I thought some of you might be interested in this. I've been working on some statistical projects, and have been using the U-16 Academy results as a proof-of-concept for one of them. This produced a rating system similar to what Jeff Sagarin does for USA Today for pro and college sports. I'm going to try to update it on a fairly regular basis throughout the season. The current ratings take into account the results through February 18. As more results become available on the US Soccer website, I'll update the ratings. I expect things will shake up quite a bit in the early going, as some of these teams have only played 2 or 3 games, making any statistical analysis on them a bit iffy. I'm not really too familiar with these teams, so it will be interesting to see how well these ratings meet the expectations of those of you who follow things more closely. Code: Ranking Team Rating 1 B/W GOTTSCHEE 5.62 2 CROSSFIRE 4.75 3 MICHIGAN WOLVES 3.92 4 U.S. YOUTH NATIONAL TEAM 3.75 5 CASL 2.90 6 PDA 2.72 7 CHICAGO FIRE 2.72 8 BRIDGE FA 2.71 9 IRVINE STRIKERS 2.32 10 SEACOAST UNITED 2.21 11 CLEARWATER CHARGERS SC 2.08 12 REAL COLORADO 2.07 13 PATEADORES 1.96 14 REAL SO CAL 1.87 15 CARMEL UNITED SOCCER CLUB 1.86 16 BALTIMORE BAYS 1.63 17 MIAMI FC KENDALL 1.62 18 ARSENAL FC 1.61 19 VARDAR 1.60 20 LAFC 1.23 21 INTERNATIONALS 1.12 22 DE ANZA FORCE 0.74 23 COLORADO RUSH 0.63 24 SCHULZ ACADEMY 0.53 25 IMG SOCCER ACADEMY 0.37 Let me know what you think, Lee
Johnh00...you are my kind of stats geek...I want to party with you cowboy. I'm sure a lot of people will rip your system and God knows it doesn't mean squat, but it is interesting to look at the numbers. An obvious big difference between the Sagarin for pros/college vs. your index for the Academy teams is the "development" dimension of the Academy. This difference should make your index less accurate relative to predicting the outcome of games (as the Sagarin tries to do). I would love to see this updated as the season progresses, as well as your assessment of its predictive abilities. Adding some supporting stats like records (total, vs. top teams, streghth of schedule as Sagarin does) would be great to see too. Again, I realize NOTHING can be concluded from these rankings, its still fun to look at and could make for some interesting discussions.
I hope for your sake that you don't do stats for a living, as what you are attempting has very little to do with stats. How can you normalize results for teams that have played 3-4 games vs. 10-15 games? Me thinks sampling error. Also, if you use game scores for developing relative strengths, you probably couldn't get a very meaningful distribution, one that can have a high degree of confidence. Now, if you have done a regression analysis that has included a high number of independent variables ( I can think of at least six), you might have something. However, you probably still wouldn't have a good fit, as the data of game scores is probably not linear, ie a 1-0 game may be more meaningful than a 3-2 game. The major point though, is, who cares? Unless of course you are setting up a book making operation. In that case, you should move from CT to NJ.
Yeah, anything that trys to predict how a u-16 team is going to play is a crap-shoot at best. As I am sure most of you have noticed, the consistency isn't there, even at the national team level. However, by taking some of the variables out of the equation (like home-field advantage, quality of schedule, etc), I think this will do a fairly good job of identifying the best teams, which may help in identifying good club programs or teams that might have more then the usual number of good prospects. I could produce a report of the supporting data, but if/when I do, I will have to publish it elsewhere - there's only so much data I can cram into a messageboard post. I sure can. I'll try to put something together this weekend. Lee
Lol...The difference in number of games played is meaningless. I won't get into the details of how it works, because it would probably bore everyone to death, but it's a bit more complicated then adding up goals scored and conceded or looking at won/loss records. When I have used the system to analyse other sports and leagues, it corresponds pretty well to other systems, such as Sagarin's "Predictor" system. There are differences in the results, but the results are similar enough to satisfy me. It also does a pretty good job of predicting results in those leagues, which is probably the most important measure of how well it works. I was using it on the U-16 data, as it gives me a reasonably small but well distributed piece of real-life data to try out some modifications with. For the purposes of this system, it is eliminating the home/away/neutral variable, and the strength of the opponents. Eventually, it may give weightings to the the dates of the games. This is actually one of the reasons for doing this - I want to test out some theories on the dates, primarily to add the dates to the input data, run it through a neural net, and see if a reasonable algorithm can be produced to weight them properly. As far as the scores of games are concerned, this has been pretty well analyzed, and you could use my system to produced over/unders, odds, etc, but as I stated above, that's pretty meaningless when you are dealing with a bunch of 15 year-olds. They tend to be all over the place in terms of the qualities of their performance. Who knows? I know posters here are always discussing the quality of various regions/clubs against each other. I'm just adding this as an "objective" piece to their discussions. It allows comparisons between clubs from different regions who haven't played each other(much). Personally, I'm interested to see if it shows that the one region has a marked advantage over another. If it shows 3 of the top 10 clubs are from Greater Chicago, or southern California, or the DC-Metro area, then maybe we should add another club from that area. One piece of info I have gleaned from it is that there is a big gap between clubs right now. This may be because of the current small sample size, but if those top four clubs continue to have such a huge advantage over the others, that's telling me something very interesting: they've got some real players on those teams. My guess is that the teams will come back to the pack as more results come in, but if they don't, someone needs to look closely at those teams and figure out what's making them so good - perhaps a few(more?) of their players need to be called into the national team setup.
No Chicago Magic or Sockers? I thought they have been doing really well so far? ( not a wind up- I thought both teams have been winning all or most of their games?)
You say the number of games played is meaningless and that can't be the case. You would classify a game as a sample of performance. Until you have reached a certain level of samples, you can't reach a confidence level, ie the samples don't indicate any "pattern" So the number of games should be at least close to equal if you want to draw conclusions. As to the scores, it is important to understand that margin of victory is a very non-linear variable. One should be very careful when using scores. Even Sagarin (who I think is a lightweight) admits this. BTW, the use of a neural net for what you propose is laughable. While a neural net may help with non-linear data anomalies, it certainly wouldn't spit out the algorithms you are seeking. Take Stats 201 before you try again.
And beat another team, as they have played 4 official matches. They also won 2 out of 3 matches in December's Academy Nike Friendlies, which do not count toward the official record. For what it's worth, of course, in tracking the performance of Development Academies whereby development is allegedly the main point.
In layman's terms, how do you explain B/W G at #1 so far ahead when all they have done is beat two 7th place teams? National Team, Fire are both 4-0 and have won by some decent margins. It would seem to me they should be ahead or at least close to B/W G. And then there's Crossfire at 7-0.
3 of your 5 teams come from conferences with losing records so on the surface your model doesn't look that promising. To get a better idea of how good it might be, why don't you compute the results for all 64 teams and then post the results for all teams with at least 4 games. In any event, any model will be pretty flawed. Not only are some of these teams pretty weak, but also the strong ones lack depth. For example there are probably quite a few teams that are noticeable weaker in the 30% of the time the back-up keeper is playing. Furthermore, maybe a few programs were actually dumb enough to believe its about development and are actually playing younger kids instead of playing their oldest ones so they can with their division and trumpet their success in the championship showcase tournament.
One primary goal of the Academy League is to put more emphasis on player development and less on winning (not none, but less). Good to know that the parents won't do their level best to subvert this goal. But hey, rankings are "just for fun" , right? AL coaches don't feel any pressure to do as they have always done (coach to win even if it comes at the expense of player development), right? Okay, feel free to pile on *bemused anticipation*
I like the idea especially as it is all in fun. And the fact that teams from a region will play those from another will make it interesting to see how accurate your method is. One thing I noticed is that you have LAFC below the Patedores and Real So Cal, teams they have beaten.....Do you give a power rating also??? Ie quality wins? keep up the good word, thanks .........
I'm guessing the results aren't what you expected because of a couple of reason. The database had only been updated through February 18th, so the Magic only had two games - a draw and a one goal win. Not good enough to put them at the top, even after adjusting for quality of competition. Sockers were 2-1-1, but their results were offset considerably by who they played. I wouldn't worry about these too much, if I was anyone. Many of these teams have very few results so far, and I would expect things to shake out quite a bit by the end of the season. This early in the season, I expect it will change quite a bit once I update the database with the latest results. Treat this like the College Football top-25 at the beginning of the year. You know, the one where a top-5 team lost to a team that wasn't even in Div 1-A? The only reason I posted this early was I thought it would be interesting to get people's feedback as the rankings changed over time.
I'm with you, right up until the last sentence. The number of games is not meaningless - perhaps I didn't make that part clear. What is meaningless is the number of games in relationship to each other. If have the teams had played 1000 games, and half had played 100 games, I would still have enough data to analyze the data as completely as you ever could given the fact that we are talking about soccer and about 15 or 16 year old kids. It wouldn't matter that half of the teams had played an order of magnitude less games then the other half. Of course margin of victory is non-linear. Pythagoras demonstrates this quite aptly in the reverse when you try to make score predictions based on an anticipated scoring rate. That doesn't mean it's meaningless, though. An analysis taking into account margin of victory has always been more predictive then one that what did not. In this case, it doesn't really matter, as I would never attempt to do predictive analysis of soccer, or of young kids. I would hope that people would just use this as an objective guide of the results that teams have had so far. It's not going to mean a whole heck of a lot in the grand scheme of things, but it's better then just looking at won-loss records or margins of victory. It does what everyone tries to do in their heads when they look analyze teams' performances. It looks at how teams did, the tries to adjust out as many variables as possible to assess their "true" quality. Really, why? I have never had any problem with analyzing data, regardless of whether the results were linear, or non-linear. Maybe I'm misunderstanding your point, but I can't see why that would matter in the slightest. While I quit the actuarial track a bit before the most advanced math, I'm a bit further along then stats 201, thanks.
lol... my sons are waaay to young to be on an acadamy team. By about 10 years. If they were, I certainly wouldn't be driving them all the way to the Bronx to play. There'd be at least 3 other teams closer.
Interesting idea. What factors are involved if not just stats are used. I realized that you are doing this as a challenge but realize some parents take teams ranking very serious. Hope it works out. Any progress on the BU18 stats Thanks
Perhaps they are prodigies. So you could still be sandbagging us. FYI, the comment of mine that you highlighted was not in reference to your ratings, was in reference to somebody else's comment. I understood why your ratings had the output that they did for that team.
There's Freddy Adu joke here somewhere, but I can't quite put my finger on it. In all seriousness, I meant this to be looked at in a "big picture" kind of way, which I guess was a bit naive of me. I've seen how parents act at my son's u-8 games and probably should have figured out based on that that people would be interested in this to see how it rated their son's team. It's only natural for us as parents, I guess. I would just ask everyone to realize this doesn't mean much - it's just an exercise in statistics to try to put some "objective" numbers to the arguments we all make when rating teams. Anyway, here's the u-18's through games of Feb 24th: Code: 1 DC UNITED 5.77 2 BALTIMORE BAYS 5.31 3 LAFC 4.45 4 CHICAGO MAGIC 4.11 5 CONCORDE FIRE 3.61 6 ARSENAL FC 3.46 7 CHICAGO FIRE 3.43 8 U.S. YOUTH NATIONAL TEAM 3.28 9 SAN DIEGO SURF SOCCER 3.21 10 CHIVAS USA 2.56 11 METRO UNITED 2.10 12 SCHULZ ACADEMY 2.04 13 REAL SO CAL 1.99 14 PATEADORES 1.96 15 POTOMAC 1.88 16 PDA 1.83 17 COLORADO RAPIDS 1.69 18 IRVINE STRIKERS 1.59 19 RICHMOND STRIKERS 1.31 20 FC GREATER BOSTON 1.30 21 PA CLASSICS 1.30 22 REAL COLORADO 1.26 23 MICHIGAN WOLVES 1.13 24 SOCKERS FC 1.05 25 FC DELCO 0.88 As with the u-16's, there is still a wider range here then I would expect. I'm guessing that by the end of the year, if this is a regular, fairly homogenous league, the top 25 will range from about .5 to about 2.5 or 3. If the top teams continue to show big #'s, that means there would be ahuge gap in quality, top to bottom. What are people's thoughts on these ratings? Which teams are considered the best at the u-18 level? Lee
Here's the u-16's updated through Feb 24th: Code: 1 B/W GOTTSCHEE 5.62 2 CROSSFIRE 4.54 3 U.S. YOUTH NATIONAL TEAM 3.82 4 MICHIGAN WOLVES 3.82 5 CARMEL UNITED SOCCER CLUB 3.07 6 BRIDGE FA 2.87 7 CASL 2.74 8 PDA 2.69 9 IRVINE STRIKERS 2.18 10 CLEARWATER CHARGERS SC 2.02 11 PATEADORES 1.99 12 ARSENAL FC 1.97 13 REAL COLORADO 1.89 14 VARDAR 1.70 15 MIAMI FC KENDALL 1.54 16 BALTIMORE BAYS 1.51 17 SEACOAST UNITED 1.40 18 REAL SO CAL 1.38 19 CHICAGO MAGIC 1.30 20 LAFC 1.13 21 CHICAGO FIRE 1.03 22 CREW 0.97 23 DE ANZA FORCE 0.56 24 COLORADO RUSH 0.56 25 SCHULZ ACADEMY 0.50
Looks like I should have waited a day, as the results website just updated with this weekend's games. Anyway, here are the U-16's, updated through games of March 2nd: Code: 1 U.S. YOUTH NATIONAL TEAM 5.09 2 CROSSFIRE 4.44 3 CARMEL UNITED SOCCER CLUB 3.47 4 PDA 3.15 5 BRIDGE FA 2.76 6 CASL 2.70 7 PATEADORES 2.34 8 VARDAR 2.22 9 B/W GOTTSCHEE 2.08 10 MICHIGAN WOLVES 2.01 11 IRVINE STRIKERS 1.95 12 CLEARWATER CHARGERS SC 1.79 13 ARSENAL FC 1.77 14 REAL COLORADO 1.73 15 LAFC 1.73 16 REAL SO CAL 1.59 17 MIAMI FC KENDALL 1.41 18 CHICAGO MAGIC 1.33 19 SEACOAST UNITED 1.25 20 SCHULZ ACADEMY 1.05 21 ALBERTSON SOCCER CLUB 0.93 22 CHIVAS USA 0.92 23 CHICAGO FIRE 0.82 24 CREW 0.64 25 IMG SOCCER ACADEMY 0.63 Many teams have still played only a small number of games, so we will probably see big fluctuations in the rankings as these teams get more action. It's good(for me at least) to see the US YNT take over the top spot. This restores a little bit of my faith in the youth national team selection/training process. For parents or fans of particular teams, the more games your team has played, the more weight you should give to your rating. If the team has only played 2 or 3 games, the rating is still very preliminary.