paper: 2010 OPR: Week 1 and 2 Regionals

[dodar]

How did you figure that out? just curious

[JesseK]

Not to rain on anyone's parade, but there's a comprehensive (and complex) spreadsheet here that has OPR, DPR, and resulting CCWM (calculated contribution to winning margin).

Bongle, and many others (I believe it started with Karthik?) have developed OPR over the years. A thread explaining OPR/DPR is here.

CCWM = OPR - DPR. See the above threads for implications.

Fine print: The latter two may prove useful for finding that key 2nd pick in a field of robots that aren't highly publicized, yet no score-based statistical rating system should be used by itself to evaluate a team's ability to win a competition.

[The Lucas]

Originally Posted by JesseK

Fine print: The latter two may prove useful for finding that key 2nd pick in a field of robots that aren't highly publicized, yet no score-based statistical rating system should be used by itself to evaluate a team's ability to win a competition.

Or DPR may be useful to find teams that dont understand the seeding system
IMHO Team Update 16 made OPR useful again with a 5 pt incentive to win, however DPR and CCWM (or PM) is still useless.
Perhaps we should make a new calculation for this year: Seeding Power Ranking (SPR) and just an alliance's seeding points for that match instead of the score (EDIT: obviously it is not as simple as that since you have to take opponents into account). That way we can find the diamonds in the rough unfortunate enough to have weak schedules with weak opponents and even weaker partners. Lets face it a tough schedule isnt necessarily a bad thing this year and it can carry you with seeding points without you really having to do much. But what are you contributing to the seeding pts versus just being with/against top seeding bots? It could be a relevant stat as seeding points are what matters in qualifiers.

[IKE]

What is scary is a couple of those near the top of the list were Week1 guys before update 16. Imagine an extra point or so due to the extensive penalties of week 1!

[JesseK]

Originally Posted by The Lucas

Or DPR may be useful to find teams that dont understand the seeding system
....
But what are you contributing to the seeding pts versus just being with/against top seeding bots? It could be a relevant stat as seeding points are what matters in qualifiers.

Yet what happens in qualifiers with regards to seeding doesn't indicate what can happen with a given bot on an alliance. Who in their right mind forms an alliance based upon robots that are best at deciding which alliance to score for (including the opposition)?

[The Lucas]

Originally Posted by JesseK

Yet what happens in qualifiers with regards to seeding doesn't indicate what can happen with a given bot on an alliance. Who in their right mind forms an alliance based upon robots that are best at deciding which alliance to score for (including the opposition)?

Every alliance should go in to a qualification match with a strategy to maximise their seeding pts (which post TU16 includes 5 pts for winning). Should every alliance so into a qual match looking to play defence? No, I'd say its generally not in the alliance's best interest to play more than situational defence unless one or more robots are incapable of moving balls. So why use a statistic that assumes that everyone is trying to play defence all match if defence isn't necessarily sound strategy.

If you are playing defence all match you better win and get those seeding points (successfully executing your strategy). Depending on the depth of your regional, the best 2nd picks probably are flexible and play a mix of offence and defence depending on the situation in their matches. From my observation bots that play "defence only" in qualifications either do so because that's their mentality/bot speciality (very few) or that's all they can do effectively (includes dead bots placed in front of one goal). Do you really want to blindly pick a "defence only" bot based on DPR statistics?

I actually dont think you can find a 2nd pick defender this for this year's game using DPR statistics nor will the statistics even point you in the right direction. Last year DPR was pretty good at identifying how much your trailer got scored on, which was very important to consider. Anecdotally, I looked back at DPR stats for the regional I attended (MD) and 2 of the top 3 DPR teams were dead in 2+ of their matches. The other had a defensive mentality and ended up an Alliance captain. However the best defender from my viewpoint was in the middle of the pack in both OPR/DPR. You have to look at machine/driver capabilities and project them into your elim strategy.

Can they change zones? Are they quick and manoeuvrable? Good driving instincts? Can they clear ball out of the zone? There arent statistics for these.

Keep in mind that your best defensive bots could be playing striker and scoring all through the quals. Defence and striker are very similar roles. Unless the defender is completely shutting the striker down, I'd pick the striker over the defender to play defence.

The seeding pts system is far from perfect. However, they should be the goal of any good qual strategy. They may be too noisy to derive a good statistic from but atleast in a way they reward successful execution of strategy. If you could normalise them by taking into consideration partner and/or opponent I think they would make a better statistic than DPR for this year.

[JesseK]

Originally Posted by The Lucas

So why use a statistic that assumes that everyone is trying to play defence all match if defence isn't necessarily sound strategy.
...
Can they change zones? Are they quick and manoeuvrable? Good driving instincts? Can they clear ball out of the zone? There arent statistics for these.

Keep in mind that your best defensive bots could be playing striker and scoring all through the quals. Defence and striker are very similar roles. Unless the defender is completely shutting the striker down, I'd pick the striker over the defender to play defence.
...

Actually, I hadn't thought about the first sentence at all. After some thought, I agree and also wonder if this is an indicator of how tough a schedule is? Yet coopertition points also somewhat take this into account since they're the un-penalized scores of the opposition (yet they're skewed since the winner receives double the loser).

The second sentence-- was my point in the 'fine print', and next reply. The only statistics for these are what the scouts collect while watching matches.

The third sentence -- that's a good point and it's actually what we were picked to do in DC. We were one of probably 7 bots that could score from midfield in autonomous, so we were asked to play defense for elims with the first move being to kick away the first 3 balls. We got the shutout we were looking for in QF1, still played good enough defense even with technical difficulties of a gimp drive train in QF2, held midfield scorers and a pusher bot to only 3 points in SF1, and yellow card aside held the offense to only 2 points for SF2 (valve went bad mid-match, thus putting a kicker out of the frame boundary for the remainder of the match). Up until elims, we hadn't played defense in a single match.