[EDIT]Please scroll down to see the updated histogram, this one is missing the Michigan events and NYC![/EDIT]
Above is a histogram (bar graph) of OPR at the end of qualifying rounds on Saturday of week 2 in 2011. OPR is a statistic based on which teams a team plays with, and their cumulative alliance scores. Here is the mathematical description. It is a rough estimate of the number of points a given robot is “worth” per match. Negative scores may be the result of frequent penalties, or the algorithm “overcompensating” if a good team has a bad match. It is far from a perfect system. The average OPR during week 2 was 9.16 points, which is slightly higher than week 1’s 8.77 points.
There are 340 teams that played in Week 2, and 220 teams that played in Week 1. This double counts the teams that played both weeks though (but I don’t believe there were many).
Hmmm… That’s a big error! I must have missed some events. :o
Week 2 I didn’t include NYC, as they still had qualifying matches to play when I did this. It seems like I’d still need another week 2 event and then 2 week 1 events to make up the difference. I’m not a programming guy, so I just went down the list of events and and copied the results into Excel and then ran it through some MATLAB scripts. I’ll check back.
OPR isn’t a great thing to compare teams across regionals. Better than nothing, (it certainly is very good at indicating who the fantastic robots are) but once you get into the pack of teams it can be hard to compare.
1114 came out swinging as the first team to break 60 though, and Wildstang was just behind them in the very high 50s.
Barry,
[strike]I can do that, give me a day or two. Anecdotally, at WPI the top two alliances were around 60 (for the entire alliance combined), most were around 45, and then the lowest two were around 30. That would indicate a “rough” average of about 15. At every event I’ve looked at, the ACs and first round picks are typically all fall within the top 20 by OPR. The 2nd round picks typically go deeper into the field, but I’ve yet to see a team with negative OPR get picked (perhaps it does do a decent job at predicting penalties?). Just going off of WPI I’d say that something like 20 of the top 24 by OPR made eliminations. My numbers aren’t in front of me at the moment though, so I’ll check that too.[/strike]
At WPI (which is the event I was concentrating on) the average OPR of a robot in the elimination rounds was 15. 2 alliances had a cumulative OPR of over 60, two had an OPR of about 45 and the remainder were between 40-30, with one 25. The top 16 robots by OPR were also the Alliance Captains and first round picks. Of the 24 teams who played Saturday afternoon, 21 of them were in the top 24 by OPR. While OPR didn’t do a very good job of actually predicting match scores, the alliance with the highest OPR won every matchup.
[EDIT] Mike, does that 449 include Israel? I did not include them either as they have yet to play.[/EDIT]
[EDIT2]Revised by elimination OPR statements after checking my notes[/EDIT2]
Thanks Kevin, I was missing the Michigan Events! :o
This new graph should include all week 2 events, including the New York City Regional and the Waterford District.
I now show 444 teams competed at week 2, which is still not quite exactly 449. The actual mean week 2 OPR was 8.35, and the Median OPR was 4.9. The 1st quartile was .407 points, and the third quartile was 12.87 points.
Updating the week 1 statistics, we find that actual mean for week 1 OPR was 8.214, and the median was 4.61. The first quartile was .413, the third quartile was 13.815. I show 298 teams competed, which is again slightly off from 301.
