Youth Soccer Rankings: Disregarding >5 Point Blowouts?

[Woodwork]

Prefacing this by saying no one should care much about these rankings or other youth rankings, for a wide variety of reasons. I would recommend looking at actual head to head results in league play and the actual team play and coaching styles before anything else.

Hypothesis: YSR disregards blowouts over 5 points, or at least limits them to 5 points.

Two examples: (1) https://youthsoccerrankings.us/team.html?teamId=3500229 (Not to rag on this team at all - they likely had a slow start coming out of the pandemic and have steadily improved and solidified. Kudos to them.). And (2) https://youthsoccerrankings.us/team.html?teamId=3478498.

In case (1): If you take the strict average of the goal differentials (e.g., for a 15 goal loss to a 30.83 team, a point score of 15.83), 214.12/9 = 23.79. However, if you limit the losses to 5 point differentials, you average 25.68. That is within .02 of the YSR score.

In case (2): If you take the strict average of their differentials (which goes back about 18 months, a seeming cutoff date for YSR), you get 410.38/11 = 37.30. But by limiting victories to no greater than 5 each, you get 396.38/11 = 36.03, which is within .01 of the YSR score.

The end result, it seems to me, is that YSR doesn't work well in ranking teams that play way out of their league or sandbag, because it cannot derive enough information from blowouts. A team that regularly wins/loses by 5 points is effectively the same as one that wins/loses by 10.

 

[PracticeWYpreach]

Prefacing this by saying no one should care much about these rankings or other youth rankings, for a wide variety of reasons. I would recommend looking at actual head to head results in league play and the actual team play and coaching styles before anything else.

Hypothesis: YSR disregards blowouts over 5 points, or at least limits them to 5 points.

Two examples: (1) https://youthsoccerrankings.us/team.html?teamId=3500229 (Not to rag on this team at all - they likely had a slow start coming out of the pandemic and have steadily improved and solidified. Kudos to them.). And (2) https://youthsoccerrankings.us/team.html?teamId=3478498.

In case (1): If you take the strict average of the goal differentials (e.g., for a 15 goal loss to a 30.83 team, a point score of 15.83), 214.12/9 = 23.79. However, if you limit the losses to 5 point differentials, you average 25.68. That is within .02 of the YSR score.

In case (2): If you take the strict average of their differentials (which goes back about 18 months, a seeming cutoff date for YSR), you get 410.38/11 = 37.30. But by limiting victories to no greater than 5 each, you get 396.38/11 = 36.03, which is within .01 of the YSR score.

The end result, it seems to me, is that YSR doesn't work well in ranking teams that play way out of their league or sandbag, because it cannot derive enough information from blowouts. A team that regularly wins/loses by 5 points is effectively the same as one that wins/loses by 10.

To that point, a flight 1 team that gets blown out in all their games is usually ranked higher than a flight 2 team that wins all their games by a few goals. It probably needs those tournaments where teams who don't normally play each other do.

 

[futboldad1]

YSR is great for the Youngers ages but once all our kids are spread across different leagues it is very inaccurate.........a team in ECRL or flight 1 who blows everyone out is ranked above that SAME club's ECNL team which makes no sense.........YSR rewards blowouts too much....winning close harder games are not recognized as much......

So I think I'm disagreeing with the OP but it is early so I might have misread his words......but I do agree with him that H2H is much better than any rankings

 

[rainbow_unicorn]

YSR is great for the Youngers ages but once all our kids are spread across different leagues it is very inaccurate.........a team in ECRL or flight 1 who blows everyone out is ranked above that SAME club's ECNL team which makes no sense.........YSR rewards blowouts too much....winning close harder games are not recognized as much......

So I think I'm disagreeing with the OP but it is early so I might have misread his words......but I do agree with him that H2H is much better than any rankings

YSR does not "reward" blowouts too much. It's a pretty simple algorithm on the surface which is taking into account as much data as it can and using the game results/goal differentials to generate an expected future game differential between teams. Imposing a goal limit will actually make YSR less accurate. What should be happening is that teams (or players) should be playing in appropriate leagues where they are not blowing out (or getting crushed) on a consistent basis.

 

[futboldad1]

Then how do you account for the same clubs RL teams being ranked higher than their ECNL teams? And teams who are in ECNL that used to thrash teams that are stuck in flight 1 now being ranked below them as the F1 teams win 5-0 every week? That makes no sense at all.......

 

[dad4]

You more want to disregard mismatches than blowouts.

If you expect an even game, but it is 8-0, don’t disregard it. You just got information.

On the other hand, if YSR predicts a 4 point gap, and it is 8-0, then you got no information.

One solution is just toss any game where the winning team was expected to win by 3 or more. It means you disregard the easier games for a team that is hanging out in a low league.

That’s what we all do anyway. If a good team goes trophy hunting in a lower tier tournament, the rest of us don’t think better of them for it. We ignore the mismatches and ask how they do against real opponents.

 

[Desert Hound]

Then how do you account for the same clubs RL teams being ranked higher than their ECNL teams? And teams who are in ECNL that used to thrash teams that are stuck in flight 1 now being ranked below them as the F1 teams win 5-0 every week? That makes no sense at all......

YSR does fine at younger ages. Once teams move off into various leagues (ECNL, ecRL, etc), the rankings start to move all over the place.

Long story short?

At older ages and when teams in different leagues the YRS reliability falls somewhat apart.

 

[Woodwork]

YSR does fine at younger ages. Once teams move off into various leagues (ECNL, ecRL, etc), the rankings start to move all over the place.

Long story short?

At older ages and when teams in different leagues the YRS reliability falls somewhat apart.

YSR depends on head to head competition within a few degrees of Kevin Bacon to compare teams that have never played each other. That does cause some variation but I think it is relatively minor compared to the apparent 5 goal differential limit.

First, that limit causes YSR to have extremely poor predictive value for teams that routinely win/lose by more than 5 goals. Two contrasting examples: (1) https://youthsoccerrankings.us/team.html?teamId=3472201 (Team that is routinely within 5 goals has fairly good predictive outcomes) and (2) https://youthsoccerrankings.us/team.html?teamId=3458985 ("Outlier team" routinely over 5 goal differential has YSR surprised every week, although the scores are consistent with its real average score of 26.7 rather than 29.66).

Second, to the extent YSR depends on a few instances of teams playing outside their league and/or level to draw conclusions about teams across the board, the outlier team will have a disproportionate impact because it is more likely to have head to head games with lower ranked teams when tournaments come around. So an outlier team becomes the measuring stick.

Assuming the purpose of the 5 goal limit is to prevent anomalies from impacting the rankings, I think the 5 goal limit is arbitrary and too low in the context of youth sports. Rather, it would make more sense for YSR to trust its own algorithm, disregarding one result outside of a standard deviation for each team (once there is a pattern, then it would be reincorporated as within the standard deviation).

 

[Woodwork]

You more want to disregard mismatches than blowouts.

If you expect an even game, but it is 8-0, don’t disregard it. You just got information.

On the other hand, if YSR predicts a 4 point gap, and it is 8-0, then you got no information.

One solution is just toss any game where the winning team was expected to win by 3 or more. It means you disregard the easier games for a team that is hanging out in a low league.

That’s what we all do anyway. If a good team goes trophy hunting in a lower tier tournament, the rest of us don’t think better of them for it. We ignore the mismatches and ask how they do against real opponents.

Or perhaps point differentials as a whole are over-rated for predictive value. Teams behave differently once they are ahead and behind. Some teams park the bus, some pile it on, some play subs and rest starters. Some tournaments have benefits for goal differential, some don't. Maybe a better system would ignore GD and base the rankings entirely on head to head W/L (still applying degrees of kevin bacon).

 

[dad4]

Or perhaps point differentials as a whole are over-rated for predictive value. Teams behave differently once they are ahead and behind. Some teams park the bus, some pile it on, some play subs and rest starters. Some tournaments have benefits for goal differential, some don't. Maybe a better system would ignore GD and base the rankings entirely on head to head W/L.

WL- There is good information in the difference between a 3-4 game and a 4-0. Going to simple W/L loses some accuracy. It also makes it even harder to shift information across regions.

Outliers-. I am not sure what an outlier is in this context. Do you mean a team that plays mismatches, or a team that crosses over to compete in a different pool? You used "outlier" for both.

Mismatch seekers are useless as data sources.

Crossover teams that enter a new competitive pool to find close games are incredibly good information sources. The pandemic caused a temporary shortage of these, which hurts.

 

[paytoplay]

The concept of “team” is problematic. In normal years, there’s some consistency within league play, but rosters for tournament play can change the team drastically. Maybe rankings should be a measure of the former, with an alternate ranking including tournaments. And how about weighting the different leagues? ECNL 10, SCDSL 6, Rec 1, etc

 

[Woodwork]

WL- There is good information in the difference between a 3-4 game and a 4-0. Going to simple W/L loses some accuracy. It also makes it even harder to shift information across regions.

Outliers-. I am not sure what an outlier is in this context. Do you mean a team that plays mismatches, or a team that crosses over to compete in a different pool? You used "outlier" for both.

Mismatch seekers are useless as data sources.

Crossover teams that enter a new competitive pool to find close games are incredibly good information sources. The pandemic caused a temporary shortage of these, which hurts.

I was using "outlier team" to describe a team whose results are not accurately predicted by YSR due to the 5 point limit. When I say an outlier team playing in a different pool, I mean the poorly-predicted team playing in a different pool.

 

[MSK357]

I was using "outlier team" to describe a team whose results are not accurately predicted by YSR due to the 5 point limit. When I say an outlier team playing in a different pool, I mean the poorly-predicted team playing in a different pool.

I would like to think when a team is up by 5, they start to take their foot off the pedal. At least moving players around to play other positions. whether its 5-0 or 10-0 its clearly one sided, and shouldn't really matter at that point. I'm sure many teams down by that much are demoralized anyway and aren't even playing to their best capability.

 

[Woodwork]

The concept of “team” is problematic. In normal years, there’s some consistency within league play, but rosters for tournament play can change the team drastically. Maybe rankings should be a measure of the former, with an alternate ranking including tournaments. And how about weighting the different leagues? ECNL 10, SCDSL 6, Rec 1, etc

Tournaments are basically the only way you can have head to head between different leagues, though. E.g., concacaf club tournament pits MLS vs Liga MX and other leagues. It becomes a comparison point for how leagues compare. Otherwise you have endless subjective speculation. I think drastic changes in rosters at tournaments happens but not enough to worry about as a whole. It isn't a systematic problem like the 5-goal differential limit.

 

[dad4]

The concept of “team” is problematic. In normal years, there’s some consistency within league play, but rosters for tournament play can change the team drastically. Maybe rankings should be a measure of the former, with an alternate ranking including tournaments. And how about weighting the different leagues? ECNL 10, SCDSL 6, Rec 1, etc

As long as there are politics in league admissions, weighting leagues hurts accuracy.

Tournament rosters are funky. But tournaments are 100% of your long distance connectivity for lower tier teams. If you throw them out, you get massive drift between different states.

 

[Woodwork]

I would like to think when a team is up by 5, they start to take their foot off the pedal. At least moving players around to play other positions. whether its 5-0 or 10-0 its clearly one sided, and shouldn't really matter at that point. I'm sure many teams down by that much are demoralized anyway and aren't even playing to their best capability.

In my experience, teams winning at the older ages rarely take their foot off the pedal. Think USA vs. Thailand. Perhaps each goal can become worth less in terms of predictive value, gradually. But there is a difference between a team that loses 12-0 and one that loses to the same opponent 5-0. The difference isn't 0.

 

[MSK357]

In my experience, teams winning at the older ages rarely take their foot off the pedal. Think USA vs. Thailand. Perhaps each goal can become worth less in terms of predictive value, gradually. But there is a difference between a team that loses 12-0 and one that loses to the same opponent 5-0. The difference isn't 0.

I think other factors can be at play also. If you have 2 games that day or game the next day, save your starters when you're up 5-0. USA vs Thailand is a little different, they probably could have started the bench and got pretty close to that. And if I remember correctly they did save some players for the next game. So in that sense, that 12-0 could have been worse.

 

[NorCalUSN]

Bringing a different perspective. (BTW I always hated the guy that started a sentence "When I was in..." but here it goes.) You may have seen my post about my daughters playing in Japan. We were at one of those games where my daughter's team was just dominating, it was a like a 20-minute half but the score was already 5-0. I thought for sure the coach would ease off the gas, change playing style or something. But the 2nd half started off like the first soon it was 8-0 and I was feeling pretty ashamed about running up the score on a U10 boys game. I asked my Japanese wife why is he not taking it easy? or something. She asked, found out it was considered more "insulting" to take it easy on the other team then to keep going at regular speed. I didn't necessarily agree but it did open up my perceptions to seeing it differently. Culture or not culture I don't think I would have done the same way but there has to be some balance.

 

[focomoso]

.........a team in ECRL or flight 1 who blows everyone out is ranked above that SAME club's ECNL team which makes no sense.........

I agree there are deficiencies in YSR's model, but this isn't necessarily incorrect. Not all clubs treat their ECNL/ECRL teams as true A/B teams. Often they're made up of existing teams that come in as a unit, so it's not crazy that an RL team might be better than the NL. (And this is especially true for boys as RL is so new.)

 

[RedCard]

One thing that's wrong with the YSRs right now is that they are not including any of the games from the year's SCDSL league season, yet all the games from this year's ECNL league and boys EA league are added.