paper: Alliance Seeds and Results
 

Thread created automatically to discuss a document in CD-Media.

Alliance Seeds and Results by BMSOTM

Re: paper: Alliance Seeds and Results
 

I am really interested in getting everyone's opinions on this, I always figured the serpentine draft helped to equal out the playing field but from this it looks like it is not the case(at least for this year would love to see this done for previous years). This is especially true since of the two 8th seeds to make it to finals at a competition one was the Orlando 8th seed alliance (which I was a part of) who would have been eliminated.

Is it time for a change of draft structure? If so how could we improve it?

Re: paper: Alliance Seeds and Results
 

I feel like this year with so much dependability on a full working team, that the serpentine draft did a good job.

At both regionals we attended, we were the 8th seeded alliance, and came within just a couple of points of beating the #1 seeded alliance both times, Giving the #1 seeded alliance the hardest fight they saw.

At Texas Robot Roundup the lower seeds eliminated at least 2 of the 4 higher opponents, and at RoboReboot we were 5th seed (there were only enough teams for 5 seeds so we were lowest) and still won the event, beating #4, #1, and #2

This year you could do a lot of damage with 3 fairly competitive robots compared to 2 great ones and a seed #24.

I have heard some reasonable suggestions before, but what is the better alternative?

Re: paper: Alliance Seeds and Results
 

I never thought lower seeds would have more success than higher seeds, but perhaps more success than they've had in previous years. Would be interested to compare a chart like this to one from 2013 or the past few years. It'd also be interesting to look at the same results from different sizes of event, smaller vs. larger.

Re: paper: Alliance Seeds and Results
 

There were a couple factors that caused me to want to look at this data. At MAR champs, we were on the #2 alliance, and we were eliminated by a strong 7th seed alliance. It made me wonder, how often does this happen?
I was also following the WVROX off-season, which was played by 24 teams. The #1, #3, and #4 alliances were all eliminated in quarterfinals. I was originally going to include event size data and look at that, but I thought this was plenty to look at. If you would like the event size info and maybe a graph or two, I can easily put that together.

Re: paper: Alliance Seeds and Results
 

Jim Z compiled previous years. This data set is more favorable to the lower half of the bracket than most years. Typically #7 is an absolute dead spot and there we're 4 victories that year.
I don't think anything needs to be done to help lower alliances. #1 earns the right to first pick. And the rest makes sense. Serpentine does help add some variability.

Re: paper: Alliance Seeds and Results
 

not that there is much skew as a result, but kind of thinking the CMP divisions are a different kettle of fish for this analysis...

the 32nd robot was a great robot at CMP -- what mattered was how the chemistry came together, not draft vagaries! :)

Re: paper: Alliance Seeds and Results
 

I don"t think that there really is a problem to fix. The #1 has earned the spot after 10 qualification matches, and eliminations aren't supposed to be a level playing field. Plus, i think that serpentine actually helps a ton, and we just can't see it. Imagine having #1 pick first in the 2nd round of selections as well! Without serpentine, #1 would probably win more like 75% of the time.

Re: paper: Alliance Seeds and Results
 

I believe the one #8 seed that won, beat the #1 alliance in the quarter finals. In those quarter finals, the #1 seeded robot had their electronics ripped out during match play and were extremely crippled in the first or second match.

It is still a testament that once they got past the #1 alliance, the next best alliance couldn't beat them.

Re: paper: Alliance Seeds and Results
 

At Center Line, when the number 8 alliance beat the number 1, the number 8 alliance was accumulating fouls against the number 1 alliance in every match (this was week 1 so some bugs in the fouls were present.) Such as driving underneath of appendages. The number 8 alliance gave the number 1 extensive damage, such as smashed frames, broken polycarbonate, and wrecked electronics. They had good chemistry, too, which definitely contributed to their winning alliance

Re: paper: Alliance Seeds and Results
 

It's not supposed to make the alliances even, it's supposed to make them more even than if the first alliance got 2 picks in a row. If we wanted it to be truly even their would be no incentive to seed higher because it wouldn't increase your chances of winning, suddenly everyone is sandbagging and throwing matches because it doesn't matter if you win.

Serpentine draft is good, there's a reason most other competitive sports use it.

Re: paper: Alliance Seeds and Results
 

And I thought I was being all original and insightful...
:p

Re: paper: Alliance Seeds and Results
 

This made me wonder about the average seeding of an alliance and how that impacts the outcome. Especially the impact of the 1st seed alliance picking the #2 seed team. I have seen it many times where the #2 "rejects" the #1 and the crowd making a big deal out of it. From this data I would think "rejecting" the #1 seed is a bad idea.

Re: paper: Alliance Seeds and Results
 

I think that this is a bad conclusion to draw from this data. Correlation is not causation. The first seeded alliance does not do well across the board because they happen to have a number 1 next to their alliance number, but because they tend to have better robots on their alliance than any other seed. Generally, the first seed has a pretty good robot, so it is beneficial for any team to accept their invitation. However, if the first seed has a robot that is not one of the top 2 robots at the event according to your pick list, there is no good reason* to accept their invitation if you are the 2 seed.

*Unless you are looking for an easier schedule in the quarter and semi finals.

Re: paper: Alliance Seeds and Results
 

If you do not believe that the first seed was qualified enough to hold that rank would you accept them? There are few cases ever where this situation is not true but generally speaking, the first seeded team has demonstrated the highest "skill" in the competition at the tournament whether that be through an excellent robot and drive team, or less likely, a great strategic team.
Also, while you may have an easier schedule through elims, would you rather get a blue banner under a team you believe will kill the competition or suck it up with someone else who YOU believe shouldn't be there and get a nice piece of hardware. I dunno about anyone else but Blue Banners are really nice and generally secure yourselves a spot at championships/

Re: paper: Alliance Seeds and Results
 

I'm assuming that you are using "you" to refer to me, since you responded directly to my post.

The number one seed at the event is literally the most "qualified" team at the event to be the number one seed. They earned their way into that seed by earning more qualifying points than any other team at the event.

What I think you are asking is something like "if you do not believe that the first seed is the best team at the event, would you accept them?"
My answer is that it depends, if I am a captain and I think there is a good chance my team will later in the draft have an opportunity to select a better team than the first seed, then of course I would have us decline. For example, if we are the fourth seed, and according to our data the first seed is the 7th best team at the event, then of course I would have us decline, because even if the first 3 alliances contain 6 of our top 7 teams (which never happens), the worst case for us is that we end up with the 6th best team.

I see that you are from Michigan, which implies that you have had much experience with the district system. While 12 matches per team at 40 team events might generally, in your experience, cause the best team at an event to be seeded first, 10 or less matches per team at a 60+ team event, in my experince, does not generally cause the best team to be seeded first. Of the last 4 regionals I have attended, only at 2 could an argument even have been made that the first seed was the best team at the event. Even in those two cases, it was pretty up in the air between 2-3 teams.

I personally would rather give my team the best possible chance to qualify for the championship event. Generally, I believe that this means playing with the best teams available, regardless of their seeds. I think teams that expend any effort trying to get on the "easier" side of the bracket are wasting their time. However, there are instances in which I might consider trying to end up on the opposite side of the bracket (e.g. at an event with both 1114 and 2056 in attendance).

I just mentioned this point in my previous post to qualify my statement that "there is no good reason to accept their invitation," like I said, I don't generally believe in trying to get on the "easier" side of the bracket.

No argument from me here. :)

Re: paper: Alliance Seeds and Results
 

I can offer a different viewpoint on this subject. However, before I do I must stress that in a venue such as a regional that boasts 60-65 participating teams the advantage of being the #1 Captain lessens considerably over regionals that have 40-50 teams.

**Note - I have not done any research, this is only conjecture!

In the case of MPLS North Star there are usually between 60-65 teams. Of these teams, there are clearly 5-8 very dominant teams that have competed throughout the qualifications. These teams have established themselves in a large field of players. If the top 8-12 teams did not do their necessary scouting, their 2nd pick could spell out their demise.

Since our inception in 2012 (first time competing in 2013), we were concerned with the #3 and #4 alliances. If the captains of these alliances had a strong scoring robot, they would be able to pick from the field 2 more versatile robots than the #2 or #1 alliances could. In 2014 that would prove this out.

As it turned out in our first season, the initial #1 captain/alliance 3928 chose #2 2574. The initial #3 967 chose initial #4 2175. When it came back around, we were ranked #20 and were chosen by the #2 alliance of 967 and 2175. We beat the #1 alliance in the finals (partly due to 3928's pick-up failing). Last season, we started slow with a 0-2-1 record and then did not lose again as we entered into the elims. Again we were picked by 967 and 2175 (we were punching our tickets to St Louis at this time); - but our #1 alliance was eliminated by the #4 alliance.

However, the #4 alliance picked a team in their second choice that wasn't even around... so they panicked. That opened the door for the #5 alliance to choose a better team in 3042 - and sealed our fate.

For two years in a row, the North Star Champion did not include the #1 qualification team. I think that in large part this is due to the large field. If it was a smaller field that had fewer great teams it could have ended differently.

I truly believe in the serpentine draft as it awards the #1 alliance their just reward. However, the #1 alliance captain needs to do their due diligence in scouting. The Champions of the North Star in 2013 can thank a bad performance by the #1 alliance - and the same thing can be said in 2014 when 2 of the robots failed in the same match paving the way for the eventual champions.

I see no reason to go away from the serpentine draft - it is only an impetus to grow the sport more in the regional.

Re: paper: Alliance Seeds and Results
 

auction draft....

Re: paper: Alliance Seeds and Results
 

Interesting, I've found the exact opposite thing to occur. In smaller events (like Districts), the only picks left at the end of the draft are often extremely poor while a strong lower seed can get 3 competent robots.

To look into this a bit more I modified Brian's spreadsheet to compare District events to Regional events. At Districts the 1st seed won 40% of the time, compared to 56.45% at Regionals. The 6th, 7th and 8th seeds won a combined 12.5% of District events, but only 4.84% of Regionals.

Even more interesting, a District 1st seed that makes it to the finals only wins 64% of the time, while a Regional 1st seeded alliance wins 77.78% of the time.


It doesn't seem like I can add an attachment to this post (I guess because it's a paper?), so pm me if you want the modified spreadsheet. Here are photos of it though: Regionals Districts

EDIT: The Region Championships are classified as Regionals in my spreadsheet

Re: paper: Alliance Seeds and Results
 

Can you elaborate on this? It looks like you're saying the #1 alliance deserves to get the last pick after the rest of the alliances have chosen their partners (ignoring the existence of fourth robots). I would consider that more of a penalty than a reward.

Re: paper: Alliance Seeds and Results
 

Someone mentioned earlier in the thread that at WVROX, three of the top four alliances lost in quarterfinals, and that there were 24 teams, on a serpentine draft. In my experience, there is usually a significant dropoff in robot quality somewhere on the back end of the serpentine draft (though this also depends on scouting abilities, more on that later), and at small events, this can spell disaster for top alliances.

On the flipside, consider the championships, with almost a hundred teams in a division, where really good teams like 971, 973, 16, 610, 1640, and 177 get picked up on the backswing of the draft (these in particular have a lot to do with scouting flukes, but good picks have been made on the back end before).

I have come to realize that for the best alliances to form, events need a) a good number of matches (to insure the best teams reach the top) and b) a good number of teams (to allow good second or third picks to form). Unfortunately, these two variables limit each other, so often events have to pick one over the other. WVROX went wrong in this respect in having only 24 teams with how ever many matches. Champs 2013 went wrong with this when they had 100 teams and only 8 matches. With the constraints in mind, the easiest way to optimize the alliances according to their rank would be to change the draft order depending on the event size.

Re: paper: Alliance Seeds and Results
 

Sorry, my mistake was in my choice of words. The larger the regional, the better for the 1st seed. The smaller the field, the less benefit for the #1 team.

@Alan Anderson - No, my meaning was that the 1st seed had proven it was capable of winning throughout the quals - no matter the make-up of either alliance. So the #1 Seed is deserving of choosing the best fit for their first pick. The larger the field, the better for the #1 seed.

I hope this clears it up! I really like the serpentine draft as it rewards the best performing teams a great first pick - but allows for a balance over the draft altogether!

Re: paper: Alliance Seeds and Results
 

It wasn't our pick-up that failed, it was our pneumatic system that became disconnected :P

From participating in 2013 Ultimate Ascent at KC and North Star, the main reason we picked how we did was that our partners fit in with our strategy.
The other thing that's special about Minnesota that I think doesn't help things very much is the fact that they have more FRC teams than hockey teams (but many teams -- although NOT ALL -- lack the support needed to have successful/quality teams/robots - mentors, resources, and knowledge - to not be overwhelmed). I know our mentors were helping a team put a kit bot together in their pit on Thursday. If anything, states that want to follow their lead should start building out their FTC programs like Iowa.

Overall, Minnesota is just an interesting state to look at as far as teams/regionals go. Don't they have like the most regionals in a state than any other state?

Re: paper: Alliance Seeds and Results
 

Texas also has 4, with far fewer teams. California has 7, also more per team than MN. Ontario has 5, with fewer teams than any of those 3 states.

We are packing our events to the brim (two at 60, two at 63). Those other states (provinces) can get away with event attendance as low as 30 or 40, and offer more 2nd plays than we can, both of which I feel make a much better experience for all teams. That along with the number of overwhelmed teams, we are definitely behind the curve ...