Yes that's right, today we finally release what many have suspected was coming: a complete North America ranking system. With North America itself providing over 1,600 teams, more than 10,000 bouts, and going back as far as 2004, we've been able to create a rating system that includes all traveling teams (both WFTDA and non-WFTDA), b-teams and home teams. The only requirement is that they play by the WFTDA rule set.

Our default landing page will continue to be the WFTDA sanctioned only system, but if you use the front-page switcher, you'll now see both a USA and a Canada selector.  These are both filters on the same North America ranking system.  They show all the week's bouts that we know about in addition to the top 20 rankings for that region.  In addition, you get all the functionality you've come to expect from us, including bout predictions, ratings over time and detailed bout stats (if you send them in) for every team.  For those teams that exist in more than one ranking system (WFTDA and North America), you'll find buttons to switch between the systems so you can compare. 

A couple of things are worth noting:

  • The data are only as good as you tell us.  We've already found some discrepancies as we've been testing this in the background, so please go check your team's results and let us know if anything's not right.
  • Home team's still have a home/away bias.  Yes I know that's counter intuitive.  I tried really hard to get rid of it, but for whatever reason, the data show something's there.  Maybe it's that previous data reporting tended to place the winning team in the home-team position.  I don't know, but when you submit bout scores, please be consistent about how you report the home and visitor status.
  • Some team's rosters can be a little fluid.  From our perspective, if  you skate under your team's name, then it goes to your team's rating.  Don't worry, over time the system tends to self-correct.  However, if you want to create an "exhibition team" that's your_team_and_friends (and your opponents are cool with that), we'll happily add that to our list  of teams.
  • Finally, be patient as we get some experience with the new load.  Some page loads may be a little slow.  Tracking all of North America brings many more bouts each week and we have no idea what the loads are going to be.  Just re-calculating the ratings after updating bout information can take over 10 minutes (although hopefully you never have to see that).

So go forth and explore.  Feel free to send us feedback or comment directly below.  While this was the really big one for us, we've still got one more ranking system in the pipeline (the prototype is already running behind the scenes) but we can't decide what to call it.  Hopefully we'll get that one out in the next month or two.  Then we can relax, sit back and enjoy some really excellent 2013 derby.


Comments

Just curious as to why teams who haven't played in years show up fairly high? Ohio Roller Girls disbanded our home teams in 2008, after 3 seasons, yet all four of them are showing up in the top 400ish of teams.

Thanks for the comment.

Once we know that a team has been disbanded we add the disband date and that team appears with a strikethrough across the team name. (I just edited the Ohio Roller Girls page: http://flattrackstats.com/teams/3393 so you can see what that looks like.) This pulls the disbanded team from the rankings list, however all of the team data stays with FTS and any past bouts that are missing should still be added because these bouts play a role in training the algorithm.

In addition to the specifics of your inquiry it is also true that we are dealing with a lot of data on a weekly basis and that we rely on public input about that data to keep it accurate and relevant. If you see anything missing or incorrect on your league’s FTS page, in the rankings, or anywhere on the site please send us an email via one of the forms at the bottom of the page under “Help Us”. We do our best to respond quickly to those questions and update the site accordingly.

Thanks for clearing that up-- I guess my actual question beyond the specifics of that situation was really more about age of data -- do the games still decay after a certain amount of time? Ultimately, what I was surprised by was teams that hadn't played since 2008 showing up higher than currently bouting teams.

No, games don't "decay". A team's rating value will only change as a result of a game. If you don't play, it won't change. The logic being that the best information we had, was the last time you played and no new bouts means no new information.  So in 2008, the OHRG home teams were playing as low-mid 600 teams.  That should be seen in context with what other teams were rated as in 2008.  What you correctly perceive is that a 600 team today would probably beat a 600 team from 2008.  So yes, this can create some anomolies if teams don't play for long stretches. 

Is there a way to see the rankings of the teams in specific leagues? Obviously, the overall North American or USA rankings are pretty useless (at least in terms of home teams). You can't compare closed systems like home leagues are. The #24 team in the country is definitely not a Sacred City home team and that home team is definitely not better than the Sacred City all-star team. If they are, Sacred needs to rethink their all-star selection process.

From a data standpoint, this update is awesome but I really just want to see the rankings within the individual leagues.

Yeah, I wondered how long it would take for someone to find that Sacred anomaly.  Basically, those home teams are hanging off of a single connection to the main system.   To answer your main question though, we do not have a filter-by-league capability.   You’re right it would be nice to see a league’s home teams all in one view.   We’ll have to think about how to implement that.

That being said, I don’t think I can agree that the “overall North American or USA rankings are useless.”  While it’s true that closed systems may not be accurately placed within the larger ranking system, I think you’d be surprised at how well connected the system is.  As of today, the North American ranking system represents 741 leagues (1432 total teams).  Of those, 1232 teams are connected through at least one bout to the primary ranking system.  There are currently 81 disconnected closed systems (with between 2 and 5 teams each) representing a total of only 199 teams (less than 14% of the total).

Interestingly enough, not all of those disconnected groups are single league home teams.  There are some young travel teams that haven’t had time to venture beyond their nearest neighbors and there are some neighboring leagues that have created a closed bubble by having their home teams play each other. 

Of course, as you correctly point out, there can be some semi-closed groups like the Sacred home teams that are only connected to the main system by one bout.  If, as in this case, that one bout happens to be a blowout or have an otherwise anomalous result, then the whole group may be misplaced relative to the overall system.  Based on the work I did over the winter, I don't think there's more than a couple of those.  However, rather than seeing this as a flaw to be hidden or avoided, I see this as the starting point for a system that will only grow better as more data and more bouts come in.  Personally I hope that Sacred’s home teams will play more outside bouts.  Those bouts will help correct the ranking system, but more to the point it will help raise the overall quality of the surrounding leagues. 

In fact, it is exactly the possibility of home teams playing outside bouts that was one of the driving reasons for making this system.  Some home teams from the better leagues would be pretty good competition for the second tier travel teams in their area.  How would they know that scheduling a bout might be worth it if they hadn’t already played them?  The sport is growing so rapidly that distances between leagues are becoming smaller and easier to overcome.  I don’t expect these closed bubbles to remain closed for very long.  And hopefully, even if our initial rankings aren’t exactly right, we’re giving teams new options for competitive bouts that they might not have previously considered.   Maybe I’m biased, but to me, that sounds useful.

That sounds like a really big cluster map.

Correction: a really big and awesome cluster map.

Totally, we've got the data ready to plot, I just have to get around to doing it!