Lopsided Divisions

[nuggetsyl]

I wish First would use a point system like what the districts have to help even out divisions. You could use the points to make sure all divisions in theory are equal in power. If anyone else has an idea throw it out there.

[AdamHeard]

Thanks

[araniaraniratul]

Thanks
::rtm::

[itsjustjon]

Having lopsided divisions is just a side-effect of random (or, at least, semi-random) assortment.

In my opinion, splitting up and assigning teams to divisions based on their performance would make CMPs less interesting.

Champs is fun because of the challenging schedules and obstacles it imposes on the teams who make it there.

This is all my opinion, though. My word is far from factual :D

[Sperkowsky]

Thanks

https://cdn.meme.am/instances/500x/25566060.jpg

[Hitchhiker 42]

https://cdn.meme.am/instances/500x/25566060.jpg

I'll second you on that one.

[snoman]

If divisions are lopsided. ( I haven't looked at them all) what ones are the most lopsided your opinion. I assume you are looking at first time teams vs teams with world experience.

[Hitchhiker 42]

If divisions are lopsided. ( I haven't looked at them all) what ones are the most lopsided your opinion. I assume you are looking at first time teams vs teams with world experience.

I believe OP is referring to the strong Newton division this year.

[Joe Johnson]

If divisions are lopsided. ( I haven't looked at them all) what ones are the most lopsided your opinion. I assume you are looking at first time teams vs teams with world experience.

I've done some analysis that I have shared in this thread (http://www.chiefdelphi.com/forums/showthread.php?p=1578136#post1578136).

Long story short. Newton is stacked. Some others, not so much (Curie & Galileo for example).

Back to the point the OP was making, yes, think too think FIRST should make an effort to balance the divisions. Whether it is via OPR or a District-like point system or another metric behind door #3, I don't think that the current system is where we want to be going forward. Year after year, FIRST ends up with obviously lopsided divisions and I think that damages the integrity of the sport -- Does FIRST care about such things to fix them? A question for another day...

Dr. Joe J.

[snoman]

Yeah that looks like a tough one. How about the " weakest " if teams made it to Champs they are good but relatively speaking.

[Joe Johnson]

Yeah that looks like a tough one. How about the " weakest " if teams made it to Champs they are good but relatively speaking.

There are some differences on the low end but nothing like the differences on the high end. (I will include the chart below so you can see for yourself). Also, once you get below a certain capability, it doesn't really matter very much how much below that you are. If The Poofs have an alliance partner that puts in 1 or 2 boulders in the low goal or if they have a one that puts in 3 or 4 low goal boulders it really isn't going to make much a difference to their score.

One interesting note there are teams with negative OPRs and one is going to St. Louis. FWIW.

Dr. Joe J.


http://i.imgur.com/YSZnCDk.jpg

[rich2202]

Group the teams by percentile (however you do that), and then assign those randomly to each division.

[GaryVoshol]

As long as they don't do something similar to that "scheduling algorithm of death" where they assumed all the low number teams were strong and the high number teams were weak.

[Hitchhiker 42]

As long as they don't do something similar to that "scheduling algorithm of death" where they assumed all the low number teams were strong and the high number teams were weak.

I'm not sure where I read this, but I'm pretty that the way they sort the divisions is they distribute rookies evenly b/w all divisions, then next year teams, etc.

Don't quote me on that, though. I might be wrong.

[Richard Wallace]

I'm not sure where I read this, but I'm pretty that the way they sort the divisions is they distribute rookies evenly b/w all divisions, then next year teams, etc.

Don't quote me on that, though. I might be wrong.

My understanding is that HQ lists teams in the order their payment for CMP is received, then deals them out to divisions in that order. This method of assigning divisions is effectively random; however, random assignments do not prevent a stacked division. Folks who have been scouting CMP for a long time can point to earlier examples. 2016 Newton is an extreme one.

[Arhowk]

My understanding is that HQ lists teams in the order their payment for CMP is received, then deals them out to divisions in that order. This method of assigning divisions is effectively random; however, random assignments do not prevent a stacked division. Folks who have been scouting CMP for a long time can point to earlier examples. 2016 Newton is an extreme one.

This method seems like it would stack divisions because theoretically the better teams are already anticipating a championship visit and are prepared to pay earlier

[Richard Wallace]

This method seems like it would stack divisions because theoretically the better teams are already anticipating a championship visit and are prepared to pay earlier

Sorry if my description was unclear. By "deals them out" I meant that the first team to pay goes into division 1, the second team to pay goes into division 2, etc. Like dealing cards from a deck.

[Hitchhiker 42]

This method seems like it would stack divisions because theoretically the better teams are already anticipating a championship visit and are prepared to pay earlier

If they are dividing them up, though, that would theoretically make for an even distribution of first early payers (best teams) all the way to latest payers.

[Doug Frisk]

This method seems like it would stack divisions because theoretically the better teams are already anticipating a championship visit and are prepared to pay earlier

No, the teams are assigned across the divisions as payment/registration happens. Team 1 goes to Archimedes, team 2 goes to Borlaug, 3.... and finally 9 goes back to Archimedes.

That's done for the non-rookie teams, then the same process is followed for the rookie teams so they are evenly distributed. At least that was the process in the past.

[SenorZ]

Is this REALLY the conversation we're having right before champs is split in two?
Any proposed "fix" will be moot for the 2017 championship event(s).

[Abhishek R]

Is this REALLY the conversation we're having right before champs is split in two?
Any proposed "fix" will be moot for the 2017 championship event(s).

Why? Just because there are two champs does not mean one division can't be more stacked than another.

[JosephC]

Any proposed "fix" will be moot for the 2017 championship event(s).

False, considering any solution found can just be used individually at both events, the only difference would be 4 divisions instead of 8.

[SenorZ]

False, considering any solution found can just be used individually at both events, the only difference would be 4 divisions instead of 8.

...and in determining who goes to which championship? Or is that just geo-based?

I think Adam's "Thanks" was in advance of the free show of pointless arguing he was about to see.

[JosephC]

...and in determining who goes to which championship? Or is that just geo-based?

I think Adam's "Thanks" was in advance of the free show of pointless arguing he was about to see.

Anything about who goes to what champ is pure speculation at this point, and really has nothing to do with a discussion about division "fairness" and its possible solutions.

99% of any discussion on CD is "pointless arguing", including your posts here, so I don't really get what you're aiming for?

_________________________________________________

IMO, the current method of assigning divisions works fine. Each team essentially get a random division (afaik, based on previous posts). If you ended up on Newton this year, that's just your luck of the draw. Is it "fair"? Pretty much. You had just as much as a chance as any other team attending worlds to make your division.

If FRC was to swap to a different system (like one based on the OP), how would they make it "fair"? OPR? Some type of District point esque system? There would always be an unhappy group, just as there is now, except they'd have an even larger excuse.

[Joe Johnson]

<SNIP>

99% of any discussion on CD is "pointless arguing", including your posts here, so I don't really get what you're aiming for?

<SNIP>

Kinda harsh. But sure, there is a lot of blathering on online but still, I think there is sometimes good that comes of it. Either a wider consensus that things are okay or perhaps enough peasants get excited enough to grab their pitchforks and storm the castle. It is not a completely useless exercise.

At least I don't think so.


<SNIP>
IMO, the current method of assigning divisions works fine. Each team essentially get a random division (afaik, based on previous posts). If you ended up on Newton this year, that's just your luck of the draw. Is it "fair"? Pretty much. You had just as much as a chance as any other team attending worlds to make your division.

If FRC was to swap to a different system (like one based on the OP), how would they make it "fair"? OPR? Some type of District point esque system? There would always be an unhappy group, just as there is now, except they'd have an even larger excuse.

<SNIP>

I am glad you're fine with things but I am unconvinced that there isn't a better system and that a better system wouldn't be worth the bother because people would still complain.

If FIRST aspires to be a robotic sport, then they should work to make it better for the same reason the NFL or the NCAA is continually tweaking the rule book and procedures to improve the experience for all.

Dr. Joe J.

[Andrew Schreiber]

Ok, let's do it by school size... or budget... or student count... or mentor count? Who wants to be in the mentor built division?

[Sperkowsky]

Ok, let's do it by school size... or budget... or student count... or mentor count? Who wants to be in the mentor built division?
Sign my team up who doesn't want to play with all of the good teams!

[bkahl]

It's all random.

[nuggetsyl]

IMO what makes the NFL great is anyone can win becuase the league leveled the playing field. No one team can buy all the good players.

[dodar]

IMO what makes the NFL great is anyone can win becuase the league leveled the playing field. No one team can buy all the good players.

So FIRST Championship needs to become Fantasy FIRST. Each division is given a leader in STEM and those Division Leaders then draft the teams they want to see play.

[FiMFanatic]

If FIRST really cared about a perfect setup for the Championship based on ranking, OPR, etc., they wouldn't invite so many teams in the first place. It is hugely watered down by some mediocre teams. 600 is too many.....

The district championships have higher level quality play (on average) unfortunately, other than the final eliminations.

If all they want is a lot of participation and enjoyment, then a randomized approach is fine. If you put too much reliance in OPR or other metrics, teams will act differently in the districts and regionals and make it a stat-based goal versus a win-based goal.

[Hadi379]

If FIRST really cared about a perfect setup for the Championship based on ranking, OPR, etc., they wouldn't invite so many teams in the first place. It is hugely watered down by some mediocre teams. 600 is too many.....

The district championships have higher level quality play (on average) unfortunately, other than the final eliminations.

If all they want is a lot of participation and enjoyment, then a randomized approach is fine. If you put too much reliance in OPR or other metrics, teams will act differently in the districts and regionals and make it a stat-based goal versus a win-based goal.


And they would only have ONE championship...

[FiMFanatic]

And they would only have ONE championship...

Agreed!

[EricH]

You guys did read the reasoning behind ChampionSplit, right?

It's to maximize inspiration.

At least, that's HQ's take on the matter. So they aren't exactly caring about the competition part being good (I mean, it is a Championship, so they have to, but it's a lower priority than inspiration).

[FiMFanatic]

You guys did read the reasoning behind ChampionSplit, right?

It's to maximize inspiration.

At least, that's HQ's take on the matter. So they aren't exactly caring about the competition part being good (I mean, it is a Championship, so they have to, but it's a lower priority than inspiration).

Agreed!

[marshall]

Ok, let's do it by school size... or budget... or student count... or mentor count? Who wants to be in the mentor built division?

Aren't all the good teams in that division? ;)

[evanperryg]

So FIRST Championship needs to become Fantasy FIRST. Each division is given a leader in STEM and those Division Leaders then draft the teams they want to see play.

This would be cool, dibs on team Andrew Schreiber

More seriously, the current system, assuming that the "registration dates" method is actually how it works, is perfectly fine to an extent. Personally, I don't think there should be any sorting based on perceived "powerhouse" nature of the team. There's plenty of top-tier teams who aren't well known, and there's plenty of big-name teams whose strength is overestimated because they're "that big-name team." Yeah, Newton's top-heavy, but the current methodology is random and considers every team without bias based on reputation. I feel bad for those top-tier teams who won't get a well-deserved trip to Einstein, but there's plenty of deserving teams every year who don't get to go. The only difference is that this year, most of them are in the same division.

[wjordan]

I was thinking about what the most lopsided divisions were since the CMP split in 2000, so I wrote a script to calculate the number of standard deviations away from the CMP average for each division's average OPR, as a rough metric of relative division strength:

(Average OPRs are pre-CMP for 2016, others are during CMP)

Rank Year Division, OPR STDVs away from CMP mean
1 2016 new 1.8792818365030972
2 2015 cars 1.5303350085259413
3 2005 arc 1.4719431806316507
4 2008 arc 1.275903384248631
5 2013 gal 1.216209400652486
6 2010 arc 1.193897925865034
7 2009 new 1.0851114986838635
8 2006 new 1.0248700734203415
9 2014 arc 1.0038012569280608
10 2004 arc 1.0012622874746064
11 2012 arc 0.9280026009838416
12 2011 cur 0.8605362826161431
13 2007 new 0.811482461068815
14 2012 new 0.791654392743334
15 2011 gal 0.723890899467247
16 2004 cur 0.6810947735598231
17 2006 cur 0.6086265070047694
18 2009 cur 0.5683687825075886
19 2014 new 0.5614091258588676
20 2007 gal 0.5525768854103231
21 2016 hop 0.5380374488668379
22 2016 cars 0.408018860269447
23 2010 cur 0.3624233446672217
24 2016 carv 0.33274494055516757
25 2015 hop 0.24442033949382544
26 2013 arc 0.2390748663971342
27 2015 tes 0.23375807978480515
28 2008 new 0.17118024613308042
29 2015 new 0.1671052847394964
30 2015 cur 0.14696320018600914
31 2007 arc 0.06155154314818187
32 2015 carv -0.08427736320166529
33 2015 gal -0.12364688580536025
34 2005 cur -0.24293216111580532
35 2013 cur -0.2777800696315326
36 2016 arc -0.2968187516006284
37 2011 new -0.302185773307164
38 2014 gal -0.30707335292203575
39 2008 cur -0.33900159574861066
40 2010 gal -0.43743412228260514
41 2006 gal -0.4667057439069076
42 2005 gal -0.5166965248897325
43 2009 gal -0.5693576991292799
44 2004 new -0.597846349430966
45 2005 new -0.7123144946261157
46 2012 cur -0.7304368298598461
47 2016 cur -0.7416191862759174
48 2016 tes -0.9674409454187557
49 2012 gal -0.9892201638673295
50 2009 arc -1.0841225820621547
51 2004 gal -1.084510711603477
52 2008 gal -1.108082034633088
53 2010 new -1.118887148249653
54 2016 gal -1.1522042028992627
55 2006 arc -1.166790836518207
56 2013 new -1.1775041974180875
57 2014 cur -1.2581370298648897
58 2011 arc -1.2822414087762262
59 2007 cur -1.4256108896273216
60 2015 arc -2.114657663723039

This is a fairly rough metric, with some possible bias towards the 8-divisional format due to a larger n and therefore a smaller standard deviation, but I think it's a good way to compare across years.

[lkg9999]

I'm a little confused how the registration date sorting works. Most of the teams from New England did not qualify until district championships, yet there are three divisions that have 8-9 New England teams and two that only have 2. Given that 90% of those teams qualified on the same date and would likely register within a short time frame, I would expect to see them more evenly distributed.

[Andrew Schreiber]

I'm a little confused how the registration date sorting works. Most of the teams from New England did not qualify until district championships, yet there are three divisions that have 8-9 New England teams and two that only have 2. Given that 90% of those teams qualified on the same date and would likely register within a short time frame, I would expect to see them more evenly distributed.

As did all the teams from MAR and FiM...

[wilsonmw04]

You guys did read the reasoning behind ChampionSplit, right?

It's to maximize inspiration.

At least, that's HQ's take on the matter. So they aren't exactly caring about the competition part being good (I mean, it is a Championship, so they have to, but it's a lower priority than inspiration).

Yep, It's like they don't care about the robot competition that much.
+1

[AndrewPospeshil]

I was thinking about what the most lopsided divisions were since the CMP split in 2000, so I wrote a script to calculate the number of standard deviations away from the CMP average for each division's average OPR, as a rough metric of relative division strength:

(Average OPRs are pre-CMP for 2016, others are during CMP)

<snip>

This is a fairly rough metric, with some possible bias towards the 8-divisional format due to a larger n and therefore a smaller standard deviation, but I think it's a good way to compare across years.

Mostly for my own sake, I went and bolded the World Champs.


Rank Year Division, OPR STDVs away from CMP mean
1 2016 new 1.8792818365030972
2 2015 cars 1.5303350085259413
3 2005 arc 1.4719431806316507
4 2008 arc 1.275903384248631
5 2013 gal 1.216209400652486
6 2010 arc 1.193897925865034
7 2009 new 1.0851114986838635
8 2006 new 1.0248700734203415
9 2014 arc 1.0038012569280608
10 2004 arc 1.0012622874746064
11 2012 arc 0.9280026009838416
12 2011 cur 0.8605362826161431
13 2007 new 0.811482461068815
14 2012 new 0.791654392743334
15 2011 gal 0.723890899467247
16 2004 cur 0.6810947735598231
17 2006 cur 0.6086265070047694
18 2009 cur 0.5683687825075886
19 2014 new 0.5614091258588676
20 2007 gal 0.5525768854103231
21 2016 hop 0.5380374488668379
22 2016 cars 0.408018860269447
23 2010 cur 0.3624233446672217
24 2016 carv 0.33274494055516757
25 2015 hop 0.24442033949382544
26 2013 arc 0.2390748663971342
27 2015 tes 0.23375807978480515
28 2008 new 0.17118024613308042
29 2015 new 0.1671052847394964
30 2015 cur 0.14696320018600914
31 2007 arc 0.06155154314818187
32 2015 carv -0.08427736320166529
33 2015 gal -0.12364688580536025
34 2005 cur -0.24293216111580532
35 2013 cur -0.2777800696315326
36 2016 arc -0.2968187516006284
37 2011 new -0.302185773307164
38 2014 gal -0.30707335292203575
39 2008 cur -0.33900159574861066
40 2010 gal -0.43743412228260514
41 2006 gal -0.4667057439069076
42 2005 gal -0.5166965248897325
43 2009 gal -0.5693576991292799
44 2004 new -0.597846349430966
45 2005 new -0.7123144946261157
46 2012 cur -0.7304368298598461
47 2016 cur -0.7416191862759174
48 2016 tes -0.9674409454187557
49 2012 gal -0.9892201638673295
50 2009 arc -1.0841225820621547
51 2004 gal -1.084510711603477
52 2008 gal -1.108082034633088
53 2010 new -1.118887148249653
54 2016 gal -1.1522042028992627
55 2006 arc -1.166790836518207
56 2013 new -1.1775041974180875
57 2014 cur -1.2581370298648897
58 2011 arc -1.2822414087762262
59 2007 cur -1.4256108896273216
60 2015 arc -2.114657663723039


5 above, 7 below. Pretty interesting.

[evanperryg]

60 2015 arc -2.114657663723039



Stop quantifying things I don't want to admit!

[BBaltrusch]

<snip>

One interesting note there are teams with negative OPRs and one is going to St. Louis. FWIW.

Dr. Joe J.



I had to look up that particular team. Won Chairman's in a week 2 regional, didn't have a 2nd event. They may not have had the best robot then but have definitely earned their spot. I don't know what kind of improvements they'll put in their withholding, but they've had 6 weeks to prepare.

[Jay O'Donnell]

Of course the three divisions I've been on are ranked 1st, 2nd and 11th...

[wjordan]

Mostly for my own sake, I went and bolded the World Champs.


Rank Year Division, OPR STDVs away from CMP mean
1 2016 new 1.8792818365030972
2 2015 cars 1.5303350085259413
3 2005 arc 1.4719431806316507
4 2008 arc 1.275903384248631
5 2013 gal 1.216209400652486
6 2010 arc 1.193897925865034
7 2009 new 1.0851114986838635
8 2006 new 1.0248700734203415
9 2014 arc 1.0038012569280608
10 2004 arc 1.0012622874746064
11 2012 arc 0.9280026009838416
12 2011 cur 0.8605362826161431
13 2007 new 0.811482461068815
14 2012 new 0.791654392743334
15 2011 gal 0.723890899467247
16 2004 cur 0.6810947735598231
17 2006 cur 0.6086265070047694
18 2009 cur 0.5683687825075886
19 2014 new 0.5614091258588676
20 2007 gal 0.5525768854103231
21 2016 hop 0.5380374488668379
22 2016 cars 0.408018860269447
23 2010 cur 0.3624233446672217
24 2016 carv 0.33274494055516757
25 2015 hop 0.24442033949382544
26 2013 arc 0.2390748663971342
27 2015 tes 0.23375807978480515
28 2008 new 0.17118024613308042
29 2015 new 0.1671052847394964
30 2015 cur 0.14696320018600914
31 2007 arc 0.06155154314818187
32 2015 carv -0.08427736320166529
33 2015 gal -0.12364688580536025
34 2005 cur -0.24293216111580532
35 2013 cur -0.2777800696315326
36 2016 arc -0.2968187516006284
37 2011 new -0.302185773307164
38 2014 gal -0.30707335292203575
39 2008 cur -0.33900159574861066
40 2010 gal -0.43743412228260514
41 2006 gal -0.4667057439069076
42 2005 gal -0.5166965248897325
43 2009 gal -0.5693576991292799
44 2004 new -0.597846349430966
45 2005 new -0.7123144946261157
46 2012 cur -0.7304368298598461
47 2016 cur -0.7416191862759174
48 2016 tes -0.9674409454187557
49 2012 gal -0.9892201638673295
50 2009 arc -1.0841225820621547
51 2004 gal -1.084510711603477
52 2008 gal -1.108082034633088
53 2010 new -1.118887148249653
54 2016 gal -1.1522042028992627
55 2006 arc -1.166790836518207
56 2013 new -1.1775041974180875
57 2014 cur -1.2581370298648897
58 2011 arc -1.2822414087762262
59 2007 cur -1.4256108896273216
60 2015 arc -2.114657663723039


5 above, 7 below. Pretty interesting.
I suspect there are 2 reasons:
1. There's a lot more randomness on Einstein, with all of the alliances there being roughly equal.
2. The powerhouse teams capable of leading an Einstein-winning alliance are more likely to be upset in stronger divisions, leaving the gap open for strong teams in fairly weak divisions.

Stop quantifying things I don't want to admit!
My (former) team was in that division, too ;)

[wazateer1]

I suspect there are 2 reasons:
1. There's a lot more randomness on Einstein, with all of the alliances there being roughly equal.
2. The powerhouse teams capable of leading an Einstein-winning alliance are more likely to be upset in stronger divisions, leaving the gap open for strong teams in fairly weak divisions.


There are actually several interesting benefits to being in a weak and strong division. My team (3506) is in Curie, which as shown is below the mean. While Pre-Scouting, we have noticed that there are some powerful teams (maybe even enough to make an Einstein winning alliance? :yikes: ). But there are not enough teams that will clearly swing qualification rankings in their favor every game. This means that, unlike a division where the entire top 8 will probably have an average of 3 or more ranking points per game (this happened in the FiM, MAR, NE, and PNW District Champs and almost happened in Indiana State Champs), there will not always be a powerhouse team on an alliance for teams to get carried by. In any one match, there is a chance that there will only be one competitive team on the field.

In a division like Newton, there are almost too many competitive teams. You can be sure that some teams that deserve to do well will not. Also, teams that are not that competitive will have almost no control over their ranking points. For them, it will all depend on whether you are going against this powerhouse team with that powerhouse team, or if you did (not) get lucky and the powerhouses are only on one side of the field. In the NC district anyway, Qualification matches in Stronghold are a two team game. Two extremely competitive teams could be paired with a box bot, and still Breach and Win (and possibly Capture, we will have to see how teams deal with the tower strength increase) against three moderately competitive teams.

The biggest side effect of being in an extremely competitive division I could foresee is Scorched Land. You can't form your perfect alliance with your first pick, because can't let any other alliance have this or that team. Once you get up to the best-of-the-best robots in Stronghold, they can pretty much do anything, but you still might end up with, say, three high-goal robots unable to cycle through the low bar, and only one which can go over the rock wall or cheval. These high goal shooters, paired with another low-bar breacher/low-goal scorer, may have been even more competitive, but from a statistics standpoint it was the best way to rob other alliances of Offensive Power. I may be underestimating the versatility of many robots, but I can see an advantage to alliance selections being more about forming a strategically perfect alliance then a statistically perfect one.

Those are just some of the pros and cons I see of being in stacked divisions.

[messer5740]

I believe OP is referring to the strong Newton division this year.

Many of the teams in this division had participated in the Greater Pittsburgh Regional. They are some of the best teams out there. Good luck to all those competing in champs!

[Lij2015]

Making all the divisions equal would take some of the fun away.

I can make an argument that there really isn't a top 4, more like a top 7 or 8 (considering I think people are underestimating 1241, 1519, 125, and probably 5172)

When I was doing a fantasy draft with some friends we had a really hard time picking who we thought was going to win Newton, and combination of those teams are going to make the finals really interesting. However I do see how this could be problematic. Some teams that in weaker divisions might be a great candidate for 2nd pick run the risk of going to the "backup" round.

However, when your fantasy first draft looks like that you know it's going to be a fun time.

E1: 254 4678 1640
W1: 5172 217 1477
C1: 67 179 3476
E2: 16 3620 188
W2: 1241 125 126
C2: 1519 118 1676

[tstew]

Frank described how teams were assigned to divisions in 2014 in a blog post. I don't know if this process is the same process that was used this year.

How We Assigned Teams to Divisions

Wondering how we assigned teams to FIRST Championship Divisions this year? First, we assigned Rookies, taking all Rookies signed up for Championship and putting them team by team sequentially in Divisions - one team in Division 1, one team in Division 2, one team in Division 3, one team in Division 4, then back to Division 1 again - in the order in which they registered for CMP, until we ran out of Rookies. This made sure no Division was over- or under-weighted with these least experienced of all FRC teams. Then, we took the Veterans and did the same thing. That’s it.

- See more at: http://www.firstinspires.org/roboticsprograms/frc/blog-Divisions-and-Preliminary-Match-Schedules

[Nathan Streeter]

I find it fun and interesting to be in one of the strongest divisions, but I've felt like 1519 is usually in the strongest division! We've attended CMP 9 times and 5 of those divisions are among the 9 most competitive divisions of all time (and 7 of 14)!

Our CMP divisions:


Rank Year Division, OPR STDVs away from CMP mean
1 2016 new 1.8792818365030972
2 2015 cars 1.5303350085259413
6 2010 arc 1.193897925865034
8 2006 new 1.0248700734203415
9 2014 arc 1.0038012569280608
12 2011 cur 0.8605362826161431
14 2012 new 0.791654392743334
18 2009 cur 0.5683687825075886
26 2013 arc 0.2390748663971342


Also, we've been in the most-competitive division for that year every year except for 2012 Newton, 2009 Curie, and 2013 Archimedes... when we were on the second-most competitive division. Hah!

Our division is on average 1.01 Standard Deviations above the CMP mean! Maybe we could be in a "weak" division some time just to see what it's like? ;-)

[efoote868]

Only 4 teams from each division make it to Einstein. Only 3 of those 4 teams on an alliance from each division are on the field at once. IMHO this just means that a non #1 seed is most likely to emerge from Newton, and that Einstein will be fairly evenly matched.

[Richard Wallace]

... a non #1 seed is most likely to emerge from Newton, ...

I think this could happen, but disagree that it is most likely. The way to bet would be on a #1 alliance that includes 254 moving on to Einstein. The seven other alliances will be very strong and will do their best to achieve a different outcome, but the favorite is still the favorite.

If you are just saying the favorite's chances are less than even, I agree with that.

[efoote868]

I think this could happen, but disagree that it is most likely. The way to bet would be on a #1 alliance that includes 254 moving on to Einstein. The seven other alliances will be very strong and will do their best to achieve a different outcome, but the favorite is still the favorite.

If you are just saying the favorite's chances are less than even, I agree with that.

My point was that if any division does not have a #1 seed advance to Einstein, it would be Newton. Not that the #1 seed from Newton was less likely than #2-#8 to win their division.

The point being that when the #1 robot picks the #2 robot in a shallow event, usually the #1 alliance will sweep.

[AndrewPospeshil]

Stop quantifying things I don't want to admit!

We need a "I was a member of the weakest ever CMP division" support group! Man, imagine if Bedford hadn't been in Archi 2016. I don't even wanna think about it.

[Richard Wallace]

My point was that if any division does not have a #1 seed advance to Einstein, it would be Newton. Not that the #1 seed from Newton was less likely than #2-#8 to win their division.

The point being that when the #1 robot picks the #2 robot in a shallow event, usually the #1 alliance will sweep.

Ok, now I see your point and agree.

[3a11]

We need a "I was a member of the weakest ever CMP division" support group!

Can we get a "I went directly from the weakest ever CMP division to the strongest ever CMP division" support group over here?

[BrennanB]

Being in a deep division just means you have a better chance at doing well.

Being in a weak division just means it's harder to get that golden second pick steal. (Because the 3rd robot is what wins/loses championships at this level.)

[wjordan]

We need a "I was a member of the weakest ever CMP division" support group! Man, imagine if Bedford hadn't been in Archi 2016. I don't even wanna think about it.
Then it'd drop to (roughly) 2.6 standard deviations below the 2015 CMP mean, so let's not.

[Citrus Dad]

Having lopsided divisions is just a side-effect of random (or, at least, semi-random) assortment.

In my opinion, splitting up and assigning teams to divisions based on their performance would make CMPs less interesting.

Champs is fun because of the challenging schedules and obstacles it imposes on the teams who make it there.

This is all my opinion, though. My word is far from factual :D

It's important not to confuse "random" with "fair" or "balanced." Random is a description of a process, not a value statement on an outcome. I think of random in these situations more at "arbitrary" although its better for the distribution to divisions (with 600 teams) than for match scheduling (with only 10 matches).

Note that the NCAA faces the same situation in its basketball tourney and it spends substantial effort trying to balance the 4 sides of the brackets. It wouldn't take much effort for FIRST to do the same (and then we can bet on how many 12th seeds will beat 5th seeds...:cool: )

[evanperryg]

Then it'd drop to (roughly) 2.6 standard deviations below the 2015 CMP mean, so let's not.

The fact that it is affected that much by one team says something :p that division was pretty sad.

[Ether]

Related discussion going on over here (http://www.chiefdelphi.com/forums/showthread.php?t=147628).

[qzrrbz]

I think this could happen, but disagree that it is most likely. The way to bet would be on a #1 alliance that includes 254 moving on to Einstein. The seven other alliances will be very strong and will do their best to achieve a different outcome, but the favorite is still the favorite.

If you are just saying the favorite's chances are less than even, I agree with that.

it might be the case that a stronger division makes the fickle finger of scheduling malaise that much more potent, as one star + two less stellar up against 2 stars + reasonable is more of an ambush. i'd argue that makes the "favorite" less odds on to run the table and end up as #1 seed. off to do the numbers to prove that :rolleyes:

[JeffB]

It's important not to confuse "random" with "fair" or "balanced." Random is a description of a process, not a value statement on an outcome. I think of random in these situations more at "arbitrary" although its better for the distribution to divisions (with 600 teams) than for match scheduling (with only 10 matches).


It's really not all that important. The argument posed in this thread is the random process is the most fair process. You can evaluate the process and make a value statement based on that process versus other possible ways to split the teams. As even you admit the small sample size makes it difficult to assign quality scores to teams, that makes it much more difficult to find a measured solution that treats teams fairly. Without having a good system to do so, it's much more fair to remain random.

[Citrus Dad]

It's really not all that important. The argument posed in this thread is the random process is the most fair process. You can evaluate the process and make a value statement based on that process versus other possible ways to split the teams. As even you admit the small sample size makes it difficult to assign quality scores to teams, that makes it much more difficult to find a measured solution that treats teams fairly. Without having a good system to do so, it's much more fair to remain random.

I didn't say that the small sample size makes it difficult to assign quality scores to teams. While there is certainly some error margin around the final results in any regional, those results do a reasonable job of sorting teams from high to low. If you don't agree with that premise, then you have have to reject the notion that the Championships bring together the best teams in general and we should just simply randomly draw from all FRC teams.

If you believe that most of the best FRC teams are being qualified for Championships, then we can create the basis for ranking those teams using information from the competitions in which they qualified. The district points system, which is used to qualify teams for subchampionships, is a reasonable proxy for that ranking.

And as I stated, an equal burden of proof resides on those supporting the current random system to show that it is fair. No one has shown that's the case that I've seen so far.

[Christopher149]

60 2015 arc -2.114657663723039

Stop quantifying things I don't want to admit!

Is that why we managed to be an alliance captain?

[efoote868]

Thinking about division assignment once again, I think I disagree with the idea that there needs to be (if I can coin the phrase) parity of ability between the teams that make up divisions.

Robots and drivers are not static through the season, let alone a single competition. I think that makes the idea that the parity of ability between divisions can be equalized ridiculous.

[Left To Beaver]

Being in a deep division just means you have a better chance at doing well.

Being in a weak division just means it's harder to get that golden second pick steal. (Because the 3rd robot is what wins/loses championships at this level.)

Well, even 2015 Archimedes let 3996 get scooped up by Bedford at the 24th pick despite being a top-15 (and possibly top-10) robot on the division. Our alliance was able to pick 4213, a reliable two-stack chute loader, at the 30th spot. They were good enough that our alliance captain graciously chose to play them in our second elim match rather than themselves, and we put up 182 points without getting any center cans.

Scout well and there will be solid robots in the later rounds of any event. Also, after watching 110+ matches of that division, I feel I should get a spot on the support group steering committee :p

[ahartnet]

Not that we need another stat to show how stacked newton is, but through roughly 7 qual matches in all the divisions I notice that Newton has 7 teams with goal points exceeding 400 (and 1241 is pretty close to 500). The rest of the divisions have 2 teams (Hopper/Tesla/Carver) 1 team (Galileo/Carson) or 0 teams (Archimedes/Curie) that have over 400 goal points.

I'd be interesting to look at some of those stats after all the qual matches are done.

[IronicDeadBird]

http://resources.infosecinstitute.com/wp-content/uploads/121411_1611_SecureRando1.png

Oh RNG how you toy with me...

[Ether]

...

[CloakAndDagger]

Newton Out, Tesla is about to take it all, poofs got upset.

*Throws bracket out window*