FAQ

Q1. We beat team A 3-0, but they are rated higher than we are. What’s up with that?

A1. Your team’s rating increased as a result of that game, and team A’s decreased by the same number of points. If you keep beating team A, or teams rated similarly to team A, your rating will eventually exceed theirs, but it takes time, and games, for that to be reflected.

Q2. My team’s rating decreased, but we haven’t played any games in the interim. Did you make a mistake?

A2. There are two possibilities.

The first is that is that we hadn’t included a tournament your team played in, and now we have. We try to keep current with the tournaments (we cover over 600 of them), but it takes a long time to go through all the tournaments.

The second possibility is that another team in your league, or state, or region, played in a tournament somewhere, and did poorly. (Ditto if your rating increased.) We use tournament performance to “normalize” the ratings nationally (so that an 800 in Idaho and an 800 in Florida reflects two teams of comparable strength). If a high-rated team goes to a tournament and loses badly to lower-rated teams, we adjust not only their rating, but also the ratings of teams they have played (league, state, regional). That’s the only way we can normalize ratings nationally. We are working on a "relaxation algorithm" that will do this automatically.

Q3. What determines whether or not you cover a tournament or a league?

A3. We only cover tournaments and leagues that post actual scores on the internet. We don’t cover tournaments or leagues (like Virginia's VCCL) that only post standings, or champions/finalists. We cover almost every division of every league, though in big leagues (like IWSL or LIJSL), we do not cover the very lowest divisions unless at least 1/3 of the teams have ratings; then we start to cover that division.

Q4. How do you deal with a game that goes to PKs?

A4. The game is entered as a tie, because it was.

Q5. Why won’t you tell us more details about the algorithm?

A5. Because we don’t want anyone to steal it and use it for Evil, not Good. If we give lots of examples, someone with some mathematical sophistication could fairly easily reproduce the algorithm, and then start doing their own ratings. We have spent about 2500 hours on this over the last three years. We don’t want someone else to make money off Mathman's idea.

Q6. Is there any way I can help you?

A6. Yes. We probably spend at least as much time trying to figure out who a team is as we do entering scores, running games through the algorithm, and figuring out updated ratings.

• If your team changed names, or changed clubs, let us know.
• Keep your damn team names the same! Don’t switch the team’s name from “Storm White” to “Storm Green” (especially if there is a team named “Storm Green” which is switching its name to “Storm White”!). If you go to a tournament, make sure your team is listed under its full name.
• Don’t just use the name of your club as your team name unless your club only has one team in your age group.
• If you’re playing up, demand that your real age group be added to your name in the tournament listing.
• Don’t give your team two names (e.g., “Warren County Red” and “Warren County Mud Dogs”). This drives us crazy.

  If we can’t determine which team is playing in a tournament, we just ignore their results (and we would rather not!).

  Q7. How do you account for injuries to key players, absences due to proms and family activities, and guest players?

  A7. We can’t, so we don’t.

  Q8. How reliable are the ratings?

  A8. For teams that play each other a lot, very reliable. The greater the geographical distance between the teams, the less reliable they are. But we do try very hard to normalize the ratings between the states and regions. This is easiest at U14, since then there are regional leagues (the MWRL, or Region I Premier, for instance) and, of course, Regional Championships and the National Championship, that allow this. The interstate and national ratings are less reliable at the younger age groups, since those teams travel large distances less often.

  For instance in SoCal’s Coast League Cup in 2006, for the U14 age group, the higher-rated team won 75% of the time, tied 11% of the time, and lost 14% of the time. If you restrict the considered games to those in which the teams ratings differ by more than 75 points, the higher-rated team won 84% of the time, tied 9% of the time, and lost 7% of the time. In those losses, there were only three games (out of about 80) where the higher-rated team lost by more than one goal.

  Q9. If a team improves greatly, how long will it take for this to be reflected in its rating?

  A9. Statistical experiments we have run show that it can take upwards of 20 games for the team’s rating to reflect its “true” rating. But in practice, if a team has improved significantly, this is reflected fairly quickly, usually within 8 games.

  Q10. Do you fiddle with the ratings to make sure your daughter’s team appears stronger than it is.

  A10. No. Our daughters do not play in any of the age groups we cover.

  Q11. Have you seen any of these teams play, and if so, does your impression of the game affect their rating?

  A11. No, hence no.

  Q12. I don’t think your algorithm reflects the strength of teams. Our team finished first in our league, yet we are rated lower than Wazoo Green Strikers Elite, who finished last in our league.

  A12. This probably happened because your initial rating was low, and Wazoo’s was high. Your rating increased greatly over the course of the season, and Wazoo’s decreased significantly. If this continues, your team will eventually be rated higher than Wazoo. Patience, Grasshopper….

  We have no interest in discussing the correspondence between strength and rating. If you don’t like the results of the algorithm, then invent your own.

  Q13. What do you think of the other soccer rating websites, i.e., Gotsoccer, USA Rank and Nationalsoccerranking?

  A13. They’re fine. They use a different algorithm. Look at their sites and make your choice.

  Q14. Why the name Mathman?

  A14. The inventor of the algorithm is a math professor (Ph.D, etc.), so it seemed appropriate.