RandomSoccerFan
PREMIER
As of 3/12/25, it looks like it has been resolved on the Girls side, not sure if this is done yet on the Boys side:
Youth Soccer Rankings ?
- Thread starter dawson
- Start date
Random observation about the rankings. I know it’s not important at all but I just realized that in these tournaments with shortened games (60 minutes vs 80 or 90) it probably helps out the lower ranked teams a bit with the lower goal differential even if the stronger teams don’t give bench players more time.
I’m assuming they can’t really normalize per 90 since the public schedules don’t usually show that info. We just “underperformed” and that thought occurred to me.
I’m assuming they can’t really normalize per 90 since the public schedules don’t usually show that info. We just “underperformed” and that thought occurred to me.
RandomSoccerFan
PREMIER
Yes - but it's the same for all teams, who also are subject to the same differing game lengths as well. There's no situation where only "my team" is penalized and other competitive teams they are comparing themselves to aren't.
I think the point was that a short game will, on average, have a lower goal differential.
So, the underdogs — as a class — will “overperform” in short games. And the favorites—as a class — will “underperform” in short games.
RandomSoccerFan
PREMIER
And my point was yes - that's how the math works - but it applies to all teams, who will also be subject to the same parameters. Nobody is getting hurt by this, nobody is getting helped by this.
I think that’s were the quality of tournament also comes to play. My observations are you typically have 1 game per day and full 80-90 minutes for top quality tournaments versus the latter. Top quality tournaments typically bring top quality teams. Play in low quality abbreviated tournaments and you get what you pay for a d the ranking outcome reflect it.
Has anyone captured the stats of tournament games to confirm or deny that hypothesis?
From a coaching and team perspective. Theres different tactics and player types for a 60 vs 90 minute 11v11 game. It also depends on the number of substitutions allowed and field size.
However the net is a team with top players on the field and bench can play at a high level in all lengths of game + sub rules. A team that has a top line and no bench will fall apart at the end of a 90 min game.
Let me try saying it a different way that won’t trigger you.
I was bored sitting in my car between 2 games after my kid’s team won a meaningless game 2-0, so I go look up how the rankings app would have predicted the outcome. It predicted a 3-0 win. The game was 60 minutes instead of 90. Most likely it would have been 3-0 or 4-0 if we had played a full 90. Score one for the algorithm in general, but the result of this accurate prediction is that will be slightly less accurate for future predictions.
So I observed that in these tournaments with shorter games that some teams play in, the most likely outcome is that each teams rating will move towards the average. The stronger team’s ranking will decrease and the weaker team’s ranking will increase, which I’m glad to see you aren’t arguing against since that’s how math works.
I do apologize for using the word ‘help’ as I can see how that might be triggering in this crazy youth soccer landscape. I will also stop referring to my kid’s team as ‘my team’. Enjoy the rest of your preseason.
No one is "hurt" because this just isn't a big deal. But, the OP is correct. The team that tends to do tournaments where they are mostly the favorites will have a slightly negative effect on the YSR ranking. But maybe they don't care about their ranking as much as they enjoy the thrill of winning. So in that way, they are like, "Yay!" and couldn't care less about their ranking. Teams that enter tourneys and maybe stretch a little trying to get in the highest bracket can get a slight boost to their ranking by - for instance - only losing 0-2 to a team that is predicted to beat them by 4 goals simply because the shorter game length didn't allow the other team to smash as many goals into the net.
But again, the benefit is only in terms of their ranking on the app. I've seen this affect certain teams over the years, but the original poster is correct that it's not a huge deal. If some coach wants to use this to get some tiny advantage in the YSR app, I won't complain.
If you are saying that the app works the same for everyone, well yeah. But it definitely can be gamed in this small way for a coach interested in doing so.
RandomSoccerFan
PREMIER
No - it's not a matter of word choice, perspective, or trying to trigger (or not trigger) someone who is trying to explain why your conclusions aren't as accurate as you think they are.
The result of this game prediction, just like the result of any of the predictions, will influence the expected outcome of the next game. All games are weighted the same - whether 60 minute, 90 minute, on grass, turf, with 2 refs or 3, or any other variable that the game has. The results of the team vs. the team on the other side of the pitch are the *only* thing that matters, always. Each is exactly as predictive as any other.
You're right - in that a 60 minute game that you win 2-0, seems like it would be a 3-0 win in a 90 minute game, in a parallel universe. You're not right - in saying that is hurting the rating/ranking of the team compared to what it could be - as the same things are happening to all of their peer teams playing similar schedules.
And yes - playing in top tournaments with more generous playtime against top-rated opponents is a great way to keep a team's rating high - as everything always comes down to how the team performs against the team on the other side of the pitch. Beat better teams - get (and deserve) a better rating.
Now you are arguing against math. Thats an interesting choice.
The rating for that team will go down in this case. If their nearest neighbors don’t play shortened games, then this team’s ranking might go down too.
Does it matter? No. Do I think anyone is actually trying to game it this way? No. Do I think the whole thing is broken and unuseful because the stats are not normalized per 90? No. Do I care that my kids team might now be ranked 49th instead of 47th because this makes their rating drop by two hundredths? No. Did I think that realizing this small detail on why the prediction was slightly off would provoke such a butthurt response? That’s something I was wrong about.
(And yes, we skipped Surf Cup this year because the dates were earlier and we are stuck in a crappier tournament. Surf Cup just has shorter games for the semis and finals. Theoretically those games would be more evenly matched so the effect of the variance would be even more minimal.)
RandomSoccerFan
PREMIER
I hear what you're saying. I understand what you're saying. There's logic in your thinking.
But it's misunderstanding/misrepresenting what a rating means in Soccer Rankings.
Think through a simple example. Walk through this scenario:
Say a team is nothing but a tournament team this season. It is a good team. It plays only 30 minute halves, and does it for 20 straight games this season. It ends up winning every single game, by an exact score of 4-2. Assume its opponents have consistent results as well. At the end of this, this team has a rating exactly 2 goals better than its opponents. If the team has a 48.0, its opponents will all have a 46.0. If you use SR to predict scores between the team and its opponents for future games, it will always predict 4-2.
Here's the thing. Say this team is now going to a tournament with 45 minute halves, or is now starting a season with 45 minute halves. Pick a team to compare it against, have it be a team that is "identical" to the ones it's played, and the predicted score will still be 4-2. There is no concept of game length, or any other adjustable parameter of the scheduled game - just the score for each team. The "real" rating of the team doesn't suddenly become 49.0; of course it remains 48.0, until the team achieves results that give score differences that are different than a 48.0 represents. If the team starts to play games and in fact starts to beat its opponents by 3 goals - their SR rating will shift, and eventually, assuming in this model their opponents stay flat at 48.0 - yes - their rating will bend toward 49.0. But it has nothing to do with the gametime shift; SR has no concept of it. All it will be doing is seeing that now games for the team are resulting in scores that then result in these new rating differences.
Mathematically, SR is looking at games that have already happened, taking all of the scores, and using them to predict future results. It's not making guesses/intuition/logic about future games - it's just fitting a line to all relevant past score differences, weighting them in such a way that will best predict the outcome of the next game, based on years of tweaking the parameters that create that line for the highest/optimal win percentage.
The accuracy of its prediction doesn't get hurt/lessened when game times change. Each game is exactly as predictive as any other game. The raw amount of its rating doesn't get lowered/raised when game times change. It isn't affected in the least. The only thing that matters, and the only thing that ever gets factored in to the rating for a team - is how it performs in games, compared to the rating for the opposing team. Not compared to what a team is expecting to maybe do in the future - but a rating for how the team has already performed.
Now someone with all of this understanding on board, can use it to make their own intuitive predictions, and build on what SR is coming up with, to potentially make even better predictions for a specific game. Say you know that for this same team, your top two goalies are hurt, and the poor kid in goal isn't likely to perform at the same level against a challenging team coming up. SR predicts a 2 goal win based on its ratings - but you think it's instead likely to be a 1 goal win or even worse - and in fact that's what comes to pass. At that point the team's rating will be affected by having this result applied, and it will move slightly to better represent the team's actual performance. Instead, say you know that you've just added 2 phenomenal recruits, and the other team has lost one of theirs. You can look to the comparative SR ratings, see what they are, but you have a hunch that your team is going to perform better than expected and beat the rating. And back to the point of this thread - say that you know the team has mainly played full-length 90-minute games for much of the season, and is now coming up to a tournament that will have 30 minute halves. Their current rating shows them beating opponents by 3 goals. You see that these opponents are also in the tournament. SR's prediction remains unaffected - they are still predicted to win each game by that 3 goal difference. You might instead predict that, "well, the playtime will be lower - so the likely score is going to be 2-0 instead". And you might turn out to be right - and maybe that's in fact what happens. And if it does - that new result will be applied to the team's rating, it will likely shift it slightly, and it will be used for predictions of the very next game.
But none of this intuition or logic goes in to the rating itself ahead of time. It can't. It's not how SR works. The only thing that matters, always, is game results - for game results that have already happened. There is no "actual" SR rating, and a "real-world" SR rating for what it should be, accounting for factors in the future game that may affect the score. A team, any team, achieves its SR rating by playing games and having their game results loaded. Get better results, get better ratings. Get higher ratings than other teams, see higher rankings than other teams.
But it's misunderstanding/misrepresenting what a rating means in Soccer Rankings.
Think through a simple example. Walk through this scenario:
Say a team is nothing but a tournament team this season. It is a good team. It plays only 30 minute halves, and does it for 20 straight games this season. It ends up winning every single game, by an exact score of 4-2. Assume its opponents have consistent results as well. At the end of this, this team has a rating exactly 2 goals better than its opponents. If the team has a 48.0, its opponents will all have a 46.0. If you use SR to predict scores between the team and its opponents for future games, it will always predict 4-2.
Here's the thing. Say this team is now going to a tournament with 45 minute halves, or is now starting a season with 45 minute halves. Pick a team to compare it against, have it be a team that is "identical" to the ones it's played, and the predicted score will still be 4-2. There is no concept of game length, or any other adjustable parameter of the scheduled game - just the score for each team. The "real" rating of the team doesn't suddenly become 49.0; of course it remains 48.0, until the team achieves results that give score differences that are different than a 48.0 represents. If the team starts to play games and in fact starts to beat its opponents by 3 goals - their SR rating will shift, and eventually, assuming in this model their opponents stay flat at 48.0 - yes - their rating will bend toward 49.0. But it has nothing to do with the gametime shift; SR has no concept of it. All it will be doing is seeing that now games for the team are resulting in scores that then result in these new rating differences.
Mathematically, SR is looking at games that have already happened, taking all of the scores, and using them to predict future results. It's not making guesses/intuition/logic about future games - it's just fitting a line to all relevant past score differences, weighting them in such a way that will best predict the outcome of the next game, based on years of tweaking the parameters that create that line for the highest/optimal win percentage.
The accuracy of its prediction doesn't get hurt/lessened when game times change. Each game is exactly as predictive as any other game. The raw amount of its rating doesn't get lowered/raised when game times change. It isn't affected in the least. The only thing that matters, and the only thing that ever gets factored in to the rating for a team - is how it performs in games, compared to the rating for the opposing team. Not compared to what a team is expecting to maybe do in the future - but a rating for how the team has already performed.
Now someone with all of this understanding on board, can use it to make their own intuitive predictions, and build on what SR is coming up with, to potentially make even better predictions for a specific game. Say you know that for this same team, your top two goalies are hurt, and the poor kid in goal isn't likely to perform at the same level against a challenging team coming up. SR predicts a 2 goal win based on its ratings - but you think it's instead likely to be a 1 goal win or even worse - and in fact that's what comes to pass. At that point the team's rating will be affected by having this result applied, and it will move slightly to better represent the team's actual performance. Instead, say you know that you've just added 2 phenomenal recruits, and the other team has lost one of theirs. You can look to the comparative SR ratings, see what they are, but you have a hunch that your team is going to perform better than expected and beat the rating. And back to the point of this thread - say that you know the team has mainly played full-length 90-minute games for much of the season, and is now coming up to a tournament that will have 30 minute halves. Their current rating shows them beating opponents by 3 goals. You see that these opponents are also in the tournament. SR's prediction remains unaffected - they are still predicted to win each game by that 3 goal difference. You might instead predict that, "well, the playtime will be lower - so the likely score is going to be 2-0 instead". And you might turn out to be right - and maybe that's in fact what happens. And if it does - that new result will be applied to the team's rating, it will likely shift it slightly, and it will be used for predictions of the very next game.
But none of this intuition or logic goes in to the rating itself ahead of time. It can't. It's not how SR works. The only thing that matters, always, is game results - for game results that have already happened. There is no "actual" SR rating, and a "real-world" SR rating for what it should be, accounting for factors in the future game that may affect the score. A team, any team, achieves its SR rating by playing games and having their game results loaded. Get better results, get better ratings. Get higher ratings than other teams, see higher rankings than other teams.
Although I agree that normalizing results on game length is a good to have, there are many other factors that affect scores more regularly. Things like best players going on vacation or the coach deciding to play the bench against a weak team have a much bigger effect than game length variations.
Its kind of a toss up.
In theory teams will score more goals in longer games. Assuming they have a full bench. Since the better tournamants attract top teams and tend to have full 90 min games. This is where youd find the highest ranked teams.
To exploit the rankings if you had a decent 1st line and no depth you'd need to look for tournaments with 60 min games, small fields, etc that highly ranked teams participate in. Then play super aggressive for as long as you can hoping to run out the clock before your 1st line collapses.
A better way to address is to develop more players to have a deep bench so you dont need to exploit specific weaknesses for wins.
I think the easiest way to exploit the rankings app is to have two teams with slightly different names. Dump all of your bad results into one team and all the good results onto the other.
Noticed a couple of teams/clubs doing this at Surf Cup.
Teams that would go to the length to do that are the ones without access to MLS next or ECNL. They have to survive somehow.
RandomSoccerFan
PREMIER
Here could be the crux of so many misunderstandings people may have about Soccer Rankings. There are a million different factors that can and will affect the outcome of a soccer game. Many of been listed in this thread - and there are many more that we could come up with.
But once all of those factors are put in the mix, the game is played, and there is a result - that is the only thing that matters from the app's perspective. All it is ever doing is taking previous results, applying some pretty basic math, and using it to predict future results. All of the factors are baked in to how the game result came to be, but all that is ever used to predict the future (and apply a rating), is exclusively the final result.
Maybe - but one also has to account for the following. In youth soccer, as the teams get better, especially towards the top 100 nationally, there are significantly *less* goals scored, compared to all games of all teams. It's also why the app is less predictive for that set of top teams compared to all teams - because less goals means that normal randomness will have more of an effect on the game outcome.
RandomSoccerFan
PREMIER
Yes - this is accurate as written. If a team can do exactly this - and the score of the games was kept close - their team rating would reflect the game results. The prediction for their future games will reliably show how they play against these same opponents in these tournaments (and anywhere else). IMO it's not a cheat code or an exploit - it's just a recognition that their game results --> their rating.
One could look at the situation, and believe "well if they played full-length games, on large fields, against these same teams, they'd probably do worse". Even if that were true (and it certainly could be), it's just speculation - until those new games are actually played.
Trust me, it's not a misunderstanding. I understand most of the variables involved, I just hadn't considered this one until now. Maybe the other team was missing as many players as we were, or maybe they were missing more. Maybe each team lost/gained a few players since last season and it's a completely different team that hasn't been reflected yet because they haven't played any games yet. Maybe they were also coming out of hell week with heavy legs. Especially since it's preseason, maybe each team was trying different formations and positions. I get it.
That's the same thing I said, no? Why are you arguing? Most teams play 80+% games at full length which weights the data for full length games. If it was possible to normalize per 90, then the ratings wouldn't shift, but we both agree there are so many other variables that it's not worth it and directionally it evens out over time.