11-17-2008, 10:22 PM
At least they got the 4A #1 right.
The Ratings (http://www.barkingcarnival.com/wp-content/uploads/2008/11/txhsfb.htm)
Second Round Preview (http://www.barkingcarnival.com/huckleberry/high-school-playoffs-second-round-preview)
In the time-honored tradition of internet message boards, I'm sure there will be questions, both polite and otherwise. :)
I'll do my best to answer them.
I noticed one thing right off the bat.. Colleyville Heritage is 9-2, not 8-2.
By the way, nice first post... And welcome to the board.
Huckleberry
11-17-2008, 10:26 PM
Pearland #59 in state...
Does that even make any sense?? :confused:
Pearland is #59 in the overall ratings partially because of their relatively weak schedule. It's not their fault that the Pasadena schools are so bad, but it hampers them here. That's obviously a very superficial explanation, but it's a key aspect.
Huckleberry
11-17-2008, 10:29 PM
I noticed one thing right off the bat.. Colleyville Heritage is 9-2, not 8-2.
By the way, nice first post... And welcome to the board.
Thank you for the kind words.
Only games against rated teams are included, so Heritage's win over Clovis East is not included although it's in the data file.
Thank you for the kind words.
Only games against rated teams are included, so Heritage's win over Clovis East is not included although it's in the data file.
I gotcha.. That makes since to me. Thanks for the explanation.
CoveMom
11-17-2008, 11:20 PM
.
The Ratings (http://www.barkingcarnival.com/wp-content/uploads/2008/11/txhsfb.htm) Let's call this one Exhibit A
Second Round Preview (http://www.barkingcarnival.com/huckleberry/high-school-playoffs-second-round-preview) And this one will be Exhibit B
In the time-honored tradition of internet message boards, I'm sure there will be questions, both polite and otherwise. :)
I'll do my best to answer them.
This one might be a little tricky:
In Exhibit A, Garland Naaman Forest is ranked #10. Very impressive.
Copperas Cove is ranked #38. A bit low it seems to me, oh well.
HOWEVER.....
In Exhibit B, Copperas Cove is favored to win the Second Round Matchup with Garland Naaman Forest.
Lord, I am so :confused:
Don't you just love rankings? :D
Huckleberry
11-17-2008, 11:34 PM
This one might be a little tricky:
In Exhibit A, Garland Naaman Forest is ranked #10. Very impressive.
Copperas Cove is ranked #38. A bit low it seems to me, oh well.
HOWEVER.....
In Exhibit B, Copperas Cove is favored to win the Second Round Matchup with Garland Naaman Forest.
Lord, I am so :confused:
Don't you just love rankings? :D
The preview page predictions are based on the power ratings on the right half of the ratings page. Whereas the Season Rankings on the left half are intended to rank the teams based on what they've accomplished to date, the Power Ratings on the right half of the page are intended to predict future performance based on past performance.
The preview page predictions are rounded, but to predict a final score, subtract Team B's defense rating from Team A's offense rating and vice versa, taking into account the homefield advantage as necessary. There are problems at the extremes (negative point predictions in extreme mismatches), but that is the method.
Copperas Cove has the #11 power rating in 5A.
teamkillbot1022
11-17-2008, 11:37 PM
what are "pmv" and "nmv"?
Huckleberry
11-17-2008, 11:40 PM
what are "pmv" and "nmv"?
Just shorthand for Pure Margin of Victory and No Margin of Victory. They are the components that combine to produce the overall ratings on the left. The PMV set uses the actual score of the game to determine the share of the game score that the winning team receives (in this version the winning team's game score can get very close to 1/2). The NMV set doesn’t look at the score and assigns a predetermined share to each winner.
Essentially the game score is a number assigned to each team representing the probability that they would win a rematch based on the actual result. That is done for every game and then a maximum likelihood estimate is obtained based on the season's results.
The Original
11-17-2008, 11:40 PM
I think we've all learned something new today.
Don't use computers.
:D
mojotrain
11-17-2008, 11:42 PM
The Ratings (http://www.barkingcarnival.com/wp-content/uploads/2008/11/txhsfb.htm)
Second Round Preview (http://www.barkingcarnival.com/huckleberry/high-school-playoffs-second-round-preview)
In the time-honored tradition of internet message boards, I'm sure there will be questions, both polite and otherwise. :)
I'll do my best to answer them.
You can find out how to empty the boot by reading the directions on the bottom of the heel.
Just shorthand for Pure Margin of Victory and No Margin of Victory. They are the components that combine to produce the overall ratings on the left. The PMV set uses the actual score of the game to determine the share of the game score that the winning team receives (in this version the winning team's game score can get very close to 1/2). The NMV set doesn’t look at the score and assigns a predetermined share to each winner.
Essentially the game score is a number assigned to each team representing the probability that they would win a rematch based on the actual result. Tht is done for every game and then a maximum likelihood estimate is obtained based on the season's results.
I like it sir.. I'm very impressed with what you have done.. Alot of thought has went in to creating these, and it seems you have covered the bases.. Very good job..:notworthy
teamkillbot1022
11-17-2008, 11:52 PM
thanks. good job, btw, this is really good. any chance we might be able to see a "theoretical matchup" program that doesnt involve a 4.0 in calculus (not that i dont have one:D (jk, i have some form of B+ or A-)) to churn out these predictions like on the massey ratings site?
i know, its asking a lot, but i was just wondering...
Huckleberry
11-17-2008, 11:56 PM
thanks. good job, btw, this is really good. any chance we might be able to see a "theoretical matchup" program that doesnt involve a 4.0 in calculus (not that i dont have one:D (jk, i have some form of B+ or A-)) to churn out these predictions like on the massey ratings site?
i know, its asking a lot, but i was just wondering...
That is definitely a goal in the future. But my web programming skills are way subpar. Even worse than some of the posters on this thread think my computer ratings are. :D
keep your day job all teams poll is just silly. The Ratings (http://www.barkingcarnival.com/wp-content/uploads/2008/11/txhsfb.htm)
Second Round Preview (http://www.barkingcarnival.com/huckleberry/high-school-playoffs-second-round-preview)
In the time-honored tradition of internet message boards, I'm sure there will be questions, both polite and otherwise. :)
I'll do my best to answer them.
keep your day job all teams poll is just silly.
Its also seperated into classifications.... Did you not see that.
teamkillbot1022
11-18-2008, 12:08 AM
That is definitely a goal in the future. But my web programming skills are way subpar. Even worse than some of the posters on this thread think my computer ratings are. :D
zing! (on ...yourself?) anyway, thatd be really cool, i just wanted to see how we'd (coppell) do against other schools, not just next week (allen) also, i wanted to see how accurate your predictions were by using it on our past games.
btw, i think youre wrong on your coppell/allen prediction, but then again, youll always piss off 1/2 the people when you make a prediction, and everybody if you say its a tie so, you know, its all good.
CoveMom
11-18-2008, 06:24 AM
The preview page predictions are based on the power ratings on the right half of the ratings page. Whereas the Season Rankings on the left half are intended to rank the teams based on what they've accomplished to date, the Power Ratings on the right half of the page are intended to predict future performance based on past performance.
The preview page predictions are rounded, but to predict a final score, subtract Team B's defense rating from Team A's offense rating and vice versa, taking into account the homefield advantage as necessary. There are problems at the extremes (negative point predictions in extreme mismatches), but that is the method.
Copperas Cove has the #11 power rating in 5A.
oh. thanks. Being a westerner I read the LEFT side of the page and jump off the cliff immediately.....:D
HtownHacker
11-18-2008, 06:47 AM
Nice job. Can you share any details of your ranking algorithm?
trojanbacker
11-18-2008, 07:59 AM
Allen at #1. Yet, the team that gave them their only loss is #2. Interesting reasoning.
Allen at #1. Yet, the team that gave them their only loss is #2. Interesting reasoning.
Well you know how polls are? Only good for bragging and doesn't really mean anything. Probably based on how many points you score on the other teams which I've been suprised 14 is usually enough for Westerburg but I guess we couldn't help it against Boyd, Jesuit & PESH. But is nice to see not all the same and we get at least one.:D
trojanbacker
11-18-2008, 09:03 AM
Not arguing with it; just found it interesting. You guys in Allen deserve to be ranked way ahead of Trinity. Seriously. I've been saying all season we need to be moved down.....
HSFootball#1
11-18-2008, 09:06 AM
Heavens, how many are there and each one has a different #1? Guess if you kept looking you could find one for almost every team to be #1.:rolleyes:
Jayhawker
11-18-2008, 09:11 AM
I think we've all learned something new today.
Don't use computers.
:D
:laugh
oh. thanks. Being a westerner I read the LEFT side of the page and jump off the cliff immediately.....:D
:laugh
Huckleberry
11-18-2008, 09:50 AM
Nice job. Can you share any details of your ranking algorithm?
The theory on the season rankings on the left is very similar to Massey's ratings, if you're familiar with those. Each team is given a game score based on the game result. In the PMV set, the winning team's game score can vary from just above 1/2 all the way to basically 1 depending on the final score. Both margin of victory and points ratio are taken into account, meaning that a 30-10 win is more valued than a 60-40 win. In the NMV set, the game's winner is given a constant game score and the loser, of course, one minus the winner's game score. The actual score of the game is not considered. These game scores are then considered to be the probability that each team would win a game between the two. The ratings are then calculated using a maximum likelihood estimate that maximizes the probability that the season's results would happen as they did based on the ratings. My specific MLE algorithm was coded by the late David Rothman and then coded into Excel by Peter Wolfe, who sent the Excel code to me. The inputs into the algorithm differ significantly from what Rothman used, though.
The power ratings are concerned only with each team's score, winning and losing is completely ignored. Each game is analyzed in the first iteration. A team's offense and defense ratings are adjusted with each game depending on how they fared compared to their opponent's averages. The first step is simply calculating each team's average points per game and points allowed per game, which serve as their initial offense and defense ratings. Then the first iteration takes place.
For example, the first game in the data file is Tascosa over Palo Duro 21-16. For the season, Tascosa scored 25.7 points per game and allowed 31.8 points per game. Palo Duro scored 15.1 points per game and allowed 23.9 points per game. So for this game, Tascosa is assigned an offense score of -2.9 (21-23.9) and a defense score of -0.9 (15.1-16). Similarly, Palo Duro is assigned an offense score of -15.8 (16-31.8) and a defense score of 4.7 (25.7-21). This is done for every team in every game. At the end of the iteration, a team's offense and defense scores are averaged. The team's offense rating is then adjusted to be their average offense score plus the average points scored by all teams in all games. This was a little over 25.27 in this version. So based on the single game above, Tascosa would have a new offense rating of 23.37 (25.27-2.9) and Palo Duro would have a new offense rating of 9.47 (25.27-15.8). The defense ratings for calculations only are obtained by subtracting the defense score from the average points scored across all games. So Tascosa's defense rating for calculations is 26.17 (25.27+0.9) and Palo Duro's defense rating for calculations is 20.57 (25.27-4.7). The final displayed defense rating on the ratings page is simply the average defense score (-0.9 for Tascosa and 4.7 for Palo Duro here).
The iterations are continued until the ratings are stabilized to a suitable degree. For this version that required 909 iterations of the power ratings after which the sum of the absolute values of all ratings changes for all 1,028 teams was less than one-thousandth of a point.
edit - I forgot to provide the links to the college (http://www.barkingcarnival.com/wp-content/uploads/2008/11/allteams.htm) and NFL (http://www.barkingcarnival.com/wp-content/uploads/2008/11/nflrate.htm) versions so people could get a better feel for the ratings.
Allen at #1. Yet, the team that gave them their only loss is #2. Interesting reasoning.
This isn't a subjective poll. If I had a vote in one, I would of course vote Trinity above Allen. I'm not particularly happy that the ratings work out like that at the top, but Trinity is favored by the power ratings in a rematch over Allen, so that part seems reasonable. Trinity is also ahead of Allen in the NMV ratings.
vBulletin® v3.8.7, Copyright ©2000-2012, vBulletin Solutions, Inc.