Week--2 All regionals OPR
 

All week 1 and 2 regional. Pdf and excel file attached.

Enjoy!!!

Re: Week--2 All regionals OPR
 

Youre doing a great service to many of us with this, thank you.

Im a bit disheartened that i had to scroll to page 3 to find my team. Oh well, we can only keep getting better!

Re: Week--2 All regionals OPR
 

If you went to balance on the coopertition bridge your opr will be lower.

As my team went almost every match for balancing on the coopertition bridge. Our opr is low.

Re: Week--2 All regionals OPR
 

How is the OPR calculated this year?

Re: Week--2 All regionals OPR
 

.WOW got a lot of catching up only on14 page week 2 2849
OPR 2.73027

Re: Week--2 All regionals OPR
 

wow, you're right. Thats odd, what makes it do that?

Re: Week--2 All regionals OPR
 

Because OPR only measures match points scored, it doesn't take the Co-Op bridge into calculation.

Re: Week--2 All regionals OPR
 

So OPR is how many points you scored?

Re: Week--2 All regionals OPR
 

We balanced on the coopertition bridge about 4 times in 10 matches and are ranked 10th in OPR.

Guess if you work out a successful plan with your opponents to balance it quickly near the match end, that leaves in more time for OPR-building goodness. :)

Re: Week--2 All regionals OPR
 

I don't think I understand. Coop bridge doesn't provide points anyway, so why would a balance on that bridge lower the number of points your robot is expected to make each match?

Re: Week--2 All regionals OPR
 

Wow. #10, ahead of 1114.
There must be some kind of mistake.:p

Re: Week--2 All regionals OPR
 

Clearly.

Re: Week--2 All regionals OPR
 

I'm fairly convinced that 2056 is superior to 1114 this year.

I'm further convinced that had 188/610/3360 not dumped 3360 off the back of the bridge in both finals matches that they could have beaten 1114/2056/1219.

Re: Week--2 All regionals OPR
 

It appears you are using data from "http://www2.usfirst.org/2012comp/events" to do the computation. If so, how are you calculating OPR from that data? i.e. what are you including in OPR?

Also, are you processing week1 and week2 separately, or are you combining week1 & week2 data?

Re: Week--2 All regionals OPR
 

great job with the stats, love seeing our team in the 19 spot after week two hopefully a sign of good things to come

Re: Week--2 All regionals OPR
 

Using oprnet to calculate opr.

Its week 1 and week 2 regionals.

Re: Week--2 All regionals OPR
 

Ouch. Page 11. I guess we got our act together and our robot in working order a little late. We should be better at St.Louis, after a little more improvement.

Re: Week--2 All regionals OPR
 

I believe Stundt1 is speaking of the opportunity cost of balancing those co-op bridges is not accounted for. Meaning if you spent a lot of time working on that while you could have been making baskets it doesnt take that into account.

Re: Week--2 All regionals OPR
 

Yes it takes to in account if you say double balanced but not coopertition balanced. Thats why its off a lot this year.

Re: Week--2 All regionals OPR
 

That's why I find this OPR really odd. I mean, I can get why 1126 is low, because like your team had scores of 0, 3, 12 and so on, you also had a 50 with 340, and the matches you had with us were mostly in the high 30s and 40s, where your robot scored the majority of those points.

It's a system that's over my head is all.

Re: Week--2 All regionals OPR
 

There appear to be many anomalies in the analysis. See attachment.

Some teams are listed twice, with different OPRs, for the same week.

Several teams are listed 3 times with different OPRs.

Re: Week--2 All regionals OPR
 

Ill fix those tommrow. Thanks for pointing out sorry for the mistakes.

Re: Week--2 All regionals OPR
 

Raw data is available to compute separate power ratings for Hybrid, Bridge, TeleOp, and Foul Points*.

Would that be of interest to anyone ?


*Foul Points would be computed for the Team committing the foul.

Re: Week--2 All regionals OPR
 

Ether I can add hybrid teleop and bridge balancing rankings tommrow, if you wish?

Re: Week--2 All regionals OPR
 

Yes! Raw data would be awesome!

Re: Week--2 All regionals OPR
 

Basically, yes... OPR (Offensive Power Rating) is an approximation at how many points your team scores in the average match. Since the co-op bridge doesn't provide any points, it's not included, even though it's a testament to the value of your robot.

OPR is calculated the same way every year... I am a little fuzzy on the details, but basically OPR is found using Matrix & Linear Algebra to quickly solve a complex set of simultaneous equations that yields teams' average contribution in points.

I believe two equations exist for each match, one for the red alliance (A + B + C = X) and one for the blue alliance (D + E + F = Y)... in which A, B, C, D, E, and F are the teams on each alliance and X and Y are the match scores. If A and D competed with each other in a later match, two more equations might exist: A + D + G = R and H + I + J = S. I believe these equations are then approximated incrementally until values are determined for each team.

That's all a combination of speculation and secondhand information, but regardless, each team is treated as a constant for an entire competition. If you start out very poorly, then become very good, there's no telling what OPR you'll end up with!

OPR would work perfectly if all robots performed perfectly consistently... If Robot A always scored 26 points, Robot B always scored 2, Robot C always scored 16, etc.... The primary reason why OPR is flawed is because it has to approximate the average number of points robots score, and that obviously is non-constant! Teams like 1126 at FLR this year (who improve over the course of an event) are generally underestimated by OPR because their offensive contribution changes so much.

Additionally, most OPR calculations (OPRnet included) don't include elimination matches... I think partially because the alliances are the same for each match, making them far less useful in terms of calculating OPR. Again, this is speculation, not the result of comparing OPR calculations of the same event that exclude and include elim matches.

For any seeking to learn a bit more about OPR, I recommend Ed Law's explanation for how it works... His statistic CCWM is calculated the same way as OPR, but looks at the winning (or losing) margin instead of alliance score. See: http://www.chiefdelphi.com/media/papers/2174?


Nathan

Re: Week--2 All regionals OPR
 

i like how the winning alliance at Oregon isn't on the first or second page yet several other teams at Oregon are much closer to the top.

seems accurate.

Re: Week--2 All regionals OPR
 

I don't understand where the error came from if the OP used oprnet to calculate the OPR.

This has been discussed before but for those that didn't catch it, I will repeat it again. We should not compare OPRs across different regionals/district events. It looks like that is what the OP did. The reason is some events are stronger than others and it is not a fair comparison. The better way to do it is what I called True World Ranking where every match including elimination rounds of every regional/district are all assembled into a giant matrix so all interactions between teams are taken into account. It is not perfect but it will give each team one OPR number that is a reflection of their whole season so far.

This year I also calculated the OPR for hybrid, bridge, teleop and cooperition separately using the same method. I publish all these data after each week in the following CD link if anyone is interested.

http://www.chiefdelphi.com/media/papers/2174

Re: Week--2 All regionals OPR
 

I pasted the week 1 events wrong ED. Thats where the error came from.

Re: Week--2 All regionals OPR
 

The CP bridge is included in OPRNet this year - I treat it as 5 points per CP point that your alliance earns, since there is an opportunity cost aspect to it. A balanced CP bridge could have been a single robot on a balanced alliance bridge.

Same with the separate 'bridge' calculation - since your robot's bridge skills are clearly related to your ability to balance the CP bridge.

I'm not sure if I still think it's a good idea including it in the overall OPR calculation, but it's too late for me to take it back out now.

Re: Week--2 All regionals OPR
 

You forgot to include the Israeli teams in it...

Re: Week--2 All regionals OPR
 

Yes, that was done in the week 1 OPR thread and it was really cool.

Re: Week--2 All regionals OPR
 

I should have been clearer. When I said "TeleOp" I meant a TeleOp hoop power rating. Be aware that the "TeleOp" in the data you are using is actually TeleOp hoop plus Foul Points.

Re: Week--2 All regionals OPR
 

What version of OPRNet did you use to calculate these?

Re: Week--2 All regionals OPR
 

Mathematically speaking, it's an exact (and unique) solution to a least-squares problem.

Good description, but replace the phrase "match scores" with whatever alliance metric is being analyzed.

No iteration is necessary since the set of simultaneous equations is linear in the parameters. It can be solved directly without iteration by forming the normal equations for the system and then using matrix methods (this is how all the solvers I have seen here on CD do it). It can also be solved by using matrix methods directly on the design matrix without forming the normal equations. This latter method is numerically more stable and provides greater precision, but it is more computation-intensive and is overkill for this application.

For those interested in the math, a short explanation of the matrix math required to solve the normal equations is given here.

Re: Week--2 All regionals OPR
 

Here is the raw match-by-match data for weeks 1 and 2 in an Excel spreadsheet.

This includes qualification, elimination, and practice data for all regional and district events which have taken place so far this season, with the exception of Oregon and Traverse City.

It includes Bridge, Foul, Hybrid, TeleOp hoops, and Coopertition data for each match.

Re: Week--2 All regionals OPR
 

Could you do it like it is in the week 1 thread? It would make it easier to read for each team.

Re: Week--2 All regionals OPR
 

The thing about Oregon is that the teams that won, won because they had a good strategy, which was also a strategy that none of them implemented before the finals. Together they pulled off triple balance after triple balance, but when they played on there own during the qualification matches, they had significantly lower scores (on average).

Basically, they have a lot of low scores to balance out the high ones they got during eliminations, while other teams in the finals (1983 specifically) were high scorers throughout the entire event.

Re: Week--2 All regionals OPR
 

I decided not to assign points for Coopertition points and just calculate OPR for Coopertition bridge as is. So that number represents number of Coopertition points per match. Teams can assign their own point value to it and add to the OPR of bridge points if they like. My philosophy has always been encouraging teams to customize the spreadsheet for their own need so I don't decide how they should modify the data.

Our first hand experience competing at an event this year shows that it is easier to balance with your own alliance partner than with someone who is over 50 feet away from you. This is the reason I want to look at balancing alliance bridge and coopertition bridge separately.

Re: Week--2 All regionals OPR
 

Thanks for clarifying where I was misleading - I know my knowledge of OPR calculations is somewhat lacking... I was hoping I could clear up some confusion and initiate someone like you, Bongle, or Ed Law clearing up where I was confused.

Thanks for all you guys bring to those of us who use and appreciate the stats but lack the software and statistical experience to generate them.


Nathan

Re: Week--2 All regionals OPR
 

Interesting numbers! I think graphs always help data make a lot more sense. I accidentally left my laptop charger at a friends, so I don't have access to my best friend Stata right now. However, Excel works in a pinch.

The first graph essentially plots the distribution of OPRs since they are ranked in the data file. You can see that there is a very sharp drop off among the truly excellent teams, a wide flat region, and then a drop off on the tail end. If I had to bet, I would say the teams that are ranked extremely poorly probably played mostly with excellent teams in bad matches, and to make the math work out ended up with negative numbers. Since penalties do not subtract from scores this year, you'd have to be intentionally awful to interfere with your partners enough to really hurt their OPR. Also, there were 1009 points in this data set, so going by 100s you can basically break OPR up into percentiles.



The next graph shows OPR versus team number (which is a fairly decent approximation for age). With an R-square of .076, it obviously doesn't explain all of the variation, but I think it is interesting to see the general trends (and watch 2056 absolutely decimate one of them).



Finally, the last one to me is the most interesting. It plots the distribution of Week 1 and Week 2 OPRs, and is normalized by the number of teams since significantly more teams played Week 1 than Week 2. I apologize for the smallness of the graph, but this computer is ancient and apparently asking Powerpoint to resize this was asking way too much as the computer crashed a couple of times just getting to this point. As you can see, the top tier teams didn't change much, and the teams that didn't score much stayed essentially the same. However, the middle of the road teams to upper middle of the road got a couple of points better. I would think this is exactly what we'd hope to see (although more lifting on the extreme low end would be nice too), it should provide more exciting balanced matches.


(one you can blow up more)

Re: Week--2 All regionals OPR
 

Has anyone run the hybrid, teleop and bridge power ratings for Week 2? (as in Week 1) Also, I hope it doesn't make too big an impact, but do the teleop ratings include foul points? As I understand it, the TPs FIRST reports in the standings include fouls, but the Twitter feed breaks them out.

Re: Week--2 All regionals OPR
 

For OPRNet (which is also the source of the below data), the TeleOp and OPR scores both include foul points accrued. It'd take too much time to parse the twitter feeds to figure out the foul points.

2056's teleop: Wow.

Edit: Initially forgot the district events. Fixed.

Re: Week--2 All regionals OPR
 

I'm always forgetting the district events because they're at the bottom of FIRST's regional listing... just a sec.

Re: Week--2 All regionals OPR
 

You changed Hybrid to Bridge. Now there is bridge twice.

Re: Week--2 All regionals OPR
 

Bongle,

I checked my data with yours and I am glad to see that they are the same. I just want to clarify for everyone on CD who uses our data why it does not seem to match.

In my spreadsheet that is found at http://www.chiefdelphi.com/media/papers/2174, I report on the OPRs for each event separately in its own tab. In column O is the overall OPR. Column AA to AD are the OPRs for hybrid, bridge, teleop and coop respectively. From what I understand, the foul points are included in the overall OPR. To calculate OPR for foul points, it would be

OPR foul = overall OPR - OPR hybrid - OPR bridge - OPR teleop

Based on my understand, you are giving 5 bridge points and 5 overall OPR points for each Coop point. Is that correct? If it is, then

Adjusted overall OPR = overall OPR + 5 X OPR coop
Adjusted bridge = OPR bridge + 5 X OPR coop

When I adjust my data this way, it matches yours based on spot checking. However I am confused with your statement "the TeleOp and OPR scores both include foul points accrued". Did you adjust the OPR TeleOp points also and add the OPR Foul Points to it? From my understanding, the cummulative TeleOp points that FIRST publish do not include the foul points in it, and I believe we are using the same numbers to calculate OPR components. I would appreciate it if you can clarify. Thanks.

Re: Week--2 All regionals OPR
 

You are correct about the bridge points.

TeleOp and overall OPR include fouls, because there is no easy way to separate them out. It's not by any concious addition by OPRNet. There aren't any adjustments made for fouls in OPRNet.

I have fixed my previous post again.

Re: Week--2 All regionals OPR
 

Sweet, 9th best team in the world. Sandwhiched between 118 and 148. :D

Re: Week--2 All regionals OPR
 

Thanks for the data! But is it me, or is it still missing the Rutgers Mid-Atlantic District?

Re: Week--2 All regionals OPR
 

Based on "http://www.usfirst.org/roboticsprograms/frc/competition-manual-and-related-documents" where FIRST describe the FMS Twitter Feed Information, they report Foul points, TeleOp points, Hybrid points and Bridge points separately. That is why I thought when they report the cummulative TeleOp points in the TP column of Standings page (e.g. http://www2.usfirst.org/2012comp/Eve...rankings.html), they are just adding up the TeleOp points from each match.

Ether, can you weigh in on this since you have the twitter data for every match?

Another reason I think this way is OPR bridge + OPR hybrid + OPR teleop does not equal to overall OPR in most cases which it should mathematically if what Bongle said is right about OPR teleop and overall OPR both includes foul points.

Re: Week--2 All regionals OPR
 

Re: Week--2 All regionals OPR
 

Ok, I looked at intermediate results in OPRNet.
To pick on 1114, because that's a theme for the week:
Total alliance points: 411 (summing up their alliance scores in every match, which would clearly include fouls)
TP: 191
BP: 60
HP: 170
Total alliance points from rankings page: 411

So the rankings page appears to be reporting TP as hoops + fouls, not just hoops. Unless 1114's opponents took no penalties all weekend, which is probably a bit unlikely.

Another comparo for 2056:
Sum of alliance scores: 502
BP: 40
HP: 178
TP: 284
BP + HP + TP: 502

Re: Week--2 All regionals OPR
 

Thanks Bongle for looking into it. That's what Joe Ross was telling me also.

In terms of assigning 5 bridge points for each coop point, one can argue that it is harder to balance on the coop bridge when your partner is more than 50 feet away and therefore deserves more points. The opposing argument is for alliance selection, teams are not looking for alliance partners who can balance on coopertition bridge so 5 points would be right.

The reason why I report hybrid, bridge, teleop and coop as is is because I use these numbers during the competition and not only afterwards. I want to know who is the strongest teams to go to the coop bridge etc.

However when I publish results from now on, in order to be consistent as a community, I will report the adjusted overall OPR and adjusted bridge score in additional columns in Excel the same way that Bongle advocated. I agree with how it is done. If somebody using my spreadsheet wants to change the number of points to be added, they will be able to easily change the number 5to something else.

I have a question for the community. When I publish the True World Ranking from now on, which one should I use? Should I use the original or adjusted overall OPR? For those of you who have not followed the True World Ranking, it is calculated by putting in every match of every regional/district including elimination rounds and every team who has played into a giant matrix. This way all interactions between teams will be accounted for and there is no uncertainty about comparing the OPR from one event to another because one event is stronger. Lumping all the OPRs calculated from separated events together and sorting them is not the best way to do it.

Re: Week--2 All regionals OPR
 

Ed, sorry I missed out on this discussion. My internet feed was down all afternoon and evening.

Here's a link to all the match data to date. It includes Practice, Qual, and Elim data for all matches at all week1 & week2 events except Oregon and Traverse City.

There are no macros in the spreadsheet to worry about, and no formulas. Just data. You're pretty adept at Excel so you can convince yourself pretty quickly about the TeleOp hoops and Foul Points question.

Note: I found 9 entries where the scores did not add up correctly. Those may be due to late fouls that were not entered into the Twitter data stream.

Re: Week--2 All regionals OPR
 

Thanks Ether for the data, I will take a look at it later tonight.

I just thought of one other point to consider about which OPR to use for World Ranking.

Since I am including elimination matches, and elimination matches have no coop bridge but you can balance with 3 robots and get 40 points. Giving 10 points to balance coop bridge during qualifying rounds is a good idea. Perhaps we should not include elimination round this year since the potential number of points are different.

Just some more things to think about.

Re: Week--2 All regionals OPR
 

I still think you should consider the elims because teams play to the co-op for quals but play to the hoops in elims so it will show who is truly the best.

Re: Week--2 All regionals OPR
 

Fixed once again.

Re: Week--2 All regionals OPR
 

Ether,

Thanks for the data. The twitter data is worse than you thought. It is not just the Oregon and Traverse City data missing. If you look at individual events, you would find that it is missing a lot of matches (e.g. FL match 105, GG match 1-8, KC match 42-61). This is not your fault but I don't think we can use this data at all. However I did confirm that the teleop points FIRST reported in standings page include the foul points.

Re: Week--2 All regionals OPR
 

YIKES! Nice catch. I had not gotten around yet to examining it for random missing matches. Now that I have, it's even worse than that. I counted 207 missing Qual and Elim matches.

I agree: for OPR it is worthless. It still has some use for locating stats like highest scores etc but due to the missing data cannot be considered definitive.

Re: Week--2 All regionals OPR
 

Sorry to intrude again, but another most interesting plot. Plotted on the same graph is the OPR distribution for bridge, hybrid, and teleop. The way OPR is calculated (from lumped sums), it stands to reason that we should expect to see similar distribution shapes for any OPR statistic be bridge, hybrid or teleop. This is what we see.



Like one might expect, the elite tier teams dominated in Teleop. It is a little surprising to me that they do not also dominate in hybrid. Perhaps if we see some autonomous bridge jousts for balls this shape will change. In the middle of the road teams, it is clear that bridge balancing is worth more points than teleop and hybrid, although not by a terribly large margin. I also find the fact that the teleop and bridge balancing curve cross again near zero -- I wonder if this is a sign that teams with inexperienced drivers make it hard for their partners to balance, versus just being unable to score balls by themselves.

On a final note, notice all three lines are quite close. The game has a beginning, middle, and end and they are all close to equally weighted. I wonder if other FRC games that are regarded as good and fun to watch share this characteristic?