What happened in Curie???

[Paul Copioli]

I thought Curie had the most "normal" alliance selection of any of the divisions. Let me share some things I observed.

First, our team agrees that 7 matches is really not eough to statistically determine who the top 8 really is. Unfortunately, teams that are clearly not top 8 caliber will get into the top 8. This is one of the reasons the FIRST in Michigan district model had 12 matches at each of the district events.

In our team's opinion, there were 4 teams in the top 8 that were not really top 8 caliber. Such is life at the Championship.

Our team scouts every match and takes the best statistical data we can for each and every team. Scouting the actual matches is the best way to definitely determine if a team will make a good partner in the elimination rounds. On Friday, we have a 3 to 4 hour strategy meeting where we make a pick list of the top 24 teams based on our scouting data and feedback from the drive team. We use feedback from the drive team to help determine if a team is hard or easy to work with. Our scouting team then focuses on the teams that are very close in performance on Friday to see how they are doing on Saturday morning matches. We really focus on teams that are performing better as the weekend moves along.

OPR and other indicators are really only needed when you can't actually watch the matches in person. We do not use OPR because OPR will not get you to Einstein. Period.

Here are some other things we do not use:

- Record in the division
- Ranking in the division
- Performance at Regional events
- Team number
- Where a team is from

I can't even tell you what 68 or 247 records or rankings were because we didn't care.

For the last 5 years, I have witnessed the strangest picks by teams in the top 8. Teams that would have been our 4th or 5th pick slipped all the way to the bottom of the draft. This year in Curie was the first time that I can remember that the teams selecting partners during the draft basically picked how we would have picked (except 68 as they should not have been around for us). With 87 teams in the division some good teams will be overlooked and there is no way aroud it.

I want to make one more point. The serpentine draft has caused a lot more teams to decline. I can tell you that our team prefers not to decline, but will if we feel that it gives us a strategic advantage. We declined 1806 (the #2 seed) not becasue we didn't think they were a good team (they were very high on our list), but simply because of the serpentine. If the serpentine did not exist, then we would have definitely accepted their invitation. Team 1806 knew prior to alliance selections that we were going to decline as I told their team that we would decline. They used a very good strategy picking us anyway so no one else could select us. They basically forced 217 and 399 to form their own alliances (they asked 399 to be their partner too and 399 declined).

Our division had only 14 teams with a positive "plus/minus". Plus/minus (for our team) is the number of points scored by the robot minus the number of points in the robot's trailer. This year it is the biggest indicator of a team's contribution. Obviously, for pure defensive bots we only look at points scored in their trailer as +/- will, by definition, always be negative. As the #7 seed, we knew we could get two very good robots to complement our playing style. We felt we could build a better alliance from #7 than from #2. And for those of you wondering, I didn't realize we had an all Michigan alliance until about 10 minutes after selection. It is just the way things worked out.

Someone made a comment about high number teams not getting picked. We had a few high number teams on our list (I was suprised no one picked 3115), but we had the #7 and #10 picks in the draft.

[sdcantrell56]

That is a great post and I will probably be saving it to help us with scouting in the future. I too could not believe that 68 was till around for you guys. On my list I had them at least in the top 5 including your team.

One other rookie team that I was slightly suprised not to see selected was 2815, but given the teams that were picked I dont see what team they could have been selected in place of. Overall I thought the selections went very well and all of the matches were fairly close as a result.

I am still disappointed that we did not have the opportunity to play with 188, 217, or 68 at all at the championship, but that is just the way the matches go. There is always next year to get the chance to play with these incredible teams.

[Noeen Kashif]

Originally Posted by sdcantrell56

That is a great post and I will probably be saving it to help us with scouting in the future. I too could not believe that 68 was till around for you guys. On my list I had them at least in the top 5 including your team.

One other rookie team that I was slightly suprised not to see selected was 2815, but given the teams that were picked I dont see what team they could have been selected in place of. Overall I thought the selections went very well and all of the matches were fairly close as a result.

I am still disappointed that we did not have the opportunity to play with 188, 217, or 68 at all at the championship, but that is just the way the matches go. There is always next year to get the chance to play with these incredible teams.

Yo guys, were really great at Atlanta as well, you guys had an amazing alliance, and put up a great and enjoyable show. We would like to work with you guys as well in the future. Like you said, there is always next year!

[Lil' Lavery]

I have to just chip in a couple points about the earlier discussion involving defense.

I don't like overly defensive games, they tend to be boring, but defense has a definite role in FIRST, in my opinion. It's a very very large part of what separates FRC from just being an "engineering competition," "race," or "skills contest."

Secondly, where do you draw the line between "defense" and "strategy?" Is pinning a robot to aid your alliance partner's scoring effort defense? Is reserving your balls until the end of the match to ensure the other team doesn't get additional ammo defense? Is swerving your trailer out of the way of your opponent's best scoring machine defense?


As for rankings, the only real solution is to play more matches. How you accomplish that has been debated and contemplated for years, but no clear solution has emerged.

[robodude03]

Originally Posted by Paul Copioli

I thought Curie had the most "normal" alliance selection of any of the divisions. Let me share some things I observed.

First, our team agrees that 7 matches is really not eough to statistically determine who the top 8 really is. Unfortunately, teams that are clearly not top 8 caliber will get into the top 8. This is one of the reasons the FIRST in Michigan district model had 12 matches at each of the district events.

In our team's opinion, there were 4 teams in the top 8 that were not really top 8 caliber. Such is life at the Championship.

Our team scouts every match and takes the best statistical data we can for each and every team. Scouting the actual matches is the best way to definitely determine if a team will make a good partner in the elimination rounds. On Friday, we have a 3 to 4 hour strategy meeting where we make a pick list of the top 24 teams based on our scouting data and feedback from the drive team. We use feedback from the drive team to help determine if a team is hard or easy to work with. Our scouting team then focuses on the teams that are very close in performance on Friday to see how they are doing on Saturday morning matches. We really focus on teams that are performing better as the weekend moves along.

OPR and other indicators are really only needed when you can't actually watch the matches in person. We do not use OPR because OPR will not get you to Einstein. Period.

Here are some other things we do not use:

- Record in the division
- Ranking in the division
- Performance at Regional events
- Team number
- Where a team is from

I can't even tell you what 68 or 247 records or rankings were because we didn't care.

For the last 5 years, I have witnessed the strangest picks by teams in the top 8. Teams that would have been our 4th or 5th pick slipped all the way to the bottom of the draft. This year in Curie was the first time that I can remember that the teams selecting partners during the draft basically picked how we would have picked (except 68 as they should not have been around for us). With 87 teams in the division some good teams will be overlooked and there is no way aroud it.

I want to make one more point. The serpentine draft has caused a lot more teams to decline. I can tell you that our team prefers not to decline, but will if we feel that it gives us a strategic advantage. We declined 1806 (the #2 seed) not becasue we didn't think they were a good team (they were very high on our list), but simply because of the serpentine. If the serpentine did not exist, then we would have definitely accepted their invitation. Team 1806 knew prior to alliance selections that we were going to decline as I told their team that we would decline. They used a very good strategy picking us anyway so no one else could select us. They basically forced 217 and 399 to form their own alliances (they asked 399 to be their partner too and 399 declined).

Our division had only 14 teams with a positive "plus/minus". Plus/minus (for our team) is the number of points scored by the robot minus the number of points in the robot's trailer. This year it is the biggest indicator of a team's contribution. Obviously, for pure defensive bots we only look at points scored in their trailer as +/- will, by definition, always be negative. As the #7 seed, we knew we could get two very good robots to complement our playing style. We felt we could build a better alliance from #7 than from #2. And for those of you wondering, I didn't realize we had an all Michigan alliance until about 10 minutes after selection. It is just the way things worked out.

Someone made a comment about high number teams not getting picked. We had a few high number teams on our list (I was suprised no one picked 3115), but we had the #7 and #10 picks in the draft.

I have to agree with Paul's comments here and say that on our end our team also does a fair amount of scouting. We usually have a 3 hour meeting on Friday with our scouting team and also list the top 20 robots that we should pick. Once we get the input from the scouting team, the drive team talks over the picks made by the scouting team and we order them based on information given to us by the team.

During this event we also collaborated with team 188 and greatly benefited from their data. One of the major data points that our driver Brad wanted to look at was the amount of balls in the robot's trailers. This influenced our picks to a point, but we stuck primarily with the information from our team members.

As for the decline, we also mirror Paul's thoughts on this. Although we would normally accept the invitation to join 1806 on the playing field it was a matter of strategy, alliance picking, and bracket positioning that caused the decline. We felt that we could create a better alliance being the fourth alliance. We knew that 217 would not be available by the time that it was our turn for the pick and it was an excellent move by 1806 to breakup the possibility for anyone else to build up an alliance with them. We felt we had a strong chance with 188 and 329 to take it to the finals and we nearly had it, taking the showdown to 3 matches We were very happy with our alliance and the effort we made for Einstein.

[Diana Gee]

Paul Copioli from 217 Thunder Chickens.....

THANKS for your post and explaination of how your Team scouted for the selection of the elimination rounds.

As a mentor for 1983 SKUNKWORKS I would be curious to understand where our Team stood in your findings or other Teams findings, and as a whole where you saw our Teams weaknesses.

THANKS!
Darin (husband of Diana)

[Jonathan Norris]

Thats how it goes, 2 points Saturday morning we would have seeded 3rd, instead we seeded 10th. As a rookie team we knew that it was going to be difficult to be selected as a second round team, and having a communication error in our last match on Friday hurt our stats for Friday night. Without that match where we were dead it put us right in with the teams selected in the second round, but we know that with the quality of teams in the hunt as second picks it was going to be tough to get picked.

We were praying that two teams would pick within the top 8 so that we could move up into the 8th seed, and I can tell you we put quite an effort into scouting and would have made a tough 8th alliance. But i've been around long enough to know that this alliance selection went as normally as it could have, and if teams picked and declined properly we shouldn't have moved up. So if anything I applaud the teams in the top 8 for making some solid alliances.

[Chief Pride]

I would have to agree with Paul. The ONLY fair way to make the ranking system more reliable is to complete more matches.

[Ed Law]

Originally Posted by Paul Copioli

Our division had only 14 teams with a positive "plus/minus". Plus/minus (for our team) is the number of points scored by the robot minus the number of points in the robot's trailer. This year it is the biggest indicator of a team's contribution. Obviously, for pure defensive bots we only look at points scored in their trailer as +/- will, by definition, always be negative.

Paul,

Thanks for explaining your scouting and selection philosophy and posting your scouting data and selection list on the other post. It helps us understand how good teams select their partners.

I must say that I am very surprised you do not take into account how many moon rocks the human players attempted and scored. According to Team 188's scouting database, your human player do not attempt to score very often because your strategy is to have the human player feeds moon rocks into your robot. However we can not say that for other teams. Using your data in the raw data sheet, if you sum column F which is Moon rock scored by robots in all the matches you get 2354. And if you sum column J which is Moon rock in trailer you get 5080. This shows that only 46% of the moon rocks are scored by robots.

If we look at Team 188's database, the human player percentage ranges from 21% to 77% with a mean of about 47 and standard deviation of about 10. Since human players scored 54% of the moon rocks, picking a team that has a 60% shooter rather than a 30% shooter in a 100 point game would mean a difference of 16 points, which is quite significant. In past years the role of human player to scoring is limited. I don't know what percentage the human player should contribute to the score in an ideal game. I feel that this year their contribution to the final score is on the high side and thus scouting data should not ignore them. Another observation I have is robots can be defended but you can not defend against a good human player. They just keep putting moon rocks into trailers.

Ed

[J Flex 188]

Wouldn't defending against a good human player simply be moving my robot to the side of the field that they weren't able to reach, if that was my concern?

I would be more afraid of a highly mobile, highly effective robot like 217 unleashing a stream of moon rocks anywhere on the field into my trailer than I would be of a good human player taking a low percentage shot halfway across it.

Originally Posted by Ed Law

Another observation I have is robots can be defended but you can not defend against a good human player. They just keep putting moon rocks into trailers.

Ed

[Paul Copioli]

Ed,

You make a good point, but for our team and our strategy we intended to starve the field of rocks. When a human player was on our alliance, the first thing they were told was to not shoot the moon rocks unless the trailer was stopped generally in front of them.

We prioritized on a high capacity dumper that could human load from the top and could human load in auton. Combined with our human loading in auton, we could keep many of our balls off of the floor and "starve" the field. In essence, almost every ball we were pciking up from the floor was not ours.

There were many ways to wn this year, but this strategy worked for us.

[MattB703]

I think that you will notice that the Championship winning alliance used this same strategy. It was also used very succesfully in the Michigan Championship eliminations.

[Lil' Lavery]

That strategy also worked very well basically everywhere it was used (so long as robots were capable of being loaded from the human players, either through the airlock or over the wall). In general, the accuracy of a top-tier scoring machine was higher than 99% of human players, resulting in far fewer balls ended up on the floor if robots were used as the primary scoring method.

[ScottOliveira]

Originally Posted by Ed Law

Paul,

Thanks for explaining your scouting and selection philosophy and posting your scouting data and selection list on the other post. It helps us understand how good teams select their partners.

I must say that I am very surprised you do not take into account how many moon rocks the human players attempted and scored. According to Team 188's scouting database, your human player do not attempt to score very often because your strategy is to have the human player feeds moon rocks into your robot. However we can not say that for other teams. Using your data in the raw data sheet, if you sum column F which is Moon rock scored by robots in all the matches you get 2354. And if you sum column J which is Moon rock in trailer you get 5080. This shows that only 46% of the moon rocks are scored by robots.

If we look at Team 188's database, the human player percentage ranges from 21% to 77% with a mean of about 47 and standard deviation of about 10. Since human players scored 54% of the moon rocks, picking a team that has a 60% shooter rather than a 30% shooter in a 100 point game would mean a difference of 16 points, which is quite significant. In past years the role of human player to scoring is limited. I don't know what percentage the human player should contribute to the score in an ideal game. I feel that this year their contribution to the final score is on the high side and thus scouting data should not ignore them. Another observation I have is robots can be defended but you can not defend against a good human player. They just keep putting moon rocks into trailers.

Ed

I would have to agree that this year the human player just had too much of a contribution to scoring. FIRST is meant to be a robotics competition, and when only half, or less than half, of the scoring is done by robots, it seems to be deviating from the purpose of the game.

I also feel autonomous mode was very undervalued this year, with the only thing to do in autonomous mode really being to try to get away from a human player! Towards the end of championship, we got autonomous scoring working, but thinking back on it, it may not have been as useful, although it was impressive. Since most teams drive towards their human players, following one in order to shoot on it puts us in range of their human, who has 13 moon rocks to our 7.

It would be nice to see more of an emphasis put on getting things done in autonomous mode (think about last year with the bonus points for running laps, or the year before with the keepers that guarantee you a spot on the rack).

It would also be nice to have something done about no-show and nonfunctional robots. Of course there is no simple solution. Personally, I would like any robot that doesn't show to get a loss, regardless of how the alliance performs. The argument against this is that it encourages teams to put an inoperable bot on the field, or rush repairs too much. However, this could be fixed by having a team who's robot doesn't work at any point in the match to receive a loss also. This might seem harsh, but it goes back to the real world, where you don't get paid for a product that doesn't work.

I would also point out that while this might seem to hurt rookie teams in particular, it is often easy for rookie teams, or any team, to get help while at competition. From the example Martin gave about the team at a regional who was seeded first with 4 no-shows, Sean and myself spent a while working on their drive train until they functioned (and we didn't finish until AFTER the match we played with them, for the match we played against them). I also always go to future alliance partners, and do whatever I can to make sure they work, if there is some sort of problem.