OPR after Week Four Events

[efoote868]

Originally Posted by animenerdjohn

Is it more likely to be over the expected result or under the expected result.

Averages out to zero, but that's to be expected.

[dodar]

Originally Posted by animenerdjohn

Is it more likely to be over the expected result or under the expected result.

I'd say that the OPR match results are more often than not higher than the actual scores.

[Ether]

Originally Posted by dodar

I'd say that the OPR match results are more often than not higher than the actual scores.

3,713 (48.4%) of the actual alliance scores were greater than the OPR "expected" alliance score. 3,953 (51.6%) were less.

But the sum of all the actual alliance scores is equal to the sum of all the OPR "expected" alliance scores.

[Ether]

Here's a spreadsheet in a format which may be somewhat easier to use for various statistical analysis endeavors.

Column Headings & meaning:

E	event
M	Match
r1	r2	r3	b1	b2	b3	red & blue alliance teams
rs	red score
bs	blue score
ors	OPR "expected" red alliance score
obs	OPR "expected" blue alliance score
drs	rs-ors
dbs	bs-obs
rgo	rs>ors?
bgo	bs>obs?
rw	red win?
orw	red win based on OPR scores?
ocp	OPR correct prediction of match outcome?

[Ether]

Here's a spreadsheet using CCWM instead of OPR.

Column Headings & meaning:

E	event
M	Match
r1	r2	r3	b1	b2	b3	red & blue alliance teams
rs	red score
bs	blue score
crs	sum of red alliance teams' CCWMs
cbs	sum of blu alliance teams' CCWMs
rw	red win?
crw	crs>cbs?
ocp	rw==crw?

[markmcgary]

Originally Posted by Ether

Here's a spreadsheet using CCWM instead of OPR.

Column Headings & meaning:

E	event
M	Match
r1	r2	r3	b1	b2	b3	red & blue alliance teams
rs	red score
bs	blue score
crs	sum of red alliance teams' CCWMs
cbs	sum of blu alliance teams' CCWMs
rw	red win?
crw	crs>cbs?
ocp	rw==crw?

Correctly predicting 83.3% of the outcomes. (I'm pretty sure.)

[Ether]

Originally Posted by markmcgary

Correctly predicting 83.3% of the outcomes. (I'm pretty sure.)

Yes, I got the same number.

[Siri]

Originally Posted by Ether

Here's a spreadsheet in a format which may be somewhat easier to use for various statistical analysis endeavors.

Thanks, Ether. Some other interesting phenomena:
- For matches predicted incorrectly, the median actual margin was 10 and the median OPR margin was 9. This is compared to the overall median margin of 27 and OPR margin of 23. (21% and 22% of all margins are in these upper bounds, respectively.) Pretty impressive!

- The overall median score error (note everything here is in absolute values) is 26%.
- The winning score error is 23%, while the losing is 32%. (For matches predicted correctly, the former is 21% while the latter is the same.)
- OPR underpredicts the winning score 60% of the time, and overpredicts the losing score 63% of the time.

I wonder if there's some way to identify good defense based on the over and under-prediction trends.

[Ether]

Originally Posted by efoote868

Can't find an easy way to say it, but:
Predicted OPR match results match actual match results 82% of the time (3142 out of 3833).

Interesting observation, and I got the same answer you did.

In 82% of the matches, the outcome of the match (win or lose) was the same according to the actual score and the OPR "expected" score.

[JohnSchneider]

Originally Posted by Ether

Interesting observation, and I got the same answer you did.

In 82% of the matches, the outcome of the match (win or lose) was the same according to the actual score and the OPR "expected" score.

Id love to see some confidence intervals on this. With a .95 CI it might be neat to try and apply it to matches this weekend before they happen and see to what accuracy an event could be predicted.

[Ether]

Here's a sneak peek at Week5 OPR World Rankings, while we're waiting for Ed to post the update to his amazing spreadsheet.

[EIT2]

Thanks Ed for helping Code Red & The Mavericks to Buckeye Regional Championship!

Been a big fan of your work for a long time!

[Jared Russell]

Originally Posted by Ether

FWIW, there seems to be a problem with the official FRC data from Mt Olive.

For teams 2577 and 3929, the Auto+Climb+TeleOp data in the Team Standings page does not equal the sum of the scores of the alliances those teams were on, as shown in the Qual Match Results page.

2577 was DQ'd in match 7, and 3929 was DQ'd in match 26 (Unfortunately neither the frcfms twitter nor the FIRST event pages log in which match DQ's happen, so this data is directly from the scorekeeper)

[Ed Law]

Sorry, while I was rushing to publish the data, I forgot to sort the World OPR ranking. The rank number in column P is correct. It is just that I usually sort all the teams by OPR rank before I publish. So ignore column A.

I am not going to re-publish the data. You can sort it to how you want it.

[DELurker]

Ed,

Do you have a more up-to-date source for the photos? The link through the blue alliance page that is listed in the instructions does not appear to be functional for this year.