With the new FTS comes lots of questions.  In the interest of fostering a better understanding by all and encouraging a lively dialog, from time to time we will select an interesting or representative question from the mailbag and present the discussion here.  

How far back does the algorithm go?  

Pain Newton Asks:
How far back does the algorithm go?  Does it give more weight to recent wins?   I personally think that games within the last 3 months would be given 100% weight and then start scaling back from there to where games over a year old have zero weight in the current ranking.  I know that eliminates a bunch of data, but anything past a year in sports is pretty irrelevant in my book when it comes to measuring a team’s success.  You could use that data if nothing else is available, but it if you removed it from the algorithm it would reward the teams that keep a full schedule versus teams that play only a couple sanctioned bouts a year.  Personally I think we as a roller derby community should try to reward those teams that are out there trying to compete....

David Answers:
We actually start our system all the way back at the dawn of WFTDA in 2005.  However, it's important to realize that we are not averaging this information, so the idea of weighting bouts becomes irrelevant.  Rather, our rating system is constantly evolving to reflect the latest bout information that we have.  Going into each bout, the best information we have is the very last performance of each team...whether that was 6 hours ago or 6 months ago.  We use that as a seed for our calculations and compare those expectations to the actual result.  If the prediction was spot on, nothing changes.  If it's off, the rating is modified....that then becomes the seed for the next bout....rinse and repeat.  As a result, there's no ability for the past to really be holding a team to any rating that is unsupported by their current performance.

I think your larger point -- that teams should be rewarded for keeping a full schedule -- is perhaps a good one, but I don't think that it really applies to the realm of statistical measures of team performance.  It's more of an ethical or philosophical position.  One can, perhaps, take comfort in the notion that those teams which play more games, will eventually become better....and that a system which is not biased by subjective perceptions will religiously track their progress up through the ranks.