Galileo - The Champion's Division
 

So I had a thought the other day. "What division has won the championship the most?"

Some quick counting revealed that:
Galileo has won 5 times
Newton has won 4 times
Archimedes has won 2 times
Curie has won 1 time

I was surprised to see it wasn't Newton, but I kind of also figured that it would be a mix of all 4 divisions.

Re: Galileo - The Champion's Division
 

You might want to check that one again.

Re: Galileo - The Champion's Division
 

Galileo has also won the past three years, making them 3-0 in the St. Louis era.

Re: Galileo - The Champion's Division
 

I think the actual numbers are G-5, N-4, A-3, C-0. (assuming there were no division before 2001, anyone who's been around longer please correct me if I'm wrong)

Also note, Galileo has won 5 of the last 6, only to be upset by Newton in 2010.

Re: Galileo - The Champion's Division
 

Intiresting, where did you get these numbers?

Re: Galileo - The Champion's Division
 

Galileo never won a championship until 2008.
Did they really win every championship from 2008-2013 except 2010?

Re: Galileo - The Champion's Division
 

Yes, this is true. That's pretty interesting.

Re: Galileo - The Champion's Division
 

Having been the VC for that field from 2006 - 2012, that sounds about right.

Re: Galileo - The Champion's Division
 

The Blue Alliance

Did you get that number from not counting the division in 2001? Because that was the year Curie won.

Re: Galileo - The Champion's Division
 

This has Newton winning 2001.

2001:N
2002:N
2003:A
2004:A
2005:N
2006:A
2007:N
2008:G
2009:G
2010:N
2011:G
2012:G
2013:G

The BlueAlliance has no data for 2001, where did you get your information?

Re: Galileo - The Champion's Division
 

What would be really interesting is knowing this with the context of when FIRST changed the sorting algorithm. (I forgot)

Re: Galileo - The Champion's Division
 

Is the algorithm they use now known?

Re: Galileo - The Champion's Division
 

I remember someone saying that the divisions used to go by team number, like team 1 in Archimedes, team 11 in Galileo, 20 in Curie, 25 in Newton and then repeating. I don't think we know for sure what it is now, but I think someone might know what year they stopped doing the old one.

Re: Galileo - The Champion's Division
 

I have no idea what their sorting algorithm is (I took a crack at figuring it out once and got nowhere). But I can tell you what I'd do if I had to solve that problem. I'd generate some random divisions and see if they looked approximately balanced. I'd be looking for balance in several ways: competitive teams, former champions, Hall of Fame teams, rookies, waitlist teams, etc. If it looked too imbalanced or had a division that was too similar to the previous year or two, I'd throw it out and keep doing random ones until I got one that looked good. That would make the "algorithm" pretty hard to solve.

Re: Galileo - The Champion's Division
 

This would never happen, but I'd like to see four experts (Karthik, Paul C, someone nominate two others?) each represent a division. They would then take turns drafting the entire division by whatever thought process they individually desired. Their own 4 teams would be not included in this draft, and then randomly assigned to the 4 after the divisions are set.

I think we'd see some incredibly balanced divisions in a way anything FIRST does could never match.

Re: Galileo - The Champion's Division
 

I really want to see the power alliances that come out of this.

Re: Galileo - The Champion's Division
 

I agree that this would be really cool.

On the other hand, I think I like a bit of randomness. Not a ton, but a bit. With randomness, there's always the chance that the top two teams in the world will be together in the same division and then end up allying. Removing that possibility would be sort of unfortunate. And having a division that's a little bit strong or weak (not a lot, but a little) is actually good for certain teams. A team that performs at the top of the second tier of teams would have a slightly better chance at making it to Einstein if they're in a slightly weaker division, even if it reduces that division's odds of winning it all. Having a strong division sets up the chance of a glorious upset. And a stronger division is, of course, better for the very best teams, because it provides a better best case scenario. This is all fun stuff to think about.

Re: Galileo - The Champion's Division
 

Assigning divisions by OPR would actually be incredibly easy and very balanced. Idk why they don't just do that if they want balanced divisions.

Re: Galileo - The Champion's Division
 

Because then people would jump down Manchester's throat because OPR is a useless/misunderstood/abused/biased/wrong/silly/number-based/ridiculous/old-fashioned/left-leaning/conservative/smelly/overrated metric.

But hey, who knows, maybe that's how they do it already! It's not like they've told us.

Re: Galileo - The Champion's Division
 

Well it's really hard to determine the "top two teams." Or top 4 and even 10. Maybe just creating a top tier is more appropriate.

When I was at World's watching Galileo, I for sure thought 118 and 1114 were the "top two teams." But sure enough, they were beaten by the likes of 1477, 610, and 1241.

Top tier teams can be beaten any day of the week (We actually played a part in giving 1477 their first match loss at LSR) and honestly Champs has a lot to do with luck (mainly the luck of the draw).

Re: Galileo - The Champion's Division
 

Because it's not about the Robot?

Re: Galileo - The Champion's Division
 

It's actually not. OPR has a ~90% correlation with actual offensive scoring ability. You just need a large enough sample size to get it reasonably accurate (at least 10 matches or more). And I know for a fact that they don't use OPR to build divisions because I compared all 4 divisions and it was extremely obvious that some of them were incredibly stacked compared to the others.

Re: Galileo - The Champion's Division
 

I think any attempt by FIRST to sort divisions by some sort of competitive metric will be a complete failure; FIRST simply isn't tuned into that kind of stuff.

Remember a few years ago when they tried to "balance" quals by sorting teams into three groups. The lowest 1/3rd number, the middle 1/3rd number and the highest. Each qual alliance had one from each group. This obviously hugely benefited high numbered competitive teams (2056, etc...).

Re: Galileo - The Champion's Division
 

An interesting idea I was toying with the other day is if you win an event you would us thr same alliance when you get to championship I can see this mixing up the alliance selection process alot.. I know that this wont ever happen or work to many logistics and issues would arid bu it's an interesting idea

Re: Galileo - The Champion's Division
 

I agree that opr isn't at all the best indicator of how good a robot is but it's probably the closest you will get for every team at championships. Heck even trying to get top strategists in FIRST to separate the teams would be hard due to not having data or video on many of the teams that compete at champs, let alone blurry webcasts that are hard to see which team is doing what at times. For the most part too, the top teams are always near the top so opr would work decent in my opinion.

Re: Galileo - The Champion's Division
 

What would you do about teams who won multiple events with different partners? :P

Re: Galileo - The Champion's Division
 

Yeah I knew that there where many problems with it and that's why it will never happen. That being one of them and also the fact that if one of the three can't make it to championship

Re: Galileo - The Champion's Division
 

Fair enough. I'll give you that OPR was quite good last year. On the other hand, there were some years (like 2012) when it was absolutely terrible. Personally, I like OPR a fair bit, but it certainly can't tell you all there is to know about a team, and often tells you significantly less than what is important (at least, important for alliance selection).

I was commenting more from the perspective that there seems to be an irrational hatred of OPR among some veteran members of the FRC community, and that those people might be against a metric like OPR controlling divisions.

Re: Galileo - The Champion's Division
 

I remember ONE time when this happened. Want to say it was in 2004, but one alliance won a regional, then managed to all end up in the same division at Championship, pick the alliance together again, and make Einstein (but not win it).

As for winning multiple events with different partners, AC picks. I've seen that one too, in 2005--I want to say it was 245 pairing with 766 (Sacramento) and 217 (one of the MI regionals) to make Einstein. (Knocked out in Einstein semis.)

Re: Galileo - The Champion's Division
 

It was indeed 2004. 67, 1126 and 340 won the Buckeye Regional together that year, then met up again for elims on Newton, winning the division and then going on to lose in the Einstein semis.

Re: Galileo - The Champion's Division
 

Who do you mean by "they"? I don't think you can blame FIRST for the Alliance Algorithm of Doom.

Re: Galileo - The Champion's Division
 

That is a separate failure, and I believe only affected week 1 regionals? The alliance of doom looped you against the teams immediately adjacent to you in team number if I recall correctly.

Re: Galileo - The Champion's Division
 

This old post by Dave Lavery might be relevant to your discussion of the 2007 match algorithm. Apparently the algorithm designers essentially decided independent of FIRST what their priorities in assigning partners were. It seems like this post is referencing both the week 1 clustering of teams and the general "low / mid / high number vs low / mid / high number" algorithm. Apparently the latter was never fixed in 2007 despite FIRST's requests.

Re: Galileo - The Champion's Division
 

As I recall, the Algorithm of Doom was exactly what Alan said it was. Now, some of the events MAY have had the result you described--but that was due to being smaller events.

Typical Algorithm of Doom match would be that you were against one or two really good teams for half your matches, and never with them.

FIRST did release a fix, but it took a few weeks and didn't do much for smaller events, IIRC.

Re: Galileo - The Champion's Division
 

Re: Galileo - The Champion's Division
 

Sorry, Adam, but you're not adding anything new. If you have new information that you didn't ALREADY state (and which I was responding to). The discussion I remember applied the term "Algorithm of Doom" to the 3-group qual setup system, partly because if you were in the same group as [insert powerhouse] you would never be with them, and probably be against them 3 or 4 times. It wasn't fixed. IIRC, there was some modification AFTER a few weeks that was a slight improvement, but not enough to make much difference.


You may be thinking of 2010, where there was a change to the seeding structure after Week 1 produced a number of 6v0 matches.

Re: Galileo - The Champion's Division
 

Week 1 of 2007 the doom scheduler had you play with/against the team above or below you in team number every match.

Later in season, Champs for sure, teams were clearly lumped into 3 groups (in order of team number) with each qual alliance having a team from each group. Punishing low number teams that were not elite, and favoring high number teams that were elite.

Re: Galileo - The Champion's Division
 

Didn't remember the first part. As a matter of fact, this is the first I remember hearing about it. What I remember is the 2nd part being throughout the season--and, as a note, the 1st part could be an artifact of the 2nd part, to an extent.

Re: Galileo - The Champion's Division
 

One interesting Einstein winning alliance was the IFI alliance in 2008.

1114
217
148

It was all coincidence how it worked out.

Re: Galileo - The Champion's Division
 

I agree, opr is often not well received, and understandably so. It's not the best, or most accurate metric, but it's better than whatever they currently use and arguably the best option that can be used to measure every single team in frc since I doubt we can get real scouting data on every single team.

Re: Galileo - The Champion's Division
 

Check out 114's schedule at PNW 2007: http://www.thebluealliance.com/team/114/2007

Re: Galileo - The Champion's Division
 

It also only existed at week 1 events, and wasn't an issue at week 2. I remember this very well as we would either have played 254 or 330 in every single match, and that didn't happen.

Re: Galileo - The Champion's Division
 

We had that issue at Trenton in 07.

Check out who we played a bunch!
http://www.thebluealliance.com/team/11/2007

Re: Galileo - The Champion's Division
 

I remember that issue! At GSR, 1058 played 1073 every single qualification match with 1277 bouncing between alliances.

It was such a shame because both were fantastic machines that would have done better if they didn't have to play each other every match.

Re: Galileo - The Champion's Division
 

192 much?