OPR vs Coopertition Strategies

[Bongle]

As a guy that wrote lots of OPR stuff last year, I'll chime in.

OPR should never be used as the only scouting mechanism. It can help point out potential diamonds in the rough, as a team with a very high OPR probably has it for a reason. It can usually predict fairly well the teams that actually get into eliminations, but not a whole lot better than simply looking at the rankings.

OPR worked very well in 2008 because 2008 was a very robot-centred game with very little defense. Since there was little active defense, it was possible to score approximately the same amount of points every time you went out, which made the algorithm work better. It worked less well in 2009 because of the immense contribution (up to 50% from some scouting estimates I've heard) of human players, and the very heavy defense in the game. A robot might score 50 points one round, then 10 the next because they were pinned the whole match. 2007 was not an OPR year because of the exponential nature of the scoring. If your presence adds 5 points through whatever mechanism to the expected score of each alliance you're on, then your OPR will be 5.

1. Does anyone see any use of OPR or DPR this year? If so, how?

I think it should still be accurate. If your robot contributes points on average to the alliance it is on (whether by feeding balls to the next zone, scoring balls, hanging each time, or even just having a good strategy), then your OPR will be higher. Since the scores are linear and human players have very little to do with it, it should be more "accurate" than 2009.

That said, I reiterate that it is nothing more than the solving of a system of equations. It is just a slightly different way of looking at the scores that teams put up to try and find robots that may have just been unlucky with their teammates or opponents. If a robot is a feeder bot and you're a scorer bot, then the feeder bot will probably be a better choice than a scorer bot with a higher OPR.


To comment on the seeding setup:
-I think only extremely confident teams will be scoring for their opponents, so it shouldn't affect things too much. The very top teams will probably have OPRs lower than theoretically possible because they'll be scoring for their opponents slightly. This will slightly raise the OPRs of low-performance teams, but not too much because the elite teams' scoring-against will be spread across all of their opponents, and will only occur in matches that are turning out to be blowouts.

[SteveGPage]

Originally Posted by Bongle

As a guy that wrote lots of OPR stuff last year, I'll chime in. ...

I figured I could get you to chime in! Thanks for the input!

Steve

[Nawaid Ladak]

i think if FIRST still records W/L/T stats, RPI (Ratings Percentage Index) might actually be a better formula to use. If your looking for teams that had tough SoS's (Strength of Schedule). Then this is something that should be looked into

after al, OPR will drastically change once the eliminations begin because of the sudden rise in defensive play.

The formula for RPI is as follows

RPI = (WP * .25) + (OWP * .50) + (OOWP * .25)

Where WP is Winning Percentage, OWP is Opponents' Winning Percentage and OOWP is Opponents' Opponents' Winning Percentage.

[Bongle]

After reading this, I've changed my mind.

Scoring for opponents will probably be quite common, given the complete shakeup of last year's rankings. A team that was 15-1-0 at the Waterloo regional last year dropped to dead last in the rankings. So that will depress top teams' OPRs since they'll spend some of their time scoring against themselves, and slightly elevate lower-performing teams' OPRs since they'll be getting some help scoring.

[s_forbes]

Originally Posted by Bongle

After reading this, I've changed my mind.

Now I'm a bit confused; this post shows that the new seeding algorithm keeps the ranking order more or less the same (using last years data). The post you linked shows a completely different result.

Which is actually correct?

[SteveGPage]

Originally Posted by s_forbes

Now I'm a bit confused; this post shows that the new seeding algorithm keeps the ranking order more or less the same (using last years data). The post you linked shows a completely different result.

Which is actually correct?

You are looking at two different sets of scores, but regardless, I think you still can see the same thing in each. Here are two cases:

Curie 2009
-- Team 668 was in 44th place. According to the new seeding, they are now in 3rd.

Newton 2009
-- Team 1629 was in 8th place. According to the new seeding, they are now in 22nd.

Those are huge changes, as how they impact the teams. If you had team that totally dominated before, 75-0, for example, that team doesn't do any better than the teams they just beat, and perhaps if there had been a penalty - they would actually do worse than the teams they just beat.

[DanDon]

Originally Posted by SteveGPage

You are looking at two different sets of scores, but regardless, I think you still can see the same thing in each. Here are two cases:

Curie 2009
-- Team 668 was in 44th place. According to the new seeding, they are now in 3rd.

Newton 2009
-- Team 1629 was in 8th place. According to the new seeding, they are now in 22nd.

Those are huge changes, as how they impact the teams. If you had team that totally dominated before, 75-0, for example, that team doesn't do any better than the teams they just beat, and perhaps if there had been a penalty - they would actually do worse than the teams they just beat.

Let's see if we can get this cleared up a bit. I believe that in Andrew's calculation he kept the seeding scores and the coopertition bonus separate, but I'm not sure. If that is indeed the case, that's where the two differ as I have the coopertition bonus added to the seeding scores in mine.

[Bomberofdoom]

I could be missing something, since I don't have much time to read the manuals (college and all), but don't ranking points COME BEFORE the coopertition points?

In games where you have a difference of 7 points, no problem over here, go get like 4 more goals to even it up, or even more if it seems possible that the opponents will not be able to turn the tide somehow from this move. You'd do that do add coopertition points to make sure you will be ranked first in you rank level (if someone else is 2nd place with you, your coopertition points will beat his, therefore you'll be above him in ranking). But if you are about 1-2 in point difference, I'd say forget about it. MAKE SURE YOU WIN, because 2 ranking/seeding points are much more important that 2 coopertition points, as it always was!

As I see it, the coopertition is just a tie-breaker for similar ranks.

Please do correct me, for I am confused of the comments in this topic.

[DanDon]

Originally Posted by Bomberofdoom

I could be missing something, since I don't have much time to read the manuals (college and all), but don't ranking points COME BEFORE the coopertition points?

In games where you have a difference of 7 points, no problem over here, go get like 4 more goals to even it up, or even more if it seems possible that the opponents will not be able to turn the tide somehow from this move. You'd do that do add coopertition points to make sure you will be ranked first in you rank level (if someone else is 2nd place with you, your coopertition points will beat his, therefore you'll be above him in ranking). But if you are about 1-2 in point difference, I'd say forget about it. MAKE SURE YOU WIN, because 2 ranking/seeding points are much more important that 2 coopertition points, as it always was!

As I see it, the coopertition is just a tie-breaker for similar ranks.

Please do correct me, for I am confused of the comments in this topic.

Nir- the coopertition bonus gets added on to your seeding score in addition to being a tie-breaker.

[SteveGPage]

Originally Posted by Bomberofdoom

I could be missing something, since I don't have much time to read the manuals (college and all), but don't ranking points COME BEFORE the coopertition points?

In games where you have a difference of 7 points, no problem over here, go get like 4 more goals to even it up, or even more if it seems possible that the opponents will not be able to turn the tide somehow from this move. You'd do that do add coopertition points to make sure you will be ranked first in you rank level (if someone else is 2nd place with you, your coopertition points will beat his, therefore you'll be above him in ranking). But if you are about 1-2 in point difference, I'd say forget about it. MAKE SURE YOU WIN, because 2 ranking/seeding points are much more important that 2 coopertition points, as it always was!

As I see it, the coopertition is just a tie-breaker for similar ranks.

Please do correct me, for I am confused of the comments in this topic.

I understand that college, and such can keep you busy, but take 5 minutes and read in the Tournament section 9.3 (especially 9.3.3 - 9.3.5). Ranking points are no longer a part of the game. There is no 2 points for a win. Totally different formula and philosophy. Please do read the manuals for this game!

[Chris is me]

Originally Posted by SteveGPage

You are looking at two different sets of scores, but regardless, I think you still can see the same thing in each. Here are two cases:

Curie 2009
-- Team 668 was in 44th place. According to the new seeding, they are now in 3rd.

Newton 2009
-- Team 1629 was in 8th place. According to the new seeding, they are now in 22nd.

Those are huge changes, as how they impact the teams. If you had team that totally dominated before, 75-0, for example, that team doesn't do any better than the teams they just beat, and perhaps if there had been a penalty - they would actually do worse than the teams they just beat.

Oddly enough, 668 was rather under-ranked on Curie, scoring 24 balls in their first match then being subject to defense all weekend long. 1629, while an elimination caliber robot roughly on par with 469, had a slightly easier schedule than many. Perhaps this system somewhat corrects for that?

I'm interested to see Andrew's data, corrected to include Coopertition Points.

[Passion]

I find this way of ranking very funny. For the first time in my life, I might be actually trying to win, but in fact I might be actually losing even more....

[DanDon]

Originally Posted by Passion

I find this way of ranking very funny. For the first time in my life, I might be actually trying to win, but in fact I might be actually losing even more....

The one benefit I can see to the new ranking rules is that I get rewarded more for beating a stronger alliance than a weaker alliance, and I get hurt less by losing to a stronger alliance than a weaker alliance.

This is all assuming a fair-played match with no collusion.

[leafy]

Originally Posted by DanDon

The one benefit I can see to the new ranking rules is that I get rewarded more for beating a stronger alliance than a weaker alliance, and I get hurt less by losing to a stronger alliance than a weaker alliance.

This is all assuming a fair-played match with no collusion.

Exactly; the new seeding score is normalized by the difference in score. In the old system, beating a team 75-10 was the same as beating them 75-74. In the new system, it rewards you for having a close score (or, if you're already high, it rewards you for scoring for the other team).