maestroFRSM
SILVER
100% of our team doesn’t care
What’s the point?
We know who the best teams are
Youth Soccer Rankings ?
- Thread starter dawson
- Start date
What’s the point?
We know who the best teams are
The point is to make it easier to set up good scrimmages.
It's hard to do that if you don't know the scores from previous games.
rainbow_unicorn
PREMIER
Cmon, let's be real. It's 100% for parent pride.
Maybe for some but it's really a process to help parents navigate the truth to youth soccer. A lot of coaches and clubs tends to sell the wins but they might not exist or they exist based on very weak opponents. It's not the ONLY information parents should use to access whether a club is appropriate for their kids, but it's one tool in the toolbox because you should go watch teams play several games before making decisions.
The more truthful information we have available to parents, the better.
Got soccer only looks at got soccer.
the app pulls data from all sources including TGS, and various other leagues, demosphere and whatever app surf cup uses and more.
Got soccer also gives weight to certain tournaments which can boost other teams got soccer points where as the algorithm used in the soccer rankings app takes a variety of data based off of match results to assign each team a score that is used as a predictive value that allows for a goal differential to be forecasted should 2 teams from any age group and gender were to play against each other.
gotsoccer ranks teams by whether they go to gotsoccer events, and how well they do when there. It is not even remotely accurate.
SR adds up all results, weighted by strength of opponent. It is mostly accurate. One limit is that, if you’re looking at the best team in a weak region, it will overestimate their strength.
RandomSoccerFan
PREMIER
I had the exact same question for my wife quite awhile back when she shared info with me about YSR, which she had initially heard about from another parent, whose kid was starting to travel the country for MLS N. What could it be possibly doing with the game data that is heavily concentrated already on GotSoccer/GotSport, and calculating things differently to show completely different - and much more accurate - rankings and ratings. I was highly skeptical.
With some curiosity, I played with it a bit, and realized that it was in fact very accurate for all of the teams we are most familiar with, and predicted game results with uncanny accuracy. Over time I spent more free time playing with the team data itself, and merging/fixing teams in the area to improve the data quality over all. When it went down for good as a website - it did seem like something significant was lost, but Mark and team rebuilt it as an app and it has since far surpassed what it was in prior generations. If interested in exactly how it works, and how its accuracy can be measured and tracked over time - here's a more detailed post from earlier in this thread.
But to jump to the end - SR is now expected to pick the right winner for a game between two rated teams, with 82% accuracy. GotSport rankings are at 55% - 60%, only slightly better than a coinflip. For the top 100 teams, GotSport does a little better, in the 70% range (probably due to the tournament bonus points system), but it's still miles behind SR. For context, if any of these systems were at 50%, they shouldn't call themselves rating/ranking systems, and should rebrand themselves as random number generators.
Yes, some people certainly do believe that - but it is quite hard to prove it one way or another. On the one side, you have this, where game data analysis on the system itself, limited only to games between teams in different states - shows that it is still essentially as accurate as games between teams in the same state. And on the other side, you now have the app providing data on the relative strength of the schedule teams have played over the past year, and there are examples of some very highly rated teams, who have played some relatively weaker schedules when arriving at their current rating. At some point - until the teams are on the same pitch on opposing sides, there are fundamental limits to how accurate future predictions can be. But these are pretty darned good. One thing to also keep in mind is that even the back-end algorithms aren't necessarily static and unchanging - they are tweaked over time to continue to optimize that predictability number. All of the weighting for how long game data should count for, how much does it decline over time, how much to weight goal differences, how to discount larger goal differences, and probably quite a few more parameters - those are the levers that can be tweaked to optimize the results.
I'd say 80% parent pride and 10% for setting up good scrimmages and 10% for determining if your next opponent is someone you're going to need to pass the ball around 10 times before shooting with your weak foot only, or whether you should prep your kid mentally for the beat down of his life. It can also be used by parents to see if the coach that is recruiting their kid is selling snake oil. There's no doubt it serves useful functions in addition to just being a massive ego stroke for parents. More so for youngers versus olders, where most everyone already knows everyone.
RandomSoccerFan
PREMIER
Forgot to add one point, that may sometimes be glossed over when discussing these differences. If there is a bad data for a team (wrong scores entered, team data incorrect, scores connected to wrong team, etc.), GotSport charges $25 customer service fee per instance to change a single piece of data. So in practice - it's a FU charge to go away, and so nothing is ever changed or corrected unless GS does it themselves through some other process. In just our own club, there are a number of problems with team data in GS with duplicate team entities actually being the same team, and that will likely stay that way until the end of time. As a contrast, SR data is pulled in electronically, but its accuracy is then crowd-sourced, and anyone with a sub ($10/year) can correct anything they see that is wrong in the app, whether their own team, another team in their club, or something else in a bracket they are familiar with, whatever.
maestroFRSM
SILVER
if MLS Next doesn’t publicly report, then SR isn’t useful (for that tier of teams), because no one knows all those scores. My original point is that self reporting isn’t reliable enough.
This.
And I’ve found it more helpful for tournaments like man city to look into teams from out of state.
some clubs have different teams (under the same team name) play different opposing teams from the same league.
For example, 2010 A team will play the better league 2010 teams and the B team (usually 2011s) will play weaker 2010 teams. Games played under same 2010 team name
that messes up the rankings
RandomSoccerFan
PREMIER
It's as useful as the data. If there is no data - eventually it won't be useful for teams that have no data. I think there's a very good chance that the scores will become available soon, and I'm also somewhat more optimistic that while self-reporting is clearly sub-optimal, if it ever has a chance of being workable - it would be at the highest levels such as MLS, rather than if it were happening to a much more casual league. We'll see soon enough.
Yes - that's my point. Any team who wants their game history and therefore their rating to be as close to reality as possible, has the opportunity to confirm it and clean it up if things are missing or miscategorized. The software does a good job of assigning the games to the appropriate team entity, and where it can't, it's usually not that long before someone comes along to merge/fix as needed if it is sitting in unrated.
Your example is correct - if there are teams that are loosey-goosey with their names and their rosters, and play a bunch of different teams as the exact same name with no differentiation - the ratings in SR, GS, or anywhere else are going to be tied to the game history of that combined team. It does seem like the more common data complication is instead that teams use random and changing names as they enter various tournaments, appending the coach's name, or abbreviating the club name differently, or stating "pool team", etc. In that case those team entities do stay on their own and may never get a rating, unless that version of the team does have enough matches in their history alone to be rated (typically 6-8 in previous 6 months).
At this amount of data (millions of games, thousands and thousands of clubs, teams, etc.), it's a given that there will be data quality concerns. But the stats of how accurate it is *right now* already are taking that into account. It's reporting the reality of one ranked team entity vs. another ranked team entity, as stored in the database today. If someone could snap their fingers and immediately all game data was 100% correct and assigned to the correct team entity that fully represented the actual players on the field that day - it's possible that the prediction accuracy, and therefore rating accuracy, would go up somewhat. But it's debatable how much better it could actually get from where it is today, it's a classic case of diminishing returns at some point. Any team that wants their rating to be accurate as possible has the tools to do it easily. Any team that doesn't care isn't significantly impacted, and the aggregate ratings/rankings aren't affected much for others.
maestroFRSM
SILVER
“Yes - that's my point. Any team who wants their game history and therefore their rating to be as close to reality as possible, has the opportunity to confirm it and clean it up if things are missing or miscategorized.
Any team that wants their rating to be accurate as possible has the tools to do it easily.”
how would you “ fix” things for the situation I gave?
Any team that wants their rating to be accurate as possible has the tools to do it easily.”
how would you “ fix” things for the situation I gave?
RandomSoccerFan
PREMIER
Have you played with the app directly? If you don't have a sub, you can't get to the data sources to see how it all fits together and can be individually corrected. It's $10 for a year, and there's a 30-day free trial, so if interested - you can see if it's useful for a few weeks before cancelling without any cost. But here are the app screens that point to team data. Let's take this example team:
That team has a couple of different events/tournaments. Scrolling down, here are all of their games:
If you click over to Sources, here's the actual data sources that they were coming from:
You can see how the names aren't the same. The first two were probably automatically matched, while the 3rd was potentially linked together manually by someone, as they had used a slightly different name for that 1 tournament.
Here's that specific tournament:
If it turns out that I was familiar with the team - and knew that it actually wasn't them - I could delete that data source by hitting the trash bin next to the invalid source:
and then hit Done in the top right, and the team data would then be corrected. That's really it. You generally can't enter your own data in - the only option is manipulating the data that SR has pulled in via these game sources, and adding/deleting them to the team entity as appropriate. If I knew there was another tournament out there with the team named somewhat different, I could hit Add Sources, search for something close to that other team name, and if it was the right data I could add it to this team entity and then save it the same way.
The workaround being proposed for MLS is interesting, and I do wonder about how the UI will accommodate creating some new data entries, and how that will be monitored over time. But I have no doubt that in this same manner - if there is bad or even malicious data, it will be able to be removed/changed by the next person that comes across it.
Now your specific example does have its conundrums. If a team really is named "The Super Bunny Puppies Academy Premier Elite", and they use that exact same name in everything they enter - but in some games it's made up the team A starting powerhouses, but in some games it's entirely made of the scrubs that are most worried about splinters from the bench - it could be hard to separate them. The first fix is to have the club registrar and/or the team manager realize that using the exact same name for essentially different teams is a bad idea for a host of reasons, two of them having a consistent game history and having objective measures about how good the team actually is - all helpful when deciding what brackets to enter for each event, and even how to place the team next season. If the team does start naming more consistently interpretable as they enter events/tournaments, and now there are "The Super Bunny Puppies Academy Premier Elite Red" and "The Super Bunny Puppies Academy Premier Elite Blue", once their first data starts showing up in SR - you can then pull in the correct events from the prior combined name into the appropriate new team entity. For most individual tournament info, this would work fine. Though it may be tedious to figure out which is which, it's doable, and it's the very foreseeable result of using the same team name to represent multiple things at once. The sticking point that isn't resolvable, is if all of some particular data is in a single source, and that single source is actually representing multiple versions of the team, with the same name, in the same source. For example if this was in league - which typically only has a single data source entry for all games for the season (but still often split by fall/winter/spring/etc.), there's no way to say let's apply games 1,3,5 in league to the Red team and games 2,4,6 to the Blue team.
The good news, is that if a team has been sharing an identical team name across non identical teams, and at some point decides that they don't want to do that anymore - starting with a new more intelligible team name for GS/SR whatever, will still result in a rating showing up for the new team name in only 6 or 7 games. One of our teams had to do this just this year, when there were some name swaps in the club from X to Y. The Y team had no rating - but within 4 weeks - they not only had a rating - but it was identical within a tenth of a point to their rating from all of the prior game data in X. It's a surprisingly small amount of rated games that can be relied upon for the rating to show up.
If you do have a specific example team that you're aware of, and wondering if it is possible to clean it up or not - PM me and if you want I'll send you screenshots of exactly what the data looks like as is, and potentially recommend how I'd fix it if I knew more about the team itself.
That team has a couple of different events/tournaments. Scrolling down, here are all of their games:
If you click over to Sources, here's the actual data sources that they were coming from:
You can see how the names aren't the same. The first two were probably automatically matched, while the 3rd was potentially linked together manually by someone, as they had used a slightly different name for that 1 tournament.
Here's that specific tournament:
If it turns out that I was familiar with the team - and knew that it actually wasn't them - I could delete that data source by hitting the trash bin next to the invalid source:
and then hit Done in the top right, and the team data would then be corrected. That's really it. You generally can't enter your own data in - the only option is manipulating the data that SR has pulled in via these game sources, and adding/deleting them to the team entity as appropriate. If I knew there was another tournament out there with the team named somewhat different, I could hit Add Sources, search for something close to that other team name, and if it was the right data I could add it to this team entity and then save it the same way.
The workaround being proposed for MLS is interesting, and I do wonder about how the UI will accommodate creating some new data entries, and how that will be monitored over time. But I have no doubt that in this same manner - if there is bad or even malicious data, it will be able to be removed/changed by the next person that comes across it.
Now your specific example does have its conundrums. If a team really is named "The Super Bunny Puppies Academy Premier Elite", and they use that exact same name in everything they enter - but in some games it's made up the team A starting powerhouses, but in some games it's entirely made of the scrubs that are most worried about splinters from the bench - it could be hard to separate them. The first fix is to have the club registrar and/or the team manager realize that using the exact same name for essentially different teams is a bad idea for a host of reasons, two of them having a consistent game history and having objective measures about how good the team actually is - all helpful when deciding what brackets to enter for each event, and even how to place the team next season. If the team does start naming more consistently interpretable as they enter events/tournaments, and now there are "The Super Bunny Puppies Academy Premier Elite Red" and "The Super Bunny Puppies Academy Premier Elite Blue", once their first data starts showing up in SR - you can then pull in the correct events from the prior combined name into the appropriate new team entity. For most individual tournament info, this would work fine. Though it may be tedious to figure out which is which, it's doable, and it's the very foreseeable result of using the same team name to represent multiple things at once. The sticking point that isn't resolvable, is if all of some particular data is in a single source, and that single source is actually representing multiple versions of the team, with the same name, in the same source. For example if this was in league - which typically only has a single data source entry for all games for the season (but still often split by fall/winter/spring/etc.), there's no way to say let's apply games 1,3,5 in league to the Red team and games 2,4,6 to the Blue team.
The good news, is that if a team has been sharing an identical team name across non identical teams, and at some point decides that they don't want to do that anymore - starting with a new more intelligible team name for GS/SR whatever, will still result in a rating showing up for the new team name in only 6 or 7 games. One of our teams had to do this just this year, when there were some name swaps in the club from X to Y. The Y team had no rating - but within 4 weeks - they not only had a rating - but it was identical within a tenth of a point to their rating from all of the prior game data in X. It's a surprisingly small amount of rated games that can be relied upon for the rating to show up.
If you do have a specific example team that you're aware of, and wondering if it is possible to clean it up or not - PM me and if you want I'll send you screenshots of exactly what the data looks like as is, and potentially recommend how I'd fix it if I knew more about the team itself.
maestroFRSM
SILVER
yes, because you can’t parse out individual games, and because a few of the teams will interchange teams under the same name, SR can be inaccurate in these situations
Would be nice if we could delete certain games we know don’t reflect the overall quality of the team
RandomSoccerFan
PREMIER
I'm not sure it would be nice at all, it would be an open invitation for people with an unquestionable bias to decide which games were relevant and which weren't. I wonder how they might choose that...
With teams that are being silly with their naming and how they register, the possible effects are real - but they may be less significant than you'd think. Pick a specific team that you're thinking of that has these issues - not a hypothetical, an actual, real team that you think is at risk of an incorrect rating due to it. You can then look in SR for that particular team entity, and review the last 15 - 20 games, and see how accurate (or inaccurate), the team's rating has been to predict their performance. It may be surprising how spot-on the results are. If you don't have a trial and can't do it yourself - DM me and I'll send you screenshots.
RandomSoccerFan
PREMIER
As of today, it looks like the 2023-2024 results from MLS Next have been pulled in to SR. Some of the teams need to be merged, but by and large all is well. Not sure if this is temporary or a more permanent fix, but all is moving in the right direction.
Modular 11 had the results up for a couple of days and then pulled them down. I suspect these updates will remain static until the the results are posted again...
maestroFRSM
SILVER
Some of those results are inaccurate
maestroFRSM
SILVER
I have picked out a real team and SR was inaccurate
Again because 2 different rosters competed under the same name in the same league