Edit: Uploaded new document for Satuday night. Latest stats in last post
With the first set of data coming in from the FMS, I have put together a spreadsheet using TheBlueAlliance API and Linear Least Squares regression to calculate the following component OPRs:
Cargoship Hatch (side)
Cargoship Hatch (front)
Cargoship cargo (side)
Cargoship cargo (front)
Rocket Hatch (top)
Rocket Hatch (mid)
Rocket Hatch (low)
Rocket Cargo (top)
Rocket Cargo (mid)
Rocket Cargo (low)
Starting position (not actually OPR but an average because FMS tracks which robots start where)
Climbing bonus (also an average because FMS tracks this on a robot basis as well)
Attached is my spreadsheet which has the stats combined for week 1, as well as by regional/district
I also created an ‘OPR’ column which is the sum of the component OPRs and the 2 averages mentioned above. This ‘OPR’ column does not match traditional OPR as it omits penalties and has more accurate numbers involved for starting position and climbing.
Let me know if y’all have any thoughts! Tried to get this out as soon as possible so teams could use it for their scouting if they wanted. I will re-run the script tomorrow night for end-of-week 1 data as well.
I don’t know where you found your number, but the opr and stats for mtl is inaccurate. 4947 and 4930 should have lower opr and overall stats, and the rest of the top 10 should have higher opr. Or the opr is just not viable to rank robot at QCMO But it’s nice to have those stats thx !
Week 1 means each team is trying to figure out how to drive and score with their robot. Variations are high which means that the LLSR calculation has a high error. Later weeks or later in the tournament will yield lower errors and higher accuracy.
Perfect level 3 climb bots: 346, 5885, 2539
Best cargo bay cargo scorer: 6377
Best cargo bay hatch scorer: 2910
Best top rocket scorer (cargo+hatch): 4414
Because this year’s score total is (mostly) a linear combination of individual robot scoring actions, OPR should work well, as it operates under the assumption that there are no synergistic effects between teams (i.e. there are no interaction terms in the model)
With no statistical evidence to back up my claim, I think it will be very strong. With all the scoring actions being completed by one team and having constant point value. I’ll be surprised if it’s not one of the best year’s for OPR.
Actually, now that I think about it, defence could also have an uneven effect on teams that are going for the rocket RP, but I think that would have a minimal effect on their OPR.
I was waiting for Israel/Turkey events to post final week 1 component OPR stats. I will do that tonight!
As for:
I actually added a column in my most recent sheet (not posted yet) that omitted climbing for that exact reason. I wanted the ability to isolate game element scoring, then add and select a climber separately.
Penalties can have a significant effect on OPR through a tournament. One big penalty match can really do some wacky things.
It boosts the teams that were awarded the penalty points roughly by (penalty)/3/(number of matches per team).
It also penalizes all of their partners for the rest of the tournament roughly by (times partnered)* (penalty)/3/square(number of matches per team)
The best example is Ketterling qual match 20 had 42 penalties. So the three robots receiving the 42 points (1506, 6086, 4994) got an OPR boost on TBA. Anyone who was with those teams got their OPR hit. 7491 was with 1506 and 4994 in other matches. That’s part of the reason they show up at 26.673 on mray190’s list and 25.24 on TBA.
330 had 3 matches at Orange County where their alliance got double digit penalty points (qual 41, 56, 64). That’s a big reason mray190 has them at 34.271 and TBA has them at 43.14.
But it’s nice to have those stats thx !