List(s) of OPR?

[microbuns]

Does anyone know of a place to quickly access (live) OPRs at a match? TBA does part of the job (top 15), but is there anywhere that does the same thing but with more than top 15?

[Dunngeon]

Originally Posted by microbuns

Does anyone know of a place to quickly access (live) OPRs at a match? TBA does part of the job (top 15), but is there anywhere that does the same thing but with more than top 15?

http://team955.org/apps/opr/

It won't be super accurate until more matches have been played than teams. This is true with all OPR calculations though.

[Ether]

Originally Posted by Dunngeon

http://team955.org/apps/opr/

How are you handling surrogates in the computation of OPRs using data from the Team Rankings table?

[Dunngeon]

Originally Posted by Ether

How are you handling surrogates in the computation of OPRs using data from the Team Rankings table?

Quite simply, we don't.

I've noticed you asking about surrogates and DQ's on every OPR thread, why?

[Ether]

Originally Posted by Dunngeon

Quite simply, we don't.

I've noticed you asking about surrogates and DQ's on every OPR thread, why?

Because I'm interested in how various OPR app developers are handling that question.

In a match with an alliance containing a surrogate team, the scores (Coop, Auto, RC, Tote, Litter) for that alliance in that match are not included in that surrogate team's totals as reported in the Rankings table. So without some algorithmic legerdemain, the computed category OPRs for every team that plays in an event as a surrogate will be lower than they "should" be.

Unfortunately, there is really no mathematically "correct" way to deal with this, because this year FIRST does not provide the data required to handle this issue properly.

[Ed Law]

Originally Posted by Dunngeon

Quite simply, we don't.

I've noticed you asking about surrogates and DQ's on every OPR thread, why?

I am glad that everybody is not doing anything special to calculate OPR. This means it is just another match and one more data point. This is how it was done starting from 2006. I am doing it the same way also. I don't think anybody should change that.

What Ether is asking everybody about has to do with calculating Component OPR only. I already explained that in my post and how I did it in case anybody wanted to duplicate that in their own program. It is very important to me that as a community we all calculate it the same way.

[Ether]

Originally Posted by Ed Law

What Ether is asking everybody about has to do with calculating Component OPR only.

Correct. And the reason I asked Dunngeon is because Team 955's (excellent) online OPR calculator does compute component OPRs.

To see what I am talking about, go here and look through the list of links at the top of the page to find the one described as follows:

ZIP file containing 47 CSV files (one for each event in weeks 1 thru 3).
CSV can be opened directly in Excel (if file association is set).
Each CSV contains all Match Results and Team Standings data
in columns. The last 3 columns in each CSV illustrate what happens when
there are DQs or surrogates.

... and click the link underneath that description to download it.

Open the ZIP file and look at the FLFO event.

Originally Posted by Ed Law

It is very important to me that as a community we all calculate it the same way.

I don't think that's going to happen until FIRST provides the data necessary to do these computations in a straightforward manner (requiring no special work-arounds for missing data).

[microbuns]

Originally Posted by Dunngeon

http://team955.org/apps/opr/

It won't be super accurate until more matches have been played than teams. This is true with all OPR calculations though.

This looks like the kind of thing I'm looking for! However the numbers look a bit low - is it calculating OPRs with a score of 0 for all future matches?

[Dunngeon]

Originally Posted by microbuns

This looks like the kind of thing I'm looking for! However the numbers look a bit low - is it calculating OPRs with a score of 0 for all future matches?

The numbers won't be correct until there are more matches than teams.

We aren't sure, but it seems that OPR is still somewhat correct for in-progress events with less matches than teams. Obviously, the numbers aren't right but they appear to be representative. For example, Waterloo shows 1114 with a higher OPR than 2056 which makes sense given their prior performances. The remaining teams flesh out pretty much where I would expect them too as well.

[Caleb Sykes]

Originally Posted by Dunngeon

The numbers won't be correct until there are more matches than teams.

Can you explain what you mean by this?

[microbuns]

Originally Posted by Dunngeon

The numbers won't be correct until there are more matches than teams.

We aren't sure, but it seems that OPR is still somewhat correct for in-progress events with less matches than teams. Obviously, the numbers aren't right but they appear to be representative. For example, Waterloo shows 1114 with a higher OPR than 2056 which makes sense given their prior performances. The remaining teams flesh out pretty much where I would expect them too as well.

Hey - just checked, and it looks still pretty low. Does 2056 seriously just have a 46 OPR at Waterloo? TBA says they are >100.

[Dunngeon]

Originally Posted by microbuns

Hey - just checked, and it looks still pretty low. Does 2056 seriously just have a 46 OPR at Waterloo? TBA says they are >100.

No, there's an issue in how we are calculating OPR/component OPR right now. It won't be fully accurate until the quals matches conclude. The guys who developed the app are looking into it.

The data does seem to be correct if you take it as representative. Once the issue is corrected, it should be more accurate throughout the event.

[Spoam]

Originally Posted by microbuns

Hey - just checked, and it looks still pretty low. Does 2056 seriously just have a 46 OPR at Waterloo? TBA says they are >100.

I don't think so. I think one of our matrices is pulling data from matches that haven't happened yet (from the match schedule) which is diluting the OPR values. Once quals are over the numbers will look correct. I think the values are correct relative to each other (so ranking should be correct). I'll find where the problem is and fix it today.

EDIT: Found the problem. We have a Tx2M and a 2MxT (T = team count, M = match count) matrix. The problem was that M is based on the match schedule, not the qual matches played so far (exacerbated by the fact that our score vector has "-1" for unplayed matches). Being fixed now, should be live soon. Numbers for partially completed events will now be accurate.

UPDATE 2: Everything is fixed now and it works; will be live soon. No more low numbers for incomplete events!

[Ether]

Originally Posted by microbuns

Hey - just checked, and it looks still pretty low. Does 2056 seriously just have a 46 OPR at Waterloo? TBA says they are >100.

Waterloo OPR after 42 43 44 matches:

1114 122.338771908684
2056 115.255462997957
5406 79.45256087752726
1676 65.22743639565435
3683 56.35364504850799
4039 52.26278702460058
4678 48.29159643066427
1285 44.16440103647798
4917 43.75946019638102
2935 30.76809371505734
4308 30.52351397120633
4618 29.35071188804886
1305 21.71971062966638
1241 18.99280792584136
865 18.79165128654639
244 17.84078561299089
4907 17.48879720367756
5158 15.5599957215479
1334 13.88992749860653
3161 12.39563411206913
4083 12.26564544482614
4617 11.19964476997339
2702 7.470182890478237
4943 4.825685466295547
4939 3.778595533262281
329 3.103572087743411
288 1.448370672968942
5719 -0.08024046235556673
3161 -0.6562681350470999
216 -0.7921025379218417
4807 -2.33763700274084

[microbuns]

Originally Posted by Ether

Waterloo OPR after 42 43 44 matches:

1114 122.338771908684
2056 115.255462997957
5406 79.45256087752726
1676 65.22743639565435
3683 56.35364504850799
4039 52.26278702460058
4678 48.29159643066427
1285 44.16440103647798
4917 43.75946019638102
2935 30.76809371505734
4308 30.52351397120633
4618 29.35071188804886
1305 21.71971062966638
1241 18.99280792584136
865 18.79165128654639
244 17.84078561299089
4907 17.48879720367756
5158 15.5599957215479
1334 13.88992749860653
3161 12.39563411206913
4083 12.26564544482614
4617 11.19964476997339
2702 7.470182890478237
4943 4.825685466295547
4939 3.778595533262281
329 3.103572087743411
288 1.448370672968942
5719 -0.08024046235556673
3161 -0.6562681350470999
216 -0.7921025379218417
4807 -2.33763700274084

Thanks a ton Ether - that's exactly what we need. Is there any chance you could update it one more time?