2014 Curie Division

[Koko Ed]

Originally Posted by joelg236

OPR based predictions to make everyone scared.

Can't really see it too good.

[joelg236]

Originally Posted by Koko Ed

Can't really see it too good.

Full predictions. Note that this is based off of Max OPR and uses assist OPR as a tiebreaker.

[Thad House]

Originally Posted by joelg236

Full predictions. Note that this is based off of Max OPR and uses assist OPR as a tiebreaker.

Do you have this for the other divisions?

[joelg236]

Originally Posted by Thad House

Do you have this for the other divisions?

Unfortunately not, but I could start working on it.

[Jared]

Originally Posted by joelg236

OPR based predictions to make everyone scared.

I did the same thing, but got different results. Where did you get your OPR data? I used team 2834's for my results.

For top 8, I had
842
469
1241
1718
2013
118
125
135


Attached is my excel spreadsheet that uses team 2834's scouting database. It's interesting to see that OPR doesn't have a great correlation with the number of matches won predicted by OPR.

[Joe Ross]

Originally Posted by Jared

I did the same thing, but got different results. Where did you get your OPR data? I used team 2834's for my results.

You used World OPR, where Joel used max event OPR. Both are available in Ed Law's database.

[joelg236]

Originally Posted by Jared

I did the same thing, but got different results. Where did you get your OPR data?

Originally Posted by Joe Ross

You used World OPR, where Joel used max event OPR. Both are available in Ed Law's database.

This.

Originally Posted by Jared

Attached is my excel spreadsheet that uses team 2834's scouting database. It's interesting to see that OPR doesn't have a great correlation with the number of matches won predicted by OPR.

I thought your graph was interesting, here is what mine looks like with max event OPR.

[magnets]

Originally Posted by Jared

I did the same thing, but got different results. Where did you get your OPR data? I used team 2834's for my results.

For top 8, I had
842
469
1241
1718
2013
118
125
135


Attached is my excel spreadsheet that uses team 2834's scouting database. It's interesting to see that OPR doesn't have a great correlation with the number of matches won predicted by OPR.

The graph is interesting. I never realized how much scheduling luck can help/hurt somebody. There is a team that won 4 matches that has a higher OPR than a team that won 9 matches. If you're basing these predicted wins purely off of opr, I would have expected different results.

Even more interesting, is team 3562, who has an opr of 31.8, and has 7 wins, predicted by OPR. Poor team 191 has an OPR of 30, and they only have 1 win predicted by OPR. You are missing team 2073 in your OPR, and have set their OPR to 35, but changing it to their real value does not make a difference in their win/loss.

[Koko Ed]

Originally Posted by magnets

The graph is interesting. I never realized how much scheduling luck can help/hurt somebody. There is a team that won 4 matches that has a higher OPR than a team that won 9 matches. If you're basing these predicted wins purely off of opr, I would have expected different results.

Even more interesting, is team 3562, who has an opr of 31.8, and has 7 wins, predicted by OPR. Poor team 191 has an OPR of 30, and they only have 1 win predicted by OPR. You are missing team 2073 in your OPR, and have set their OPR to 35, but changing it to their real value does not make a difference in their win/loss.

Yeah, not so crazy about Jared's predictions but we exceeded the predictions last year so we'll see if we can do it again this year.

[magnets]

A few more interesting (and for me, pretty eye-opening) statistics- Remember, that in these predictions assume that a team always scores exactly their OPR in each match, so OPR (in these predictions) is a perfect representation of robot performance.

Team 1884 has an OPR of 1.18. On their simulated performance, based solely on OPR and the matches they will play in, they will win 5 matches. Team 447 has an OPR of 74.14. On their simulated performance, they will win only 4 matches.

The average OPR of teams that win 5 matches is lower than that of teams that win 4.

There is little to no correlation between a teams ability to score and their ability to seed in the top 8.

1/2 of the top 8 scoring robots will not end up in the top 8. The 29th best robot places 5th, the third best robot places 10th, and the 5th best places 17th.

[Koko Ed]

Originally Posted by magnets

A few more interesting (and for me, pretty eye-opening) statistics- Remember, that in these predictions assume that a team always scores exactly their OPR in each match, so OPR (in these predictions) is a perfect representation of robot performance.

Team 1884 has an OPR of 1.18. On their simulated performance, based solely on OPR and the matches they will play in, they will win 5 matches. Team 447 has an OPR of 74.14. On their simulated performance, they will win only 4 matches.

The average OPR of teams that win 5 matches is lower than that of teams that win 4.

There is little to no correlation between a teams ability to score and their ability to seed in the top 8.

1/2 of the top 8 scoring robots will not end up in the top 8. The 29th best robot places 5th, the third best robot places 10th, and the 5th best places 17th.

I remember a match last year where it was predicted we were going to get annihilated and one of our partners didn't bother to show up to boot and we won bigtime (partially due to one of the robots falling over). The predictions are just speculation. Not fact.

[Robotmmm]

I don't think there is any way to account for it, but remember that teams in this division who were at the Mount Olive District played matches that were 20 seconds short. This will have a negative impact on their OPR.

[Gregor]

Originally Posted by Robotmmm

I don't think there is any way to account for it, but remember that teams in this division who were at the Mount Olive District played matches that were 20 seconds short. This will have a negative impact on their OPR.

Since these predictions are using Max OPR (best OPR from a single event), it shouldn't affect too many teams, since Mount Olive was a week one. Most teams play better at their later events in the season.

It would only affect teams whose Max OPR was from Mount Olive, or would have been from Mount Olive had the matches been the correct length.

[XaulZan11]

Originally Posted by magnets

There is little to no correlation between a teams ability to score and their ability to seed in the top 8.

You mean there is no correlation between a teams' OPR and their ability to seed in the top 8. *insert all the limitations of OPR*.


(There are far too many people concerned with OPR and spending far too much time analyzing, adjusting OPR to show various things and make predictions. OPR is good for a very quick reference when you have no other information about a team. If you are spending more than 15 minutes on OPR, I suggest you actually watch matches of teams. You will get a far better picture of who is good and who isn't. Better yet, you will find what their strengths/weakness are.)

[magnets]

Originally Posted by XaulZan11

You mean there is no correlation between a teams' OPR and their ability to seed in the top 8. *insert all the limitations of OPR*.


(There are far too many people concerned with OPR and spending far too much time analyzing, adjusting OPR to show various things and make predictions. OPR is good for a very quick reference when you have no other information about a team. If you are spending more than 15 minutes on OPR, I suggest you actually watch matches of teams. You will get a far better picture of who is good and who isn't. Better yet, you will find what their strengths/weakness are.)

I do not think that a team's OPR is a good indicator of their strength. However, in these predicted matches, it is a perfect representation of their strength because the score for each match is calculated based on the the OPR, and only the OPR of each robot.

My point has nothing to do with individual teams, but it has to do with the schedule and number of matches. If you go out, and you score exactly 30 points per match, you may win one match, or you may win seven, depending on your schedule. That's what the data shows. I'm trying to prove how much of a difference your schedule can make. In these predictions, OPR is a perfect representation of a robots performance. In the real world, that is not true. I would generally expect that robots that score more points will rank higher, but it is not true.