02 Ranking by soccer.stat.05

I don't know who has seen the 02 rankings below that soccer.stats.05 was nice enough to do. They are posted in the 04 section and I decided to copy them into the 02 section so we could see them and thank soccer.stat.05 for the work.

Apparently there is some sort of challenge between Gophers and soccer.stat concerning their two methods of rankings and which formula will be a better predictor of game outcome or qualifying or something along those lines. The discussion appears to involve engineering and statistics, so it lost me after the first couple of sentences, but I did see the competition involves beer to the winner! The topic is in the 04 section in the FBR rankings 3-27-12 post.

Again, thanks to Gophers and soccer.stat.05!

Thanks for the 02 data. I ran the numbers (below). I'd like to include the Fall games/tournaments as well so if you can forward that data, I will include in the calculations.


Rankings

Expected Goal Differential

EyesWideOpen wrote:I don't know who has seen the 02 rankings below that soccer.stats.05 was nice enough to do. They are posted in the 04 section and I decided to copy them into the 02 section so we could see them and thank soccer.stat.05 for the work.

Apparently there is some sort of challenge between Gophers and soccer.stat concerning their two methods of rankings and which formula will be a better predictor of game outcome or qualifying or something along those lines. The discussion appears to involve engineering and statistics, so it lost me after the first couple of sentences, but I did see the competition involves beer to the winner! The topic is in the 04 section in the FBR rankings 3-27-12 post.

Again, thanks to Gophers and soccer.stat.05!

Thanks for the 02 data. I ran the numbers (below). I'd like to include the Fall games/tournaments as well so if you can forward that data, I will include in the calculations.



Rankings

Expected Goal Differential

Whats the key concept differences between the methods? Can it be summed up in some basic bullet points?

This list is so dramatically different once you get past the top 10 that I'm suspicious. There are teams ranked high who have lost to teams ranked low. So either he's using a really funky formula, or (more likely) he's including ALL data from the last two years rather than clearing old scores like bwgophers, or he hasn't entered updated scores. Because there are teams in the top 30 that don't belong there, and there are teams in the bottom 20 that don't belong there. With the amount of player movement right now changing team dynamics, I think that makes gopher's list more accurate.

If anything, it is the lack of data. I only have games scores from Jan 1 2012 for the 02s (provided by bwgopher. thanks!). If I get the Fall data, I will incorporate it. That would further refine the ratings with twice as many games.

Data that is older will slant rankings the wrong way though. For teams that have been established for a few seasons maybe not so much, but for teams that have just jumped in the mix or had roster overhauls in the last six months or so there is little histroy to incorporate. There is also a lot of movement right now so how would you weight that?

That is the beauty to the SRS model....there is no weighting. It is all objective. Goal differential and strength of schedule are the only variables. For the 05s, I used the typical soccer year of August to June. I did have concerns about the mobility of players, but surprisingly, it did not seem to cause problems. I would think the mobility would be the same or less at the 02 level especially rolling into LH qualifying.

Just looked at the 02 scores you guys posted.

SRS got 8 correct (including calling a tie)...and 1 incorrect (TFC vs LP Thompson it missed). The macro goal differential for all 9 games was off by 1 goal total for SRS. BTW, a tie is called if the the two teams have rating differential of less than 0.5.

FBR got 7 games correct and 1 incorrect (same game of TFC v LP). The other game that was a tie, I'm not sure how FBR handles calling ties, but the two team were right next to each other on the rankings. To me that would be close enough to predict the tie for that specific game. So the two systems (FBR and SRS) were the same in calling the winner/ties. Not sure how FBR handles predicting the scoring differential for the winning and losing teams. Have to let Gopher comment on that piece.