Strongest Regionals/District 2016?

[Andrew Schreiber]

Originally Posted by cadandcookies

Well, they were fine when 254 decided to visit in 2014, so, barring things I'm unaware of (of which there are many), I don't see 148 being in the mix shaking up 2056's chances of keeping their streak alive significantly.

There are a lot of ways to quantify "competitive". To me, it boils down to two questions: How difficult is it to win an event and how difficult is it to play Saturday (or Sunday) afternoon at an event. I'm not sure what the best way to quantify each of those is (and I'm pretty sure better minds than mine have tried), but it seems to me that one could come up with a number of statistics that might correlate to "competitiveness" at a regional.

I'm also not sure many of the previous posts in this thread are what I would say are most competitive overall. Perhaps in a given region, yes, but I wouldn't be likely to put one of our MN regionals on the same level as many Michigan District events, or even Wisconsin Regional, which always seems to be very deep, even if the ceiling isn't as high as places like Waterloo. It's quite difficult to make a judgement on the whole however, when I, like many people, only really have a strong knowledge of regionals in my area.

This is one of those questions I've been trying to figure out for a while. So, I've actually built out a system to provide what I'm calling a Dynasty rating for teams. It's effectively based on the team's ability to play in the afternoon. I've assigned winning an event as 100 points. From there, each level is divided by the number of alliances reaching it. (Finalist 50, SF 25, QF 12.5) I also reverted towards the mean by 5% each year to help offset teams that had a great year in like 2006 and haven't been consistent.[1] I've also added a minimum threshold of 5 events, teams are derated accordingly prior to that because they are unknown. So, my theory is that this shows an overall ability to play in eliminations.[2] These were averaged over a year.

After that I started asking myself how I could quantify competitiveness of an event. Frankly, I enjoy watching an event where the winner isn't a known thing[3]. So I did a bit of research, for those of you unfamiliar with the Gini Coefficient, it's a measure of income disparity in nations. But, at it's core, it's a measure of equality of distribution. So, I got to wondering if I could leverage the math behind it to find events with lots of close teams.


Ok, this post is dragging on a lot longer than I wanted it to, so the TLDR - you can, it's actually fairly simple math given a sorted array. So, by week, the most "exciting"[4]:

(If they were real close (within a hundredth) I included the top 2

Week 1

Guilford County 	 2016ncmcl 	
Kettering University 	 2016miket

Week 2

Tippecanoe 	 2016inwla
Central Valley 	 2016cama

Week 3

Walker Warren 	 2016inwch

Week 4

Central Maryland 	 2016mdedg

Week 5

Hartford 	 2016cthar

Week 6 [5]

Pine Tree 	 2016melew

This was all based on every official event from 2006 to present. [6]


I'm tired, but if folks want the other order I can do that tomorrow.

[1] Number felt right, not really based on much than the that.
[2] Spot checking teams felt right. In theory there's prolly a way of using this for Fantasy FIRST type stuff.
[3] Almost known thing... whatevs
[4] Read as, events with the lowest Gini coefficient since 0 would be perfectly equal distribution.
[5] Actually, NCCMP was flagging as the most even this week, but since it's a qualify only event I assume it's a TBA bug?
[6] Which surprises absolutely nobody that I have...

[dodar]

Andrew, would that be possible to do for Regionals?

[page2067]

Originally Posted by Andrew Schreiber

This is one of those questions I've been trying to figure out for a while. So, I've actually built out a system to provide what I'm calling a Dynasty rating for teams. It's effectively based on the team's ability to play in the afternoon. I've assigned winning an event as 100 points. From there, each level is divided by the number of alliances reaching it. (Finalist 50, SF 25, QF 12.5) I also reverted towards the mean by 5% each year to help offset teams that had a great year in like 2006 and haven't been consistent.[1] I've also added a minimum threshold of 5 events, teams are derated accordingly prior to that because they are unknown. So, my theory is that this shows an overall ability to play in eliminations.[2] These were averaged over a year.

After that I started asking myself how I could quantify competitiveness of an event. Frankly, I enjoy watching an event where the winner isn't a known thing[3]. So I did a bit of research, for those of you unfamiliar with the Gini Coefficient, it's a measure of income disparity in nations. But, at it's core, it's a measure of equality of distribution. So, I got to wondering if I could leverage the math behind it to find events with lots of close teams.


Ok, this post is dragging on a lot longer than I wanted it to, so the TLDR - you can, it's actually fairly simple math given a sorted array. So, by week, the most "exciting"[4]:

(If they were real close (within a hundredth) I included the top 2

Week 1

Guilford County 	 2016ncmcl 	
Kettering University 	 2016miket

Week 2

Tippecanoe 	 2016inwla
Central Valley 	 2016cama

Week 3

Walker Warren 	 2016inwch

Week 4

Central Maryland 	 2016mdedg

Week 5

Hartford 	 2016cthar

Week 6 [5]

Pine Tree 	 2016melew

This was all based on every official event from 2006 to present. [6]


I'm tired, but if folks want the other order I can do that tomorrow.

[1] Number felt right, not really based on much than the that.
[2] Spot checking teams felt right. In theory there's prolly a way of using this for Fantasy FIRST type stuff.
[3] Almost known thing... whatevs
[4] Read as, events with the lowest Gini coefficient since 0 would be perfectly equal distribution.
[5] Actually, NCCMP was flagging as the most even this week, but since it's a qualify only event I assume it's a TBA bug?
[6] Which surprises absolutely nobody that I have...

Note not only close competition at Hartford,
But Hartford has 10 of 38 competitors that were in eliminations at championships last year, and second to top on average qualification score of all regionals or districts - so competitive and at a high level. None of the super-star teams maybe (is 195?), but solid to the core.

Interesting analysis Andrew.

[jajabinx124]

Originally Posted by cadandcookies

I'm also not sure many of the previous posts in this thread are what I would say are most competitive overall. Perhaps in a given region, yes, but I wouldn't be likely to put one of our MN regionals on the same level as many Michigan District events, or even Wisconsin Regional, which always seems to be very deep, even if the ceiling isn't as high as places like Waterloo. It's quite difficult to make a judgement on the whole however, when I, like many people, only really have a strong knowledge of regionals in my area.

My lists for a couple regionals were definitely a bit biased since I know the local teams/powerhouses better.

Originally Posted by tindleroot

No, 148 will definitely shake up OP's chances. In 2014 there was a little bit of luck that went their way to keep 2056 on the winning streak. Namely, the fact that 4039 seeded first and picked Simbotics allowed 254 and 2056 to pair up. In many other situations, 254 and 1114 could have paired up and ended 2056's streak. Not to mention, if 148 had come to Waterloo this past year and 1114's canburglars were better, then the two probably would have paired up.

I'm sure 2056 is up to the challenge and excited to fight for another regional win this year at Waterloo.

I agree. The presence of 148 will shake 2056's chances of continuing their win streak. We definitely can't make accurate predictions now since we don't know what the game is and what their robots are on a competitive level, but we can guess. Regardless Waterloo will be a fun regional to watch!

[Caleb Sykes]

Originally Posted by Andrew Schreiber

...lots of good info...

If it is easy to do, could you try this method on just the top 24 teams at each event?

[Andrew Schreiber]

Originally Posted by Caleb Sykes

If it is easy to do, could you try this method on just the top 24 teams at each event?

https://gist.github.com/schreiaj/e87...file-top24-tsv

It does change things up, makes a lot of events look more balanced. For reading the Gini column - 0 is balanced, 1 is unbalanced.

Originally Posted by dodar

Andrew, would that be possible to do for Regionals?

https://gist.github.com/schreiaj/e87...b#file-all-tsv That actually included regionals, but districts seem to be more balanced.

[Caleb Sykes]

Originally Posted by Andrew Schreiber

https://gist.github.com/schreiaj/e87...file-top24-tsv

It does change things up, makes a lot of events look more balanced. For reading the Gini column - 0 is balanced, 1 is unbalanced.

Thanks, this is really interesting. I'm surprised by how much more balanced the top 24 are when compared to all teams. I suppose that every 40+ team event has at least 10 teams that almost never make playoffs, which really increases the disparity.

[Karthik]

The two metrics I like to look at when gauging the top end teams at a regional during the preseason are the number of previous year Championship elimination round teams and Einstein teams. I've only counted for the events we're attending, but I'm sure someone could run the numbers across all events.

GTR East:
Champs Elim Teams: 6/52
Einstein Teams: 4/52

Waterloo:
Champs Elim Teams: 8/32
Einstein Teams: 3/32

I'm guessing no other regional is going to match the 25% number that Waterloo has, but that's probably artificially inflated by the size (or lack thereof) of the event. That being said, that 25% number also means you're not going to run into many matches without a high level team. So the event becomes that much more webcast/spectator friendly.

[XaulZan11]

Originally Posted by Karthik

I'm guessing no other regional is going to match the 25% number that Waterloo has, but that's probably artificially inflated by the size (or lack thereof) of the event.

I'd also assume that 25% is partially due to the abnormally high percentage of teams from Canada making the championship due to the smaller regionals.

[Karthik]

Originally Posted by XaulZan11

I'd also assume that 25% is partially due to the abnormally high percentage of teams from Canada making the championship due to the smaller regionals.

That would explain a larger number of teams who attended Championship in the previous year, but it wouldn't explain why the teams were so successful at Championship.

[cadandcookies]

Originally Posted by Karthik

That would explain a larger number of teams who attended Championship in the previous year, but it wouldn't explain why the teams were so successful at Championship.

Unless we had data backing a causative (or at least correlative) relationship between attending championships and future success on the field, which I don't think is a link anyone has shown, but is something I think many assume.

[XaulZan11]

Originally Posted by Karthik

That would explain a larger number of teams who attended Championship in the previous year, but it wouldn't explain why the teams were so successful at Championship.

Agreed, but when over 40% of teams make eliminations at the championship, the more teams you send, the more will make eliminations.


As an aside, I think it would be intersting to compare the percent of teams making championship eliminations based on region. I'm not sure that's been done before.

[Richard Wallace]

Originally Posted by Karthik

... the event becomes that much more webcast/spectator friendly.

Team quality is one reason for this. Field volunteers are another.

Wish I were going ...

[indubitably]

Originally Posted by Karthik

GTR East:
Champs Elim Teams: 6/52
Einstein Teams: 4/52

Waterloo:
Champs Elim Teams: 8/32
Einstein Teams: 3/32

I'm guessing no other regional is going to match the 25% number that Waterloo has, but that's probably artificially inflated by the size (or lack thereof) of the event. That being said, that 25% number also means you're not going to run into many matches without a high level team. So the event becomes that much more webcast/spectator friendly.

The ratio is effective at determining the competitiveness of the average qualification match but everything playoff related including seeding is better represented by the actual number of Championship playoff teams.

Originally Posted by Karthik

That would explain a larger number of teams who attended Championship in the previous year, but it wouldn't explain why the teams were so successful at Championship.

The more teams that get sent to champs, the less there is a chance of a good team getting snubbed which happens quite frequently in areas without districts or fewer/bigger events.

[page2067]

Originally Posted by Karthik

The two metrics I like to look at when gauging the top end teams at a regional during the preseason are the number of previous year Championship elimination round teams and Einstein teams. I've only counted for the events we're attending, but I'm sure someone could run the numbers across all events.

GTR East:
Champs Elim Teams: 6/52
Einstein Teams: 4/52

Waterloo:
Champs Elim Teams: 8/32
Einstein Teams: 3/32

I'm guessing no other regional is going to match the 25% number that Waterloo has, but that's probably artificially inflated by the size (or lack thereof) of the event. That being said, that 25% number also means you're not going to run into many matches without a high level team. So the event becomes that much more webcast/spectator friendly.

Hartford District -
Champs Elims 10/38 (26%)
Einstein - 0/38