Week 2 Regionals: Veterans Still Have No Advantage

[JamesBrown]

Alright I finally got around to investigating a hunch I have. The results we are seeing, the fact that team number has little to do with record has nothing to do with the game or how good the new teams are.

In each match with the exception of the last one or two in each round (ie when all except 6-12 teams have played the same number of matches) the sum of the team numbers on the red alliance is very close to the sum of the team numbers on the blue side. I am writing a program now to gather data from all of the regionals but from the three I have looked at it seems that in order for one alliance to win they must give losses to three teams with simmilar numbers. This Prevents any group of numbers from getting significantly more wins than any other. It seems that any data you have collected relating win percent (or seeding) to how well these groups are doing is invalid.

There really isn't any merit to any of the claims in this thread.

[dtengineering]

Originally Posted by JamesBrown

In each match with the exception of the last one or two in each round (ie when all except 6-12 teams have played the same number of matches) the sum of the team numbers on the red alliance is very close to the sum of the team numbers on the blue side. I am writing a program now to gather data from all of the regionals but from the three I have looked at it seems that in order for one alliance to win they must give losses to three teams with simmilar numbers. This Prevents any group of numbers from getting significantly more wins than any other. It seems that any data you have collected relating win percent (or seeding) to how well these groups are doing is invalid.

There really isn't any merit to any of the claims in this thread.

While I agree with you that having a seeding algorithm that averages team numbers across each alliance prevents any one group of numbers from getting significantly more wins than another, I disagree with your conclusion that this makes the data invalid. In fact, that is the whole point of the data!

Allow me to demonstrate by making two assumptions:

1) A team's performance (robot quality, strategy, resources, etc.) should be at least roughly reflected in their performance in qualifying matches. Note that I'm not saying that this is happening this year, but rather that, in general, your top seeded teams should be relatively better than lower seeded teams.

2) That veteran teams should have some advantages over rookie teams. Not, perhaps, an insurmountable advantage, but one would hope that as teams develop resources and learn more about robotics that they should become better each year. This is pretty much implied, if not explicitly stated by FIRST and is the logic behind the "rookie only" awards at each regional.

Now if one were to take those two assumptions, one could hypothesize that veteran teams should rank higher in qualifying matches than less experienced teams. The fact that this is not happening (at least as far as I can tell) is really rather important.

It suggests that at least one of the two assumptions is flawed. So either teams don't improve with time, or else the qualifying rankings do not adequately reflect the quality of a team and their robot. If we don't get better... why bother? And if the better teams aren't on top (in general...) then why do we give awards like "top rookie seed" and allow only the top eight alliances to pick their partners?

To be clear, I did not mean to suggest that veterans had absolutely no advantages... merely that they demonstrated no statistical advantage in qualifying matches, data that poses a very signficant and valid challenge to the two assumptions listed above.

If FIRST were to implement a scheduling algorithm that ignored team number then we could run a test to see which -- if either -- of the assumptions are invalid. (And probably end up with a much better representation of relative robot quality following the qualifying matches.)

Jason

Just a quick P.S.... If the scheduling algorithm is going to be biased for team number -- or anything other than operational measures such as equal numbers of matches and reasonable breaks between matches -- then that should be stated in the rules at the beginning of the season.

[AdamHeard]

Well... I can certainly tell you that the LA regional winners were veterans.

254 and 330 for sure. I know Last year was 4's first season, but they had a large experience base that came with them; I can't them as a veteran team.

But, some high numbered teams had awesome robots.

1717 was the best vertical lifter I have seen. two platforms about 4'x4' that lift in 2 seconds.

1622 was also a very impressive scorer.

[dtengineering]

Taking a look at last year's second week regionals, I calculate that there is a .16 correlation between team number and seeding. This is a noticable difference from this year, where the correlation was -.026.

I did not do any tests to prove whether this is significant or not, however, it is much greater than this year's correlation, and looking at an x-y chart of team number and seed, you can sort of force yourself to start to see a trend. (nothing but the finest statiscal analysis from this guy!)

I suggest this adds weight to the "common sense" hypothesis that veteran teams DO have some advantages... certainly not overwhelming ones... but that experience DOES help, and that something in this year's game (perhaps the scheduling system) is preventing veteran teams from achieving their traditional success in qualifying rounds.

Jason