Why rank youth teams (including 10 and 11 year olds)?
Good strength rankings allow one to construct good league and tournament brackets for all levels. Teams in the PNW play in many different leagues. Tournament and league directors have no way to know the strengths of all the different teams that register to play in their tournament or league. Without good rankings, some teams will be very mis-matched in their league or tournament bracket. The strength ratings apply across ages, so teams also know what level to play at if they play up in a tournament. In addition, many teams find the game to game performance analysis and the match predictor (left navbar) features useful for team development and goal setting.
This must take a lot of time.
Yes it does. That said, many volunteers and many volunteer hours are involved in making youth soccer tournaments and leagues happen. The LP strength ratings hopefully make that work easier and the brackets more even for all. You can help by making sure your team (or teams) use distinct and consistent names when entering tournaments. Club, Team Designation, Age. So something like Arsenal City FC Red B06. Using something like ACFC (no team, no age) or Red B06 means we have to spend a lot time figuring what team you might be.
What goes into these rankings?
Only the home score, away score, home team name and away team name. Nothing else---unless I add an effect of say surface (grass/turf) or a home/away advantage term. Those apply equally to all teams. We attempt to make the database as comprehensive as we can, so every select league and tournament game. The league a team plays in does not enter the calculation---only who they played and the scores. A team in a lower division can have a high ranking by consistently beating weaker teams by wide margins and rarely allowing weaker teams to score on them.
My team played up in age in a tournament or league. How do you take that into account?
Age is not part of the calculations. Age is just used to figure out what table to list the team on. In a way age-groups are no different than league divisions. To estimate strengths, it is only necessary to have enough cross-group (in this case age) games so that no groups are 'isolated'. Isolated means not enough games between one group (age) and other groups (age up and age down). Summer tournaments with mixed age divisions and teams that play-up in age provide these cross-age games.
Why didn't my team go up more when they just beat XYZ team at the top of the standings?
Well, just because XYZ team is stronger than your team doesn't mean they are going to win. Sport wouldn't be much fun if that were the case. Weaker teams do upset stronger teams. The strength ratings are based on a probabilistic model that computes the probability of a win/loss/tie and the goal differential. In order for the strength rating to go up the weaker team must upset stronger teams more often than expected---or score more against them more than expected. One upset will not be enough to increase the strength rating that much. The team must continue to beat stronger teams---or score more than expected---for multiple games.
How can XYZ team be ranked higher than my team if we beat them?
See comment above. Note also that strength ratings are strongly influenced by goal differentials. So if you have more wins but a lower goal differential, your strength rating will be lower. Also note that strength differences of less than 50 are not meaningful. You and the other team are basically the same strength and games could go either way.
The score of our last game doesn't reflect our strengths. We dominated the other team even though the score was 0-0.
One game alone will not determine your strength rating. If your team is truly stronger than the other team, then your team will tend to score more and allow fewer goals. A weaker team might have a very strong defense by employing a 9-1 formation and keeping many stronger teams to 0-0, but they will not score many goals that way and their attack strength will be low and their overall strength correspondingly low.
You are missing some of our games!
Unlike a points chase, you don't get more 'points' by playing more games. You only increase in rating by scoring more than expected and allowing fewer goals than expected. More games helps the algorithm get a better estimate of your team's strength, but missing games in an of itself should not affect your team's estimate. Having matches mis-assigned to your team, however, could definitely change your team's ranking. I say 'could' because a few mis-assigned matches just adds noise. With enough correctly assigned matches, your team's estimates will still be roughly correct.
Can a strength rating be 'gamed' by playing harder or easier opponents?
I don't think so. Winning or losing per se does not enter the calculation, just the goals scored and goals allowed. The value of goals scored or allowed depends entirely on the strength of the other team. So blowing weaker teams out of the water does not get you 'more', since they are weaker. However, not scoring as much as you can against weak teams will make your team look weaker. So in that sense, routinely playing much weaker competition might lower a team's rating a bit if they are 'pulling back the guns' when the score gets over 8-0. One thing I want to add later is to have game strength be a function of the opponent's strength. This will allow me to model coaches rotating on players from the bench when they face weaker opponents. Note also that losing to a stronger team does not hurt you in this algorithm relative to winning against a weaker team. This algorithm computes the relative attack (scoring) to defense (being scored on) rate so what lowers a team's rating is allowing goals or not scoring goals---winning or losing per se does not enter the calculations. That said, there is, unsurprisingly, a very strong correlation between winning games and high goals fielded to goals allowed ratios.
How are teams ranked across leagues that don't meet each other very often?
There are a lot of cross-league matches in the tournaments. More than you might expect just based on your team's experience alone. But basically teams A and B (who never meet) are being ranked against each other based on their performance against similar strength teams---even if they are not playing exactly the same teams.
You are not including team skill. Some of these lower level teams can win but they do it by playing 'bad' soccer.
The objective of this work is to quantify a team's rate of scoring and allowing goals. It is one data-point about a team. Certainly, some teams focus on skills that might not translate at age 10, 11, 12 to high score rates, but the coaches of those teams probably don't want to be playing in leagues and tournaments where they lose almost every game. That doesn't help development. Having a strength rating for the way the team plays now allows a coach to choose a good level for the team.
Is this like BCS or RPI?
It is like BCS in the sense that BCS also uses statistical modeling as part of its ranking. I do not know the specific algorithm used so cannot comment on similarities or differences. However BCS also uses polls and extra 'points' for playing in certain competitions. The algorithm I am using (multivatiate poisson regression) only uses the match scores and who you play, where you play (e.g. what tournament or league) is irrelevant. RPI uses the percent wins and a strength of schedule weighting. It is not a statistical analysis and does not use the goal differentials. The strength of schedule term is also up to interpretation and is not an estimated quantity. The algorithm I am uses implicitly uses the strength of schedule, but it is an estimated quantity. Manipulating one's strength of schedule should not affect the rankings in my algorithm while it has a strong effect on RPI rankings. A team can win all its games, but in my algorithm, it will never be ranked highly unless it wins against stiff opponents. One criticism of the type of algorithm I am using is that it is affected by manipulation of goal differentials (by the teams) that happens when there is significant gambling. I would assume (and hope) there is not gambling occurring regarding youth soccer matches.
Good strength rankings allow one to construct good league and tournament brackets for all levels. Teams in the PNW play in many different leagues. Tournament and league directors have no way to know the strengths of all the different teams that register to play in their tournament or league. Without good rankings, some teams will be very mis-matched in their league or tournament bracket. The strength ratings apply across ages, so teams also know what level to play at if they play up in a tournament. In addition, many teams find the game to game performance analysis and the match predictor (left navbar) features useful for team development and goal setting.
This must take a lot of time.
Yes it does. That said, many volunteers and many volunteer hours are involved in making youth soccer tournaments and leagues happen. The LP strength ratings hopefully make that work easier and the brackets more even for all. You can help by making sure your team (or teams) use distinct and consistent names when entering tournaments. Club, Team Designation, Age. So something like Arsenal City FC Red B06. Using something like ACFC (no team, no age) or Red B06 means we have to spend a lot time figuring what team you might be.
What goes into these rankings?
Only the home score, away score, home team name and away team name. Nothing else---unless I add an effect of say surface (grass/turf) or a home/away advantage term. Those apply equally to all teams. We attempt to make the database as comprehensive as we can, so every select league and tournament game. The league a team plays in does not enter the calculation---only who they played and the scores. A team in a lower division can have a high ranking by consistently beating weaker teams by wide margins and rarely allowing weaker teams to score on them.
My team played up in age in a tournament or league. How do you take that into account?
Age is not part of the calculations. Age is just used to figure out what table to list the team on. In a way age-groups are no different than league divisions. To estimate strengths, it is only necessary to have enough cross-group (in this case age) games so that no groups are 'isolated'. Isolated means not enough games between one group (age) and other groups (age up and age down). Summer tournaments with mixed age divisions and teams that play-up in age provide these cross-age games.
Why didn't my team go up more when they just beat XYZ team at the top of the standings?
Well, just because XYZ team is stronger than your team doesn't mean they are going to win. Sport wouldn't be much fun if that were the case. Weaker teams do upset stronger teams. The strength ratings are based on a probabilistic model that computes the probability of a win/loss/tie and the goal differential. In order for the strength rating to go up the weaker team must upset stronger teams more often than expected---or score more against them more than expected. One upset will not be enough to increase the strength rating that much. The team must continue to beat stronger teams---or score more than expected---for multiple games.
How can XYZ team be ranked higher than my team if we beat them?
See comment above. Note also that strength ratings are strongly influenced by goal differentials. So if you have more wins but a lower goal differential, your strength rating will be lower. Also note that strength differences of less than 50 are not meaningful. You and the other team are basically the same strength and games could go either way.
The score of our last game doesn't reflect our strengths. We dominated the other team even though the score was 0-0.
One game alone will not determine your strength rating. If your team is truly stronger than the other team, then your team will tend to score more and allow fewer goals. A weaker team might have a very strong defense by employing a 9-1 formation and keeping many stronger teams to 0-0, but they will not score many goals that way and their attack strength will be low and their overall strength correspondingly low.
You are missing some of our games!
Unlike a points chase, you don't get more 'points' by playing more games. You only increase in rating by scoring more than expected and allowing fewer goals than expected. More games helps the algorithm get a better estimate of your team's strength, but missing games in an of itself should not affect your team's estimate. Having matches mis-assigned to your team, however, could definitely change your team's ranking. I say 'could' because a few mis-assigned matches just adds noise. With enough correctly assigned matches, your team's estimates will still be roughly correct.
Can a strength rating be 'gamed' by playing harder or easier opponents?
I don't think so. Winning or losing per se does not enter the calculation, just the goals scored and goals allowed. The value of goals scored or allowed depends entirely on the strength of the other team. So blowing weaker teams out of the water does not get you 'more', since they are weaker. However, not scoring as much as you can against weak teams will make your team look weaker. So in that sense, routinely playing much weaker competition might lower a team's rating a bit if they are 'pulling back the guns' when the score gets over 8-0. One thing I want to add later is to have game strength be a function of the opponent's strength. This will allow me to model coaches rotating on players from the bench when they face weaker opponents. Note also that losing to a stronger team does not hurt you in this algorithm relative to winning against a weaker team. This algorithm computes the relative attack (scoring) to defense (being scored on) rate so what lowers a team's rating is allowing goals or not scoring goals---winning or losing per se does not enter the calculations. That said, there is, unsurprisingly, a very strong correlation between winning games and high goals fielded to goals allowed ratios.
How are teams ranked across leagues that don't meet each other very often?
There are a lot of cross-league matches in the tournaments. More than you might expect just based on your team's experience alone. But basically teams A and B (who never meet) are being ranked against each other based on their performance against similar strength teams---even if they are not playing exactly the same teams.
You are not including team skill. Some of these lower level teams can win but they do it by playing 'bad' soccer.
The objective of this work is to quantify a team's rate of scoring and allowing goals. It is one data-point about a team. Certainly, some teams focus on skills that might not translate at age 10, 11, 12 to high score rates, but the coaches of those teams probably don't want to be playing in leagues and tournaments where they lose almost every game. That doesn't help development. Having a strength rating for the way the team plays now allows a coach to choose a good level for the team.
Is this like BCS or RPI?
It is like BCS in the sense that BCS also uses statistical modeling as part of its ranking. I do not know the specific algorithm used so cannot comment on similarities or differences. However BCS also uses polls and extra 'points' for playing in certain competitions. The algorithm I am using (multivatiate poisson regression) only uses the match scores and who you play, where you play (e.g. what tournament or league) is irrelevant. RPI uses the percent wins and a strength of schedule weighting. It is not a statistical analysis and does not use the goal differentials. The strength of schedule term is also up to interpretation and is not an estimated quantity. The algorithm I am uses implicitly uses the strength of schedule, but it is an estimated quantity. Manipulating one's strength of schedule should not affect the rankings in my algorithm while it has a strong effect on RPI rankings. A team can win all its games, but in my algorithm, it will never be ranked highly unless it wins against stiff opponents. One criticism of the type of algorithm I am using is that it is affected by manipulation of goal differentials (by the teams) that happens when there is significant gambling. I would assume (and hope) there is not gambling occurring regarding youth soccer matches.
