Quote:
Originally Posted by Navid Shafa
You know the normal qualification system needs work if 9/76 teams attending IRI didn't play at the World Championship.
|
Perhaps a better way to put this is that the IRI committee considered these 9 teams more qualified than the 28 CMP qualifiers who applied and were not accepted. Three of these 28 were 1285, 341, and 193, who weren't accepted for other reasons already addressed. Meanwhile, 2 of the 9 teams you refer to are 234 and 1024, who are hosting the IRI...
Also, this statistic could be far worse. That's only 9 teams (maybe as many as 20 non-CMP qualifiers including those who didn't apply for the IRI) who are "better" than some of the 400 CMP qualifiers. And by this IRI metric (if we're gonna call it that), this 5% (20/400) is only "better" than 10% (28, or ~40 teams) of those who qualified. I feel pretty comfortable dismissing this 5% error as statistical noise, rather than claiming a correlation to an inaccurate qualification system (particularly since the IRI and CMP both have about 5-10% noise built-in with auto-qualifications).
Edit: it might be neat to actually figure out some of the statistics here. We could call the qualification system "error" the difference between some ranking that qualifies you for CMP and some other ranking that we like better (such as qualifying for the IRI, worldwide district points, etc.). The above suggests the error distribution is rather narrow around zero (assuming IRI acceptance is the gold standard of qualification systems...). At the very least, even if both CMP qualification and IRI acceptance are horrible measures of team or robot quality, the discrepancy between the two measures is small.