I never thought it would take us this long to release these (it’s been six months of FTS now!).  It was something we intended on rolling out immediately because I think it demonstrates nicely what’s subjective about computerized ranking.  It also offers some different perspectives on rankings, and maybe there’s something here that is closer to how you think teams should be ranked.  Such things like full blowouts, just wins/losses, no tournaments, and the daily sampled bout decay are all being experimented with over there.  See the new Ranking Variations page.

It should be noted that these variations are isolated changes to the standard FTS ranks.  I believe that if you wanted to pursue the philosophy of one of these variants as a ranking system, there’s potentially more to be done to make the system more robust (or starting from scratch on a system that foregrounds that concept).  For instance, it’s interesting to see what the FTS system looks like if you take out extra tournament bonuses.  But if we wanted to seriously pursue a system that ignored tournament results, we would probably tweak the way other parts of the system work.  Since the FTS system relies on tournament reputation for some amount of it’s weighting, simply removing it in this variation produces some illogical results because a significant balancing element is missing.

But knowing that, I think the variations are helpful.  For now they may just serve a perspective relative to the standard FTS ranks, but maybe in the future, rebuilding some of these philosophies from the ground up could be a way to go.

Just wanted to give a heads up that Tucson’s drop 27 points in Today’s Ranks is a combination of two things that happened this weekend - their loss to Windy City, and the disappearance of a win to Madison from one year ago.  If you’ve been following along since we released the new ranking system, you’ve noticed teams moving despite having not played in the last week.  This is a bout from exactly a year ago that just popped outside of our view (full explanation here).  Today (Monday), it actually looks pretty clean.  Only Tucson and Madison move in this way, and everyone else had a bout this weekend.

We’re working on some ways to make this a lot clearer - to differentiate between movement from recent bouts, movement from disappearing bouts, and today’s extra confusing scenario where both things happened.  Until then, watch out for it!

Attention WFTDA voters!  Recent derby craziness making you fear your looming task?  Maybe we can help.  We’ve written a tool that lets you drag-and-drop your own custom ranks, quickly bring up recent bouts, and analyze your ranks for logical errors.  Choose last quarter’s WFTDA ranks (or any other preset) for a starting place if you like.  Then Export the plain text to your people and call it a day!  Of course, non-WFTDA voters are welcome to play with it too.

Let us know how it works for you, or if you have ideas for comparing teams, and coming up with awesomely accurate WFTDA ranks!

Note: Older browsers or computers may have some trouble with the drag/drop.  Firefox 3 should work without a hitch though.

The updated ranking system is finally done and online today.
These are the major changes, which I’ll document more fully in the coming days:

• One Year Bout Decay:  Bouts played longer than 365 days from today’s date are ignored.
• Re-seeding with all bout data:  All bouts are re-run starting from 2005.  Even though these are outside of the bout decay, they can effect later games by propagating down and informing what an opponent’s skill may be.  This doubles the dataset with an extra year and a half of data, as previously we were only looking at 2007 to now.
• New Tournament Results:  We calculated all tournament placements equally and by our own method Seeding By Outgoing Ranks (fully documented over here)
• Tournament Decay and Weighting:  This had never been an issue before, but is necessary when reseeding from 2005: Reputation from a tournament season lasts until the next tournament replaces it.  Another way of saying this is the national champions remain the national champions until they are dethroned.  This effect already existed, but the system now knows what to do with historical tournaments, and the upcoming tournament season.  Also the tournament bonus is weighted less overall, which is one factor that makes the rankings more mobile in the upper ranks.
• Tightened Blowout Adjustment:  Tightened up this curve a bit so the largest blowouts are even more equivalent to normal size blowouts.  Or to put it another way, bigger blowouts mean less.
• Low Participators as Opponents:  When we introduced what we affectionately refer to as the Kiddie Pool, we didn’t have a sound method of addressing what happens when a team plays a low participator.  The new rule is that if you are in the Kiddie Pool as a newbie (2 or less bouts), your calculated skill to an opponent is the worst rating in that quarter’s ranks.  If you are in the Kiddie Pool as a veteran (more than 2 bouts ever, but bout decay gives you only 1-2 in the last year), your calculated opponent skill is not penalized.

If you’ve been following along things will look a lot different.   Teams are in general a little more mobile (this was one of our goals), so if you’re used to a certain outcome generally giving +5, you’ll probably find a new set of norms.

There is one major caveat that I’d like to point out in using a bout decay.  Since last year’s regionals happened 6 months apart (!), Eastern division teams get their tournament bouts counted, while Western division teams do not (because the 2007 Dust Devil is more than one year ago).  The affect is Eastern teams get pushed up a lot more than Western teams (at least those that had nice outcomes at Regionals).  Rat City probably feels that effect the most and drops a few places for it.  We debated whether we should artificially include Dust Devil games to even up the top ranks, but this just feels like ignoring the reality of the situation, and the fact that this upcoming season fixes the issue.  At the very least, you can use the filters to see Western and Eastern compared amongst themselves.

If you’d like to know about the gritty details of these decisions, keep your eye on the Development Blog (and subscribe if you like).  I’ll be unloading the hows and whys of each of these things over the next several days, plus a bunch of observations we’ve come to on how these can be be better, and where to go from here.

Edit: More posts from the Dev Blog

• Bout Decay
• Low Participators as Opponents

New feature in FTS land: Add/Edit Bouts*. It’s now super easy to keep us updated with bout information, whether it be a bout that ended 20 minutes ago, or loading us up with bouts from two years ago. We’ve gotten a lot of info from folks over the months via the feedback form, but we wanted to make it even easier for you (and us). Please verify your team’s history if you haven’t already done so, otherwise keep doing what you’re doing, and this form should make the process much more streamlined. Also feel free to ask us if you’re not sure if a bout belongs on here or not. You can see the basic outline of the matter here.

*[Actually the Edit Bouts part is just a tease - that will be for another day. Send existing errors the old fashioned way.]

Ok still no new rankings yet. We’ve got a couple kinks still yet and we want to release them in their best possible state. But since it’s been a while since anything new has shown up around here, and we’ve been sitting on a couple goodies, we figured they should be introduced. It all can be found on a team’s page.

First, there’s a graph showing a team’s performance over time. Note that switching between Absolute and Relative mode changes the scale of the graph (hover the ? for more interactive things you can do).

Secondly, the Bout History section got an overhaul. We’re now tracking Unofficial and Upcoming bouts, and we’ll be doing some cool stuff with that data in the future. Right now Unofficial bouts are Mixed Bouts between a travel team and a local league team, or a bout that for some reason is not considered by WFTDA to be official. See the FAQ on official bouts for more info on that. We’re not currently tracking bouts between two local league teams.

The unofficial bouts are probably not so complete, and we’re hoping you can help us fill them in. Soon we’ll have an even easier way for updating us with new new bouts or correcting existing ones. Until then you can always send us data free-form via the feedback page. It should be said that unofficial bouts probably won’t be affecting aggregated stats, and almost certainly will not be affecting rankings. We do think it’s valuable information to consider when thinking about a team though.

We’re decidedly late in matching a quarterly ranking to WFTDA’s newly released rankings. We have some changes to make to our ranking system, and we want to use this new quarter to phase them in, so it could still be a bit while we iron that out. Until then, you can still play with the new WFTDA ranks.

The WFTDA History section is showing the new ranks (as well as ranks back to 2006). The FTS-WFTDA Compare page now shows the new rankings, and a snapshot of our ranks on April 1st. When we make the changes to our system, we’ll likely replace this FTS 2008 Q1 with the new snapshot, so keep in mind that this page is temporary.

I’m adding a second blog to act as half site changelog/half development blog. I’ll be keeping track of all the features (big or small) that creep onto this site as we add them, as well as any decisions we make on the algorithm.

We’ll still post about the major features or changes in this blog, but I wanted to have a place to document the more technical minutia. If you want to know all the gory details, subscribe to the feed over there.

Aaron and I took some time last night to create a helper page to identify logical errors in Flat Track Stats rankings. As some of you may have noticed, there are some instances of teams being ranked lower than a team that they have beaten, or inversely, ranked higher than a team they have lost to.

The Logical Errors page looks at all bouts within a year of the viewing date, and according to each win and loss, confirms that the team is above of or below their opponent logically. If the ranking is illogical, it spits out an error for that bout. This page does not argue for any specific re-ranking of teams, but simply displays any logical errors that exist in the current rankings to help understand their placement better. While the errors may point out where our rankings need improvement, most errors are inevitable (upsets, old bouts, or inconsistent bouting record).

We also realized that the same tool may be helpful if applied to the quarterly WFTDA ranks, but that will have to wait for another night.

We added all old bouts before 2007 today. We’re now going for a complete dataset of all official travel team bouts under WFTDA, so if it’s not here let us know. Unofficial bouts of WFTDA teams are coming soon too. For now, this for archival and stats purposes only - it does not affect our rankings. We are going to explore re-seeding our rankings from the beginning of time when we a get chance to.

Also, we want to thank everyone for all the feedback and fixes we’ve received on this site. It may not be apparent on the surface, but we’ve made a lot of fixes to the bout data itself, based on errors that people are finding. The feedback has been fantastic, and it’s exciting to see this data really get cleaned up with everyone’s eyes on it!