Girls 02 Rankings

[Unregistered]

True but we're stuck here and like to think our players are "all that" and that they'll end up w/full rides to D1 schools and start on the national team. Hey, they go to US Training Center camps, after all! It's just all the other kids on their teams who are dragging them down....

[Unregistered]

These rankings are pretty cool.

Don't put too much weight on them though.

I know of several anomalies that are pretty dramatic on other age groups. That is probably more related to shuffling of players among teams.

[Unregistered]

These rankings are pretty cool.

Don't put too much weight on them though.

I know of several anomalies that are pretty dramatic on other age groups. That is probably more related to shuffling of players among teams.

Toss out summer tournament records, every tournament you've got players out on vacations and different rosters wiith coaches hopefully focused on moving players around preparing for league.

[Unregistered]

Toss out summer tournament records, every tournament you've got players out on vacations and different rosters wiith coaches hopefully focused on moving players around preparing for league.

Every game counts. Teams with depth can handle absences better and are therefore stronger teams.

[Unregistered]

Every game counts. Teams with depth can handle absences better and are therefore stronger teams.

The ranking on Last Planet are actually surprisingly good at predicting out comes in other age groups.

[Unregistered]

Toss out summer tournament records, every tournament you've got players out on vacations and different rosters wiith coaches hopefully focused on moving players around preparing for league.

And so does every other team. Deal with it or make excuses. Your choice.

[Unregistered]

girls 02 ranking

toss out summer tournament records, every tournament you've got players out on vacations and different rosters wiith coaches hopefully focused on moving players around preparing for league.

run whatcha brung !

[Unregistered]

run whatcha brung !

That's all great for kids getting experience but if you're looking for statistical relevance in those quant models you may want to rethink your data samples. Would you include pre-season football games in your analysis of regular season match ups?? Of course not, so you mental midgets may want to rethink how relevant some of these results are to current match ups. Vegas would make a fortune off you chumps!

[All for One]

Chumps or Champs?

That's all great for kids getting experience but if you're looking for statistical relevance in those quant models you may want to rethink your data samples. Would you include pre-season football games in your analysis of regular season match ups?? Of course not, so you mental midgets may want to rethink how relevant some of these results are to current match ups. Vegas would make a fortune off you chumps!

Thank you for all the compliments. Trust me, much thought has gone into the statistical modeling and data sampling over the last 3 years that the LastPlanet rankings have been in use. The effects of including and excluding various types of games have been tested and what we've found is that the most accurate predictions occur when the full dataset is used to calculate each team's attack and defense strengths.

As just one small example, we recently predicted the results of the NWCL games for the G01 age group (where we have the most complete dataset). The accuracy of the predictions was over 90% for outcomes and over 50% for actual scores. I suspect we could do very well in Vegas with results like that :)

As for teams having different rosters for different games within the year, our position is that this is going to happen with all teams at all times of the year. If a team plays 50 games in a year and uses 15 different rosters due to vacation, injury, illness, etc, was that 15 different teams or just 1? Are the Portland Timbers not still the Portland Timbers when they don't have Will Johnson in the lineup? Further, the best teams typically have the deepest benches (talent wise) and therefore continue to have success when they are missing players.

One last note. The LastPlanet rankings are time-weighted. That means that more recent results carry more weight in the ratings calculations than do results in the distant past. So, if a team is not ranked where you think it should be right now, don't blame those mid-summer losses. Those have much less impact on a team's ranking than do the games that were played last weekend.

I hope this clarifies things for you.

[Unregistered]

Thank you for all the compliments. Trust me, much thought has gone into the statistical modeling and data sampling over the last 3 years that the LastPlanet rankings have been in use. The effects of including and excluding various types of games have been tested and what we've found is that the most accurate predictions occur when the full dataset is used to calculate each team's attack and defense strengths.

As just one small example, we recently predicted the results of the NWCL games for the G01 age group (where we have the most complete dataset). The accuracy of the predictions was over 90% for outcomes and over 50% for actual scores. I suspect we could do very well in Vegas with results like that :)

As for teams having different rosters for different games within the year, our position is that this is going to happen with all teams at all times of the year. If a team plays 50 games in a year and uses 15 different rosters due to vacation, injury, illness, etc, was that 15 different teams or just 1? Are the Portland Timbers not still the Portland Timbers when they don't have Will Johnson in the lineup? Further, the best teams typically have the deepest benches (talent wise) and therefore continue to have success when they are missing players.

One last note. The LastPlanet rankings are time-weighted. That means that more recent results carry more weight in the ratings calculations than do results in the distant past. So, if a team is not ranked where you think it should be right now, don't blame those mid-summer losses. Those have much less impact on a team's ranking than do the games that were played last weekend.

I hope this clarifies things for you.

Just wanted to say thank you for all your hard work! Keep rockin on!

[Unregistered]

Thank you for all the compliments. Trust me, much thought has gone into the statistical modeling and data sampling over the last 3 years that the LastPlanet rankings have been in use. The effects of including and excluding various types of games have been tested and what we've found is that the most accurate predictions occur when the full dataset is used to calculate each team's attack and defense strengths.

As just one small example, we recently predicted the results of the NWCL games for the G01 age group (where we have the most complete dataset). The accuracy of the predictions was over 90% for outcomes and over 50% for actual scores. I suspect we could do very well in Vegas with results like that :)

As for teams having different rosters for different games within the year, our position is that this is going to happen with all teams at all times of the year. If a team plays 50 games in a year and uses 15 different rosters due to vacation, injury, illness, etc, was that 15 different teams or just 1? Are the Portland Timbers not still the Portland Timbers when they don't have Will Johnson in the lineup? Further, the best teams typically have the deepest benches (talent wise) and therefore continue to have success when they are missing players.

One last note. The LastPlanet rankings are time-weighted. That means that more recent results carry more weight in the ratings calculations than do results in the distant past. So, if a team is not ranked where you think it should be right now, don't blame those mid-summer losses. Those have much less impact on a team's ranking than do the games that were played last weekend.

I hope this clarifies things for you.

How do you adjust for the brackets teams play in? There are some lower division teams ranked higher then premier teams.

[All for One]

How do you adjust for the brackets teams play in? There are some lower division teams ranked higher then premier teams.

The short answer is that we don't adjust rankings to compensate for the division or bracket a team has been placed in. The only data that is used in the calculation are:

Date
Home team score
Away team score
Field type (grass or turf or unknown)
Format (8v8, 9v9, or 11v11)

What happens is that the program works its way through the database of game results and predicts what the score of each game should be (based on previous results of the 2 teams in question). It then compares that prediction to the actual score and adjusts each team's strength ratings (attack and defense) accordingly. It takes about 7-8 games for the ratings to converge and stabilize and from there the changes are much smaller from game to game (this is why we don't display rankings for teams that haven't played at least 8 games).

For example, if the program predicts the score of a game to be 3.5 to 1.2 for the home team and the actual score is 2 to 1, the home team will lose some attack rating (only scored 2 vs 3.5 predicted) but game some defense rating (only allowed 1 vs 1.2 predicted). Similarly for the away team, they would lose a tiny amount of attack rating (only scored 1 vs 1.2 predicted) while gaining a fair amount of defense rating (only allowed 2 vs 3.5 predicted).

After the program has run through all of the games and calculated every team's strength ratings it then uses those attack and defense ratings to calculate an overall strength rating. This overall rating is what is used to rank the teams.

That's the long way of saying that it's quite possible for a team in a lower division to be ranked higher than a team in a higher division. It all depends on the strength of the teams you're playing and on how well you do against those teams. There are no artificial factors or adjustments to make sure that Div 1 teams are all ranked above Div 2 teams, for instance. Last year in G01 RCL league there was a team in division 3 that was ranked higher than some teams in division 1. This might seem odd until you realize that that team is now playing in division 2 and will be promoted to division 1 in a few weeks. This was a case of a very good team being improperly placed (due to league rules or DoC mistake) in the wrong division.

Hope that helps.

[Unregistered]

The short answer is that we don't adjust rankings to compensate for the division or bracket a team has been placed in. The only data that is used in the calculation are:

Date
Home team score
Away team score
Field type (grass or turf or unknown)
Format (8v8, 9v9, or 11v11)

What happens is that the program works its way through the database of game results and predicts what the score of each game should be (based on previous results of the 2 teams in question). It then compares that prediction to the actual score and adjusts each team's strength ratings (attack and defense) accordingly. It takes about 7-8 games for the ratings to converge and stabilize and from there the changes are much smaller from game to game (this is why we don't display rankings for teams that haven't played at least 8 games).

For example, if the program predicts the score of a game to be 3.5 to 1.2 for the home team and the actual score is 2 to 1, the home team will lose some attack rating (only scored 2 vs 3.5 predicted) but game some defense rating (only allowed 1 vs 1.2 predicted). Similarly for the away team, they would lose a tiny amount of attack rating (only scored 1 vs 1.2 predicted) while gaining a fair amount of defense rating (only allowed 2 vs 3.5 predicted).

After the program has run through all of the games and calculated every team's strength ratings it then uses those attack and defense ratings to calculate an overall strength rating. This overall rating is what is used to rank the teams.

That's the long way of saying that it's quite possible for a team in a lower division to be ranked higher than a team in a higher division. It all depends on the strength of the teams you're playing and on how well you do against those teams. There are no artificial factors or adjustments to make sure that Div 1 teams are all ranked above Div 2 teams, for instance. Last year in G01 RCL league there was a team in division 3 that was ranked higher than some teams in division 1. This might seem odd until you realize that that team is now playing in division 2 and will be promoted to division 1 in a few weeks. This was a case of a very good team being improperly placed (due to league rules or DoC mistake) in the wrong division.

Hope that helps.

Well when you look at those lower teams being ranked above premier team and you actually see teams play you know the rankings are not as accurate as one thinks. If you added in recreational teams say one that never looses then it could be possible that a recreational team could be ranked quite high?

[All for One]

Well when you look at those lower teams being ranked above premier team and you actually see teams play you know the rankings are not as accurate as one thinks. If you added in recreational teams say one that never looses then it could be possible that a recreational team could be ranked quite high?

For that to happen the group of teams that this hypothetical rec team belongs to would have to be completely isolated from all of the other teams in the database. If that happens, and it can happen, then those teams are excluded from the rankings. What tends to happen, though, is that over the course of a summer/fall there is enough cross-play between teams in different leagues, divisions, states, etc. that all teams become "connected" to all other teams through one or more of these "cross-play" games. In the G01 rankings, for instance, there are over 1500 teams and none of them are isolated from the rest of the teams. The same is true with the G02 rankings although the number of teams is considerably less.

If you have a specific team or teams that you believe are ranked too high/low let me know and I'll double check the game data to make sure there are no errors. Occasionally we get games applied to the wrong teams because most teams use several different names for leagues, tournaments, etc, making it difficult to match them up later.

[Unregistered]

Thank you for all the compliments. Trust me, much thought has gone into the statistical modeling and data sampling over the last 3 years that the LastPlanet rankings have been in use. The effects of including and excluding various types of games have been tested and what we've found is that the most accurate predictions occur when the full dataset is used to calculate each team's attack and defense strengths.

As just one small example, we recently predicted the results of the NWCL games for the G01 age group (where we have the most complete dataset). The accuracy of the predictions was over 90% for outcomes and over 50% for actual scores. I suspect we could do very well in Vegas with results like that :)

As for teams having different rosters for different games within the year, our position is that this is going to happen with all teams at all times of the year. If a team plays 50 games in a year and uses 15 different rosters due to vacation, injury, illness, etc, was that 15 different teams or just 1? Are the Portland Timbers not still the Portland Timbers when they don't have Will Johnson in the lineup? Further, the best teams typically have the deepest benches (talent wise) and therefore continue to have success when they are missing players.

One last note. The LastPlanet rankings are time-weighted. That means that more recent results carry more weight in the ratings calculations than do results in the distant past. So, if a team is not ranked where you think it should be right now, don't blame those mid-summer losses. Those have much less impact on a team's ranking than do the games that were played last weekend.

I hope this clarifies things for you.

I assume the N value shown for each team is the number of games used as your sample. What is the reasoning in there being such a wide discrepancy in the N value between teams? Some have an N in the single digits and others are in the high 20's.