FIRST has used a largely “random” method for assigning teams to Championship divisions. Frank Merrick posted a couple of years ago (I’m sure someone can find his comment) that FIRST assigns teams based on when they submit their entry, with a special consideration of rookie teams (that I can’t remember at the moment.)
The problem is that random is arbitrary, not fair, unless there are many repeated trials (probably at least 30 based on standard statistical principles.) No student is going to experience more than 4 Champs, so we can never approach true fairness for any individual student.
This season we could see in Houston how divisions could have concentrations to higher performing or lower performing robots that skewed ranking results and playoff opportunities. Newton had 5 of the last 8 world championship teams (118, 120, 330, 1678, 5012), and either of the final two alliances in Carver would have been equally competitive on Einstein. Last year, the Newton alliance was so stacked that the alliance power was too dissipated to bring the strongest alliance to Einstein. (We may have been the beneficiary of an overly strong Curie field in 2013 as well.) This is a clustering phenomenon that is more likely to occur when a group is segmented into smaller and smaller subgroups (as occurs as we move from 4 to 12 divisions.)
Instead, FIRST should consider assigning teams to divisions based on a performance metric such as district points. The Michigan State Championship already does this and has an algorithm approved by FIRST. The Blue Alliance reports district points earned at each regional event, so that calculation would be simple. Because this is only about assigning teams to divisions and NOT about qualifying for Champs, the pre-event ranking need not be perfect or precise–it’s only intended to mitigate the clustering effect of random assignment.
(I have a different solution to the scheduling problem of random assignment, but that’s for another day.)
I’d modify it so that it is by state or region as well where applicable. If we want teams to have an experience of playing with teams from half the world at the half champs, a given state (MI, MN, CA) or region (CHS, NE) that has more than 6 teams that qualified should be separated.
I fully agree. I can tell first-hand that Carver felt like an absolute bloodbath with several Einstein-worthy competitors, so it makes sense to divvy up the teams with more reasoning.
The end goal would be to eventually move all FRC teams to districts. However, since that is unrealistic in some regions (Hawaii, Cali, Minnesota to name a few) I imagine we should come up with an alternative system that can still provide a unified ranking system without the logistical issues a full district would have.
If each region had a unified ranking system and sent a proportional amount of teams to champs based on team population, it would allow FIRST not only to balance the divisions out better but also to spread out teams from each region in a better planned way.
Here’s a comparison of some stats across Houston divisions, compared to MSC divisions… Houston divisions had a huge disparity in RP accumulation in quals (16.1%, 25.5%, 37.5%, 39.1%, 42.4%, and 59.4%), while MSC variations in RP accumulation in quals were much smaller (33.1%, 38.1%, 49.4%, and 51.3%).
Also of note, is that the RP rates in Houston Qual matches perfectly predicted the relative frequency of the bonuses in Playoffs! Furthermore, the three divisions with the highest Qualification RP rate were the three highest ranked alliances on Einstein after Semifinals.
This indicates that the random allocation of teams to divisions - at least in Houston - resulted in very misbalanced divisions, which resulted in very misbalanced playoffs, which resulted in misbalanced Einstein matches.
Do you want a districts centithread? That’s how you get a districts centithread.
Just using the districts point system as a way to allocate teams and balance divisions is a great idea. While the current algorithm is fair in some sense of the word, any team could be in any division, not biased for/against any one team, it isn’t fair with respect to difficulty, which I think would be a desirable trait. Minnesota already applies (a slight modification of) the district point system to teams in their regional system, to qualify for their state high school league competition. I don’t see why FIRST couldn’t use this to balance divisions.
The thing I’ve noticed with 12 divisions and split champs is clustering divisions. Turing was a large part PNW teams, Darwin has 1/3 of all canadian teams, almost all of the Colorado teams were in Galileo. People don’t travel to the world championships to play with people they’ve already played with. This should probably be addressed too.
I agree. In the Daly subdivision, 23/68 teams are from Michigan (give or take a few, may have miscounted). Granted, Michigan sent 91 teams to champs this year, but even still, divided among the 6 subdivisions, that would only be about 15 teams per field. In addition, many of the Michigan teams on Daly have been some of the top competitors in the state this season (2767, 3620, 27, 4967, 910, 2960, just to name a few). I think if the way divisions are assigned changes, eliminating over-representation of a certain area in one division is something to look at.
I know last year MN was clustered heavily on one field, but this year we’re actually spread out almost evenly (2 divisions with 4, 2 with 5, and 2 with 6), and there’s no clustering of our teams based on how they qualified - 5 chairman’s teams across 4 divisions, at least 1 event winner in each division (and no more than 2 in any division). I prefer being spread out like this, as my team gets to interact with more new faces… but as a volunteer without a team last year, it was nice to have the one field I could watch with a bunch of familiar teams.