Quote:
Originally Posted by Caleb Sykes
Do you calculate OPRs differently than Ed Law does? According to his spreadsheet, 1023 had an OPR of 34.3 at MISouthfield.
|
Yes we do actually. The biggest difference is that we decided including Coop scores in the OPR calculation would be a misleading measure of a robots value in elims (since a robot that can get 40 points through coop in a certain amount of time is very likely not to be able to score 40 points in elims with that same amount of time through conventional stacking).
Our solution was to calculate sub-oprs and total them (without coop) to get a "special" OPR value which we felt was more useful (and also not readily available in other places). Now, it's definitely not perfect as it penalizes some teams who spend time on coop they could spend scoring conventionally, but this problem is only significant for teams that are exceptionally fast at stacking and scoring (like 2056). For run-of-the-mill teams, it keeps coop from inaccurately inflating their OPR. Of course, the difference in the math for what we do vs what Ed Law does makes our values slightly different inherently, but they both run on the same foundation of linear least squares estimation that OPR is generally associated with. Adding coop sub-opr to our "special" OPR gets roughly comparable numbers. We ended up liking our numbers a lot more over say thebluealliance though, so we ran with it.
If you're interested, head over to
http://team955.org/opr. We also have a global rankings page based on our numbers
here.