TheStr1kerz
SILVER
How is everyones season going so far...Lots of movement from what I have heard...Blues lost two keys players....who else?
2023 Season
- Thread starter TheStr1kerz
- Start date
I don't think you're going to get a lot of people that are going to come in here and announce moves to/from their team. It'll be interesting to see how teams shape up as we head into the summer because from my understanding there's a lot of indecision from a lot of people about where their DDs will end up. A lot of the moves that I've heard of involve players on free tuition which also means that they haven't invested into the team and have zero obligation to stay. Regardless of which players have left, Blues Lime were so far ahead of everyone else that maybe this closes the gap slightly but honestly I don't know enough about their roster build-up to say for sure. Tournaments are coming soon - going to be a lot of fun.
TheStr1kerz
SILVER
You guys never want to share any of the drama lol
No need for drama, the soccer storylines are enough
For reference, here are the pre-tournament season CA Top 25. Will be interesting to see how these rankings change over the summer.
You know it's ok to laugh at yourself. I remember U10-13 rankings. After that, the only rankings that matter is your kids GPA, SAT and what college they will attend.
They sure are! Straight from the Soccer Rankings app. If you haven't checked it out, you can find it here. They rank teams as young as 7 years old right now.
The app also has a prediction tab. For some of the bigger tournaments, I've compared the predictor scores to the actual game scores. It's extremely accurate - kind of weird.
JustaSpectator
BRONZE
Love the prediction tab and can't wait for the player rankings to go live! The gentleman has done a great job with that app.
Willy Falcon
SILVER
Don't know a ton about this age group. But here is some drama.
Blues did lose a couple of key pieces I was told. I believe they landed at TFA to play with boys. I could be wrong, but that's what I thought I had heard.
Empire SC fell apart and the coach (Diego) went to Blues. Some of his key 2012 players and a couple of 13's, also went to Blues. I'm
hearing he'll have the 2nd F1 teams for the 12's and 13's. Good coach and respected across the board.
Most the remaining players from Empire landed over at Sporting CA. The Sporting CA team should be vastly improved with several player additions.
RandomSoccerFan
SILVER ELITE
I wouldn't get pre-excited about the player rankings until we see what's really possible. As I understand it, it will be an interface to keep track of your own kid's stats (or as many kids on your squad whose stats you are keeping), and it can be recorded within the app. If enough people do it, and consent to share that individual information, perhaps there could be enough data to create player rankings.
It's a very different situation than game result data, which is available very quickly and very accurately at scale.
It's not weird, it's just math and probability!
What's weird is just that it turns out that youth soccer game data turns out to be very, very predictable - much more so than most would believe by intuition, believing that since anything can happen in any game, there would be much more randomness.Yup, Diego did a great job with the Empire team. That's a great coach for Blues to get. Surprised that the girls ended up at Sporting as opposed to Legends.
irishvirus
BRONZE
Some tournaments...
City SC won the Express Cup in Oceanside...beat DMCV Sharks and Eagles in the process...
Blues Lime won the MSA Tournament...only shock was the final vs LA Surf Kobozart was only 2-0...
Rangers White won the Legends 2012 9v9...beat Sand and Surf in the final...Sand and Surf knocked out Rebels Elite in the semis
City SC won the Express Cup in Oceanside...beat DMCV Sharks and Eagles in the process...
Blues Lime won the MSA Tournament...only shock was the final vs LA Surf Kobozart was only 2-0...
Rangers White won the Legends 2012 9v9...beat Sand and Surf in the final...Sand and Surf knocked out Rebels Elite in the semis
I think at the highest levels the ability to win or lose week after week becomes very tight when playing other high level teams + could go either way. If a top team playes a lower team you know what's going to happen 99% of the time.I wouldn't get pre-excited about the player rankings until we see what's really possible. As I understand it, it will be an interface to keep track of your own kid's stats (or as many kids on your squad whose stats you are keeping), and it can be recorded within the app. If enough people do it, and consent to share that individual information, perhaps there could be enough data to create player rankings.
It's a very different situation than game result data, which is available very quickly and very accurately at scale.
It's not weird, it's just math and probability!
Where things become more variable is at the 2nd and 3rd levels. At these levels a new coach, a single new player, traffic, etc etc etc become more of a factor.
RandomSoccerFan
SILVER ELITE
I know that people believe what you state. It's just that it turns out that the math disproves it. At the highest level of youth soccer (defined here as top 100 teams in nation), the games are tighter, there are less goals scored, and the teams themselves are closer together in skill and likelihood to win. The factors you describe may have a larger effect as you describe. But it's nowhere close to a "could go either way". A higher ranked team in SR (that is in the top 100 nationally) beats a lower ranked team 75% of the time. Not 50%. For comparison, a higher ranked team in SR (including the entire data set) beats a lower ranked team 83.4% of the time. (these numbers were shared for this thread as of beginning of April).
Does that mean it gets every game right? Of course not. 1 out of 4 games it will pick the wrong winner for the top 100. 1 out of 6 games it will pick the wrong winner for all teams. But it's as far from "could go either way" for top teams as it is from being 100% perfect - in fact it's right in between the two. Perhaps Mark could run the percentages again for the top 10 in nation to see if there is a difference just at the top of the top - but looking at the ratings distribution, it's unlikely to be much different. In the age/genders I looked through, there's typically a 3 goal difference from #1 to #100, and a 1.5 goal difference from #10 to #100. (top 100 is about 2% of active teams in lower ages, going up to 5% in the upper ages)
SR's Rankings of the Man City Cup teams. SD Surf are definitely the favorites. According to SR they haven't played a competitive game in over 2 months. Will that lead to some rust? (I don't think so). The newly retooled Blues Oliveros are the favorites in Group B. Going to be interesting to see how things go- a lot of games are toss-ups with the app predicting several ties.
TheStr1kerz
SILVER
Dont be shocked if Surf gets beat..when teams add on new players and havent gelled anything can happen
RandomSoccerFan
SILVER ELITE
OK - Mark has re-run the predictive calcs including only the top 20 teams nationally. For comparison, there were ~750 games in the past two weeks between 2 top 100 teams. There were only 76 games in the past two weeks between 2 top 20 teams, so it's already a pretty small set. But here are those results. As a reminder, predicting the right winner between two rated teams is 83.4% for the entire data set, 75% for exclusively the top 100, and (drum roll please) 73% for exclusively the top 20 teams nationally. By any measure, SR is still highly predictive for even the best of the best.
Code
SILVER ELITE
Now we just need people to go into the Unranked Teams section and sort out all those competitions that didn't get assigned correctly to the Ranked Teams. Good Luck!
RandomSoccerFan
SILVER ELITE
You're not wrong. The more accurately game results are assigned to the proper teams, the more accurate the results/predictions will be for those specific teams - and all of their opponents - and their opponents - and so on. But these calculated accuracy numbers are with the data as it is now, accounting for all of the accuracy/inaccuracy that's baked in. They don't represent what would be the result if everything were perfect - they represent what happens when a team that is rated in the system, plays another team that is rated in the system - how often will the system predict the correct result between those two teams.
And the max guesses it could get right if it picked every game perfectly is 100% - and it's picking almost 84%. The theoretical optimal guess of a future game outcome given perfect data is somewhere between that 84% and 100% - but I still don't believe it would be significantly higher than the current real-world result. 87%? 90%? It's a thought experiment more than something that can be tested and proven - but the main point stands. The better the data, the better the predictions, but there are diminishing returns as it improves up to some maximum - which will be significantly less than being able to predict the future perfectly.
Given the messiness of the incoming data and the data processing shenanigans that are necessary to obtain, deduplicate, normalize, and uncorrupt from the hundreds of different data inputs - it's amazing it works as well as it does. It also showcases the fact that even without perfect data, the assessment of how a team plays turns out to be quite relevant with game data of only 6-10 games in the past 3-6 months. Any more is a bonus. As an example, earlier this year the socal state cup games were reported in a slightly different format, and all teams had duplicate game entries for many of the games - in some cases assigned to an existing team, and in other cases sitting in unranked, unconnected to an existing team. It didn't affect the ratings much at all with the bad/duplicate data, and the ratings didn't change again once the data problem was identified/corrected to remove all of the dupes. If there's a belief that it only gets accurate/useful if you can be sure the last 50 games are 100% accurate - it turns out to be much more forgiving.
If I do come across a data problem while browsing the app - and it's unquestionably clear from the existing available game sources what the reality should be - I do make those corrections. Anyone with Pro account can do the same thing - and should feel free to do so if they choose. But if someone tweaks things carelessly, or is found to be making changes that either aren't in good faith or are unintentionally careless/incorrect, there's a very real risk that Mark will reverse the changes and block the account from editing.
And the max guesses it could get right if it picked every game perfectly is 100% - and it's picking almost 84%. The theoretical optimal guess of a future game outcome given perfect data is somewhere between that 84% and 100% - but I still don't believe it would be significantly higher than the current real-world result. 87%? 90%? It's a thought experiment more than something that can be tested and proven - but the main point stands. The better the data, the better the predictions, but there are diminishing returns as it improves up to some maximum - which will be significantly less than being able to predict the future perfectly.
Given the messiness of the incoming data and the data processing shenanigans that are necessary to obtain, deduplicate, normalize, and uncorrupt from the hundreds of different data inputs - it's amazing it works as well as it does. It also showcases the fact that even without perfect data, the assessment of how a team plays turns out to be quite relevant with game data of only 6-10 games in the past 3-6 months. Any more is a bonus. As an example, earlier this year the socal state cup games were reported in a slightly different format, and all teams had duplicate game entries for many of the games - in some cases assigned to an existing team, and in other cases sitting in unranked, unconnected to an existing team. It didn't affect the ratings much at all with the bad/duplicate data, and the ratings didn't change again once the data problem was identified/corrected to remove all of the dupes. If there's a belief that it only gets accurate/useful if you can be sure the last 50 games are 100% accurate - it turns out to be much more forgiving.
If I do come across a data problem while browsing the app - and it's unquestionably clear from the existing available game sources what the reality should be - I do make those corrections. Anyone with Pro account can do the same thing - and should feel free to do so if they choose. But if someone tweaks things carelessly, or is found to be making changes that either aren't in good faith or are unintentionally careless/incorrect, there's a very real risk that Mark will reverse the changes and block the account from editing.
TheStr1kerz
SILVER
RSL AZ and Legends AZ both tied Surf in regular time...