There is another great part of the app. You can click on any team on it and see the game results for that team. Nationals proved to me that the algorithm isn't perfect and I think there should be more weighting to the "difficulty of schedule". I say that because of the dominance of SoCal. We play great teams and a lot of other clubs play very sub-par teams, beat them 9-0, and the app gives them way too much credit. Really good teams can absolutely score goals at will against really weak teams and you learn nothing playing those teams if the score is 4-0 or 14-0.
Club Team Tiers
- Thread starter Rollonyoubears
- Start date
The other side to this coin is good teams that generally play terrible teams in league get penalized.
What will happen is when good teams from not so good leagues play against poor teams from good leagues they'll destroy them.
I would suggest weighting teams ranking by geography not league. CA TX then everyone else.
RandomSoccerFan
GOLD
Both of these things can't be true at the same time. Good teams playing poor competition either get too much credit (and have inflated ratings), or not enough credit (and have deflated ratings). It doesn't mean either of you are seeing things incorrectly, and it might be a a direct interpretation of actual games that you're seeing.
But I think it might be helpful to separate the predicted score / difference between two teams (Did we win by 3 goals), and the win/loss result (Did we win). The ratings are optimized so a higher rated team will beat a lower rated team. If there are circumstances where higher (better) rated teams are losing to lower (worse) rated teams, the ratings of both of the teams will adjust in the expected direction. If it were happening enough, the percentage of successfully predicted wins would be expected to be noticeably lower than it is.
I think we'd all agree that a team that can beat another by 5 can also beat them by 10, and the inherent differences between the teams don't matter much whether the score differs by 5 points, 10 points, or 20 points.
The problem with what you're describing is after a team gets up by 5 goals they tend pull back from scoring.
RandomSoccerFan
GOLD
Right - but that's the point. Regardless of why that difference isn't as large as it might be in a perfect world (A better team might theoretically keep on scoring to go up by 10 if there weren't a stigma), it doesn't matter. The results are the results, the wins are wins, and the scores are recorded.
I don't agree with that.
If you're a better team and you can score 10+ goals on an opponent at around 5-6 unanswered goals coaches have a decision to make.
Go all in and keep your starters in to score as many goals as possible or pull back and let players try different positions to keep the score even.
RandomSoccerFan
GOLD
We're still not communicating. I am not disagreeing with your thought process here. Of course coaches will have all sorts of legitimate reasons for either running up the score or not, or defending at all costs or not. Any and all of these decisions may be made for a whole variety of legitimate reasons.
But at the end of all of that - there is a score at the end of the game. And whether it is a 4 point difference, an 8 point difference, or a 20 point difference - the difference between those three outcomes doesn't provide much data about exactly how different in quality the two teams are.
But aggregate all of those results, figure out which team wins, and the rating is what it is. And it's at this point where everyone thinks they are getting screwed. "I should be ranked higher because I could have won by more" "They should be ranked lower because getting extra goals over a terrible opponent shouldn't matter much". As long as everyone is whining, it's probably working OK.
But at the end of all of that - there is a score at the end of the game. And whether it is a 4 point difference, an 8 point difference, or a 20 point difference - the difference between those three outcomes doesn't provide much data about exactly how different in quality the two teams are.
But aggregate all of those results, figure out which team wins, and the rating is what it is. And it's at this point where everyone thinks they are getting screwed. "I should be ranked higher because I could have won by more" "They should be ranked lower because getting extra goals over a terrible opponent shouldn't matter much". As long as everyone is whining, it's probably working OK.
Where do you guys play that teams pull back when the score is lopsided? I haven't encountered that timeline. Most coaches we've had, even the softer ones, take that opportunity to work on specific tactics, but not to stop scoring.
Years ago, my son's team was up by a ton at half. The coach told the boys they had to make 15 passes before they could go to goal. Guess what happens when you make 15 passes before attacking... you score more goals. It got so bad he told them not to shoot at all, but that was even worse. They'd get right on the doorstep with an open net and turn around and pass back. It made the other team feel even worse. So he put the defenders up top and they really ran up the score because it was their only chance.
The coach told me later that he was never going to pull them back again. "Just play our game."
Years ago, my son's team was up by a ton at half. The coach told the boys they had to make 15 passes before they could go to goal. Guess what happens when you make 15 passes before attacking... you score more goals. It got so bad he told them not to shoot at all, but that was even worse. They'd get right on the doorstep with an open net and turn around and pass back. It made the other team feel even worse. So he put the defenders up top and they really ran up the score because it was their only chance.
The coach told me later that he was never going to pull them back again. "Just play our game."
I don’t really care much about any team “getting screwed” in the rankings. The rankings are certainly good enough to help flight a tournament, which is nice.
That doesn’t mean the algorithm can’t be better. Lowering the weight given to expected mismatches would help.
Why? Teams do weird stuff when they’re expecting a blowout. The weaker team can give up before the first whistle. The stronger team might start the entire bench, or loan their starters to an older team with injuries.
In effect, neither team really cares whether the final score is 8-0 or 14-0. So why should your algorithm?
RandomSoccerFan
GOLD
First, not my algorithm. I'm just using the app like you. Second, you (or I, or likely anyone else other than the developers), have no idea whatsoever about what the algorithms are already doing to discount the weights given to different amounts of score differences. Of course it is going to put less weight as blowouts get worse. Of course at some point the amount of goal differences don't matter. Of course I imagine that this was thought of years ago when it was being developed, and continued to be massaged over the years and now with the app. It's tweaks to all of these things, specifically weighting the results of real matches vs. predictions to make sure that the predictions, and therefore the ratings, are as accurate as they can be given the incoming data.
If folks are curious about the details - the support email is right there in the app. They are incredibly responsive, and I imagine if there's anything that they can share, they would - just ask!
If folks are curious about the details - the support email is right there in the app. They are incredibly responsive, and I imagine if there's anything that they can share, they would - just ask!
If they keep underperforming, I'm pretty sure that's exactly what happens.
They keep underperforming just by enough to not get the dreaded red score line... but also never good enough to get the coveted green score line when they win
you know if you know
We were supposed to lose today's game 4-1.
We were up 1-0 at the half.
Parents were optimistic on the sidelines.
But I trust the algorithm.
We lost 4-1. LOL.
Guys guys, I think he was joking (right right??)
Let's all get along.
I get on here to feel better and banter with strangers so I don't have to talk to other parents on my kid's team
This is absolutely true.
I've seen some teams specifically try to blow out teams as large as possible even against teams not at their level. They consider it a pride to get a 15-0 score line out there. I'm not judging. I think there are pros and cons.
But I know my kid's team even if we did randomly meet a team in league that's "scoreable on" at will... the coach instructs the kids to stop scoring when it gets to 7-0 or something... they COULD hang 15 on some of those games... but they don't and so it doesn't get reflected as such in Gotsport and consequently on the app...
LOL this did happen a couple of years ago in a tournament where they clearly promoted a flight 2 team up to the top bracket because they needed to fill the space....
It was 10-0 at the half and so coach instructed everyone has to touch the ball before they could score... so everyone would take turns and check in to the ball and get in a touch and then they'll go ahead and score... haha didn't really help...
I think this is the better approach... just play bench players more but play their regular game... let them do their own thing...
RandomSoccerFan
GOLD
Right - but your team is not a rarity in behaving this way. Many teams do the same. Most teams (IMO) do the same. And yes, there are teams that seem to be going for a high score like an arcade game. I don't believe the differences in outcomes, from beating a team soundly, to beating them mercilessly, make much of a difference in the team's rating, as at some point the weighting of blowouts continues to go down. FOMO because there aren't 15-0 score lines isn't really needed.
In the youngers, it was more common to play in a league or participate in a tournament that had "mercy rules", and would penalize a team if they won by more than 7 goals. Coach had a code word he'd scream at the team, basically forbidding them from scoring another goal - as if they did so, they'd actually lose a point in the tournament scoring. That is more rare once they get to the larger field and 11v11, most tournaments we've seen no longer include this for those age groups and up.
Rollonyoubears
BRONZE
Pretty much! Haha. But we all love it.
I agree and at Surf, our coaches are trying to balance winning and playing time. Rosters are too big in my humble opinion and the coaches know that and will jump at every opportunity to get his (or her) non starters in the game. Scoring for most players are difficult and your starters almost always include your scorers. When your scorers are taken out of the game, scoring goes down. If you don’t take out those payers that score, the score runs up. There is the problem with rankings. Some teams have essentially enough players to field two teams, and the starting 11 might be in the top 10 and the next 11 might be in the top 50. So, rankings would have to know who is playing, etc etc etc to be really accurate. I love soccer rankings with its flaws because any ranking system has flaws. ECNL rankings have much bigger flaws.
RandomSoccerFan
GOLD
I don't believe this is the case, but I imagine that we're posing different questions. The ratings/rankings/predictions attempt to be as accurate as feasible, given the incoming data of scores of the games. If it could infer more accurate outcomes given the incoming scores - it would - that's what all of the improvement over the years is focused on. The predictivity results of it right now, are exactly what they are - no better, or worse. Throwing out the model because conditions aren't exactly average for that game, isn't necessary or appropriate - the predictivity results already include all of that variability - because it has no idea it's there in the first place. The predicted results are what the team is predicted to do given all prior results, in all conditions.
Now - if the model had additional incoming data, like which specific roster was on the field, the weather, injuries, etc., I don't doubt that it would up the predictivity a bit compared to today. The question is how much better, as eventually there would be diminishing returns no matter how much data it has, and there will always be variability in results. Would it go from 83% to something like 90% or even higher? Who knows - but there is definitely an upper limit, and it's hard to say how far (or close) it is from that right now.