Alliance Selection Strategy

[Tknee]

Originally Posted by themccannman

Scouting, if you guys don't already have a scouting system set up I would just use OPR data to select teams.

I think it is the word "just" that I and others have a problem with. The mental image of a team picking solely based off OPR seems almost as wrong as picking the next highest seed. Granted OPR is a much better metric than the ranking system, but great alliances aren't composed of offense alone.

Even without a scouting system providing hard numbers of a robots capabilities, a lot of useful soft information can be gleaned by talking to teams in their pits as well as talking to spectators.

[MrForbes]

Originally Posted by Tknee

Granted OPR is a much better metric than the ranking system, but great alliances aren't composed of offense alone.

Our scouting at the AZ regional was based pretty much totally on offense alone, we counted up auto, teleop and climb points for each robot, summed them for all the Friday matches, and used that to generate an ordered list. Result....we picked the #1 ranked OPR team first, and a very underrated team second. All three teams were offensive most of the time, but our second pick (3944) got drafted into defense mode for our semi's and shut down a full court shooter. Three offensive robots can do just fine at this game, as long as they're good offensive robots. Scoring is how you win matches.

Your regional may vary.

[Ether]

Originally Posted by Bill_B

The OPR score also cannot show a trend in a team's scoring

It could, but nobody has ever broken it down that way.

If a team competes in two regionals (or two districts) the trending OPR could be calculated as follows:

1) first 2/3 of first regional

2) last 2/3 of first regional

3) last 1/3 of first regional and first 1/3 of second regional

4) first 2/3 of second regional

5) last 2/3 of second regional

... and so on, if a third regional is involved.

[Ed Law]

Originally Posted by Jack_O

I have one main concern with OPR. I only had time to briefly look over how it is calculated, but what I saw troubled me. One teams OPR is influenced by their alliance partners. This influence may not be very big, but it can add up. If a team goes to an event where the teams are generally "good", then their OPR will be higher. The opposite happens when the teams are generally bad. Most OPR stats are seperated by less than one point. As a result, one match with bad partners could screw everything up.

That is just my two cents. If I am wrong, or if I misunderstood something, please let me know.

First of all, I just publish the numbers and provide means for other teams to use the data and to calculate the OPR during a competition. How they use the data is up to them.

My team does pit scouting, collect data of every robot of every match and combine that with OPR/CCWM to come up with a pick list. I don't advocate using only OPR. There are two circumstances I can think of that it is useful. When we were a rookie team, we had 6 students on the team, 4 of them were on the field or in the queue line most of the time, leaving only 2 students in the stand. In between matches, they were fixing and getting the robot ready for the next match. That year we used strictly OPR for first pick and CCWM for second pick. It was better than nothing.

The other way I would use OPR is if I was not at the competition so other means of scouting is not possible. The data tells me a lot about the teams. This data is very useful at the Championship when you have not seen how the other 99 robots in your division play, especially in the early matches.

The last time I check, the winner of the alliance is the one who can score more points. Offense is definitely important. However another way to look at it is if my alliance plays in such a way that we can create a positive point differential, my alliance will win. Three defensive teams, no matter how good they are at it, will never win a match. The best they can do is to tie it. You need at least one robot on the alliance that has sufficient offensive power to win.

Oh I almost forgot to answer this new posters question. You are absolutely wrong. Let me give you a simple example. I see you are also from Michigan. Let's say the district event has 40 teams. Let's say all teams except yours score exactly 20 points per match and your robot score 10 points. The calculated OPR is 20 for every other team and 10 for your team. Let's say you go to your second district and every robot there can score 40 points per match and your robot score 10 points per match. The OPR will be exactly 40 for all other teams and your team's OPR is still 10. So you are wrong to think that your OPR will be higher because you play in higher scoring matches. If you don't contribute to the higher score, you get nothing higher in your OPR.

[Siri]

Originally Posted by Jack_O

I have one main concern with OPR. I only had time to briefly look over how it is calculated, but what I saw troubled me. One teams OPR is influenced by their alliance partners. This influence may not be very big, but it can add up. If a team goes to an event where the teams are generally "good", then their OPR will be higher. The opposite happens when the teams are generally bad. Most OPR stats are seperated by less than one point. As a result, one match with bad partners could screw everything up.

That is just my two cents. If I am wrong, or if I misunderstood something, please let me know.

Welcome to ChiefDelphi!

You are in some respects correct, with some caveats and corrections. Yes, OPRs are based on calculations that involve the full alliance's score. However, "add up" is not the correct turn of phrase for potential inaccuracy--the more you play, specifically the more different alliances you (and your allies) play with, the less a role they should play in determining your OPR. As a simple example:

1114 + 000 + 999 = 150 ...this could mean that 1114 did nothing, and 000 scored 150 points alone
000 + 9999 + 8888 = 10 ...this makes it less likely that that happened
1114 + 7777 + 8888 = 180 ...again, it's getting clearer that 1114 has something going on here...
Imagine this continuing for another half dozen matches, and you'll see how more data tends to yield more accurate rather than less accurate results.

OPRs are also not as close as you think they are. For instance, at my last event, the standard deviation was 17.5, with a range of 69. Of 33 teams, the largest set within a 1pt range was 3 (13.7, 13.1, 13.0). Even at 60-some team events, the largest I found was 6. While they are somewhat sensitive, particularly at regionals with fewer qual matches, the amount that they'd change would probably be lower than the rating's useful precision anyway. And again, if your partners in that match are normally that bad in their matches, it certainly won't screw everything up for you.


EDIT: Or, you know, what Ed said. That too. (Except ever since EDIT: after 2011, 3 defensive robots can win matches.)

[Ed Law]

Siri,

When I say defensive robots, I don't mean they can score all their autonomous points and then play defense and then get the end game points. I meant pure defensive robots that do not score. You cannot win a match that way.

The type of robot that score well in autonomous and end games and play good defense in the middle will usually have a decent OPR. However their CCWM will be high because they will contribute to a big point differential. That is why I use CCWM to prescreen 2nd round pick and I never pick a team in the second round with negative CCWM. We do not use strictly OPR or CCWM to pick teams.

[Siri]

Originally Posted by Ed Law

Siri,

When I say defensive robots, I don't mean they can score all their autonomous points and then play defense and then get the end game points. I meant pure defensive robots that do not score. You cannot win a match that way.

Nor do I--but this year and in 2012, fouls committed by the opposing alliance add to your score. (I've actually seen matches--albeit not elim matches--won this way.)

[Ed Law]

Originally Posted by Siri

Nor do I--but this year and in 2012, fouls committed by the opposing alliance add to your score. (I've actually seen matches--albeit not elim matches--won this way.)

Ha ha, I forgot about that. A team that is good at getting the other alliance commit foul and get a certain number of points will have the same OPR as another team who actually score those points.

[themccannman]

Originally Posted by Tknee

I think it is the word "just" that I and others have a problem with. The mental image of a team picking solely based off OPR seems almost as wrong as picking the next highest seed. Granted OPR is a much better metric than the ranking system, but great alliances aren't composed of offense alone.

Even without a scouting system providing hard numbers of a robots capabilities, a lot of useful soft information can be gleaned by talking to teams in their pits as well as talking to spectators.

That's not what I was trying to communicate, my bad. I was using "just" as a verbal filler which I probably shouldn't. I was trying to make that clear in my other posts that if they have no other options to only rely on OPR for what it is, offensive ranking. Rarely do you ever pick a third robot to score points and be offensive. I realize I could have been clearer with my original posts but I hope this has clarified most of it.

[waialua359]

Ed,
first of all, thanks for posting the OPR data week after week, year after year.
More than anything, I find it useful to see how robots have done AFTER tournaments.....especially if our team is competing the same weekend (not being able to watch other webcasts).

Nothing beats the good 'ole fashion way of scouting by watching robots in matches.
In recent years, we have either been the #1 seed or picked by the #1 seed most of the time which has kept us busy on friday evenings after event socials.

We always look for robots that either best complement our strategy OR robots that we dont want to face in eliminations the most.

In this year's game, if we went to a regional where we were the best shooter far and above the second best shooting robot, I would NOT consider finding the second best shooter at the tournament......but instead, look if there is a 50 point climb/dumper robot (even if that is all they do!).
With that kind of combination, it would be hard pressed for any opposing alliance to overcome.....assuming autonomous points are equal and full court shooters can be blocked.
I beg to differ that 30 point climbing even if it takes 1 minute, doesnt make a difference. At many regionals this year, selecting a 30 point climber CAN make a difference. Or, it can be a pain to try and overcome if climbers are on the opposing alliance.

[Jack_O]

Okay, thanks for clearing that up. It helps a lot.

[efoote868]

Originally Posted by themccannman

That's not what I was trying to communicate, my bad. I was using "just" as a verbal filler which I probably shouldn't. I was trying to make that clear in my other posts that if they have no other options to only rely on OPR for what it is, offensive ranking. Rarely do you ever pick a third robot to score points and be offensive. I realize I could have been clearer with my original posts but I hope this has clarified most of it.

At the championship event when the division is ~100 teams deep, OPR from all previous competitions can be a valuable tool for the 2nd pick. At IRI when the event is ~70 deep, OPR might be valuable for the 2nd pick but probably won't for the 3rd pick (assuming they let captains pick 3).

[themccannman]

Originally Posted by efoote868

At the championship event when the division is ~100 teams deep, OPR from all previous competitions can be a valuable tool for the 2nd pick. At IRI when the event is ~70 deep, OPR might be valuable for the 2nd pick but probably won't for the 3rd pick (assuming they let captains pick 3).

I assume what you mean by second pick is the second alliance robot (the alliance captains first pick). I wouldn't consider using OPR for picking your third alliance robot unless you're in an unusual situation where you are looking for a high scoring third robot.

[Lil' Lavery]

Originally Posted by themccannman

I assume what you mean by second pick is the second alliance robot (the alliance captains first pick). I wouldn't consider using OPR for picking your third alliance robot unless you're in an unusual situation where you are looking for a high scoring third robot.

The "3rd pick" was in reference to IRI, where alliance captains draft their own "backup bot," essentially forming 4-team alliances.

[efoote868]

Originally Posted by themccannman

I assume what you mean by second pick is the second alliance robot (the alliance captains first pick). I wouldn't consider using OPR for picking your third alliance robot unless you're in an unusual situation where you are looking for a high scoring third robot.

No, second pick means the 3rd member of the alliance.

I expect there to be a good number (maybe 10 or more?) of robots at the championship event that will find themselves in the top 8 in each division, even though they'll be middle of the pack in OPR.

Their strategy should be to pick the two best offensive robots they can, so that the alliance captain in effect becomes the 3rd bot for defense. OPR in this case would be a good tool for them to help rank teams.

Also, depending on the team's strengths, an alliance of 3 offensive capable bots could be more effective than an alliance of 2 offensive capable bots and 1 defensive only bot. If this is a team's strategy, OPR is a good tool to help rank teams (or at least identify 20-30 teams to focus on in scouting instead of 90+).