paper: Predictions of Championship Divisions by OPR

[stingray27]

Thread created automatically to discuss a document in CD-Media.

Predictions of Championship Divisions by OPR (http://www.chiefdelphi.com/media/papers/3003?) by stingray27

Working on uploading Curie and Newton. Will be up as soon as possible

[Bryce Paputa]

I (and some others) also made one of these on google docs, https://docs.google.com/spreadsheets/d/1CL6b8fOjVPt_gK-CImpa9loBYOHlK-sQEg2O870PAYI/edit?pli=1#gid=363754164. Right now it is only Archimedes, but it wouldn't be hard to do the same for other divisions.

[stingray27]

I just want to reiterate what I just stated in another post: (http://www.chiefdelphi.com/forums/showpost.php?p=1376132&postcount=138)

All of the data analysis everyone is doing, including and especially myself, is just for rough estimation of what to look for at championship.

OPR values should be used to your/your team's discretion. Personally, I think it is just interesting to look at the numbers and get a glance ahead of time what to look at for match strategy. THEN, I believe you/your team should actually go out and do the scouting yourself to see how a team can best fit into your strategy.

The reason I am doing all of this is because it is interesting to look and I find it fun to run through all of my calculation in excel.

I am by no means obsessed or stuck on OPR values and their meaning. I am just providing stats for people to look at. I am sorry if people are getting frustrated by how much content has been put out recently regarding the subject.

[SteveGPage]

All of the data analysis everyone is doing, including and especially myself, is just for rough estimation of what to look for at championship.


Quoted for truth.

A big "Thank you!" to all of those who put these numbers together! These numbers should give you a rough idea of what to look for. The real work is done by real scouting - in St. Louis!

"Winning" the rankings in OPR, CCWM, etc... is like winning the SuperBowl in the pre-season. There is a reason we still play the game!

Strategy beats OPR!

[Dan_Karol]

Mike,

I noticed there is one team expected to get a score of 108 auto points...
You may get more accurate data by capping Auto OPR at 75 points. The data still looks good though. It will be really interesting to see how accurate these predictions appear.

[stingray27]

Mike,

I noticed there is one team expected to get a score of 108 auto points...
You may get more accurate data by capping Auto OPR at 75 points. The data still looks good though. It will be really interesting to see how accurate these predictions appear.

Good call! That should make things a little more interesting!!

[hannaners]

Are all the OPRs used from the regional the bot participated in? I know our highest OPR actually came from district event; just curious!

[stingray27]

Are all the OPRs used from the regional the bot participated in? I know our highest OPR actually came from district event; just curious!

Some events are throwing errors in my database and therefore, I have to omit those entries. 90% of the teams are unaffected in the predictions though. Some of the MAR districts are giving me issues - hence why your team may be affected.

[hannaners]

Some events are throwing errors in my database and therefore, I have to omit those entries. 90% of the teams are unaffected in the predictions though. Some of the MAR districts are giving me issues - hence why your team may be affected.

That makes sense. Thank you so much for putting this all together!

[Ether]

Some events are throwing errors in my database

Would you like help finding the cause and possible solution?

Identify the problem events and post the data you are using for those events (preferably in whitespace or comma-delimited plain text format).

[themccannman]

Strategy beats OPR!

OPR isn't doesn't even truly reflect performance in the first place. There's so many factors that skew OPR and cause it to be inaccurate that it really can't be used for much more than a getting a general idea of which teams score a lot of points, which teams score some, and which teams don't score very many.

[stingray27]

OPR isn't doesn't even truly reflect performance in the first place. There's so many factors that skew OPR and cause it to be inaccurate that it really can't be used for much more than a getting a general idea of which teams score a lot of points, which teams score some, and which teams don't score very many.

I agree, hence my statement in post #3 above.....

[stingray27]

Would you like help finding the cause and possible solution?

Identify the problem events and post the data you are using for those events (preferably in whitespace or comma-delimited plain text format).




Thanks for offering me your help! After I finish studying for my exam, I will take a look at it more and get you some details. The info that I use if you wanna take a look at it ahead of time is in my 2014 Advanced Stats and OPR paper. Any regionals that are not in that excel sheet cause issues in my breakdowns of divisions