"Random" match Schedules

[wolfj]

Of course its not random. My team played in the 9th match of every group of 10 for the entire qualifier. They have to change it soon.

[AcesPease]

Originally Posted by wolfj

Of course its not random. My team played in the 9th match of every group of 10 for the entire qualifier. They have to change it soon.

Sounds like there is still a lot of emphasis on time between matches, but improvements have been made. As I understand how any program that sets the matches will work, if you limit the time between matches to 8-12 rounds (for a field of 60), then you will be "randomly" matched with no more than 30 different teams and probably less. It probably is not practical to have a 5 round spread for smaller fields, so the number of "random" teams will decrease dramatically on small fields. I recall being frustrated at a small field regional a couple years ago, because we faced or teamed with the same team many times and did not play quite a few of the other teams. Anyway, the matches seem more mixed when the opponents are allowed to shuffle. The perpetual opponent algorithm makes an imperfect system worse, in my opinion, so I hope the event staffs are using the other option.

Let's try to have some fun out there

[JA-Low]

Well both of first problems this year bit us. In our last 5 matchs we had the same alliance partner. They either could not move or did not show because of a failed banebot. This made us have a 3 vs 2 match and we lost 3 out of 5. We droped for 2nd in the rankings to 17 witch hurt in the final selections. Please make the match selections more "random".

[chinckley]

We saw one team 4 out of 8 matches and many teams 2 or 3 times, even in a row. There were many, many teams we never saw. People were talking about this at GLR. Also we had one match we were on the floor and they were calling Last call for our next match (only two between them). We told the person at the gate that we were still on the floor and would be back soon. That was a fast one.

[lenergyrlah]

I just ran a lot of statistics from the VCU Regional. There were 66 teams.

The oldest third & newest third were against the same team each round. The pairings were done in numerical order. Example: 116 vs. 122 every match, 339 vs. 343 every match, 345 vs. 346 every match.

Each team in the middle third alternated rounds against the team right above them and the team right below them (in numerical order). So team 928 alternated rounds between teams 900 and 975. Since team 620 was the oldest team in the middle third & team 1522 was the newest team in the middle third they alternated against each other. So team 620 alternated between 623 & 1522, and team 1522 alternated between 620 & 1413.

The only thing "random" was which pairs were matched against which, keeping in mind that the 3 pools were completely separate.

Team 1731 was matched against 1655, & in each match every team in our alliance was newer than their opposing team (i.e. 346, 928, & 1731 vs. 345, 900, 1655). If FIRST honestly thinks that the lower the team # the better the robot then we should have lost every single match. How is this fair?

[Nate Edwards]

Originally Posted by lenergyrlah

I just ran a lot of statistics from the VCU Regional. There were 66 teams.

The oldest third & newest third were against the same team each round. The pairings were done in numerical order. Example: 116 vs. 122 every match, 339 vs. 343 every match, 345 vs. 346 every match.

Each team in the middle third alternated rounds against the team right above them and the team right below them (in numerical order). So team 928 alternated rounds between teams 900 and 975. Since team 620 was the oldest team in the middle third & team 1522 was the newest team in the middle third they alternated against each other. So team 620 alternated between 623 & 1522, and team 1522 alternated between 620 & 1413.

The only thing "random" was which pairs were matched against which, keeping in mind that the 3 pools were completely separate.

Team 1731 was matched against 1655, & in each match every team in our alliance was newer than their opposing team (i.e. 346, 928, & 1731 vs. 345, 900, 1655). If FIRST honestly thinks that the lower the team # the better the robot then we should have lost every single match. How is this fair?

Sounds like what happened at PNW... Were the pits the same? (ie .. the teams that were always against each other removed by one pit)

ie team #1 (middle team #1) Team #2 (middle team #2)
Team #1 vs Team #2
middle team #1 vs middle team #2

At PNW it seemed that the top third and bottom third had the same problem... examples again: 272-360, 488-492, 957-997,956-955, 847-753, etc..... I am not sure of the specifics on the rookie teams but when we were alligned with them they seemed to be against one team alot.

Our team was in matches 70, 61, 52, 43, 34, 25, 16, and 7 ... Except for match 16 we were always on the red side... Always against 997

[Tom Saxton]

Lots of people on this thread have said they want a random schedule, but really what they want is a schedule that lets them play many different teams with as few duplicates as possible. This doesn't happen with a random distribution, it requires an algorithm that specifically builds a schedule to minimize duplication while keeping successive matches far enough apart that teams can turn their robots around.

FIRST now has two match scheduling algorithms.

Most of the week one regionals used what's being called "perpetual opponent" which does the scheduling where all of the teams in the high and low third (by team number) are paired with an adjacent team who is their opponent in every match. The middle third teams have two teams and alternate matches with them as opponents. There is no duplication of partner teams and great spacing between successive matches for each team.

After all of the complaints about the week one schedules, FIRST added a second algorithm called "random" which does a better job (but not a good job) of mixing opponents, but also duplicates partners and schedules matches very close to each other, even back-to-back.

Two of us (my wife and I) have implemented an algorithm that produces the type of schedules that people are saying they want: minimum duplication of partners and opponents, and good separation between matches. An analysis of the problem and our proposed solution are documented in a white paper on our team site. We'd love to have FIRST use our algorithm. We've sent them mail but haven't heard back.

[Michael Corsetto]

So, has anyone heard anything official from FIRST about the Veteran/Intermediate/Rookie constraint that they sort of snuck into the "random" match pairing system this year?

As I've stated earlier, from the perspective of a slowly-improving, learn-as-you-go, student run team, this new system for forming alliances takes our low team number way out of context. I consider every team as a worthy adversary, and I base my evaluation of their performance on their robot, not some arbitrary number that FIRST assigns every team that registers for competition. Sure, some teams might have past successes (254 comes to mind ), but that in no way means that they will have a killer robot this year (which they do, but thats beside the point ). And some teams, such as us, haven't even had "winning" success in past years, yet they treat us like we are the same level as the Poof's as one of the elite "A" teams.

If I was a rookie, I would be insulted that FIRST only considers you as a "C" team and that you need an "A" team to help you through your qualifying matches to make sure you don't lose badly every match. I saw plenty of inspiring rookie bots at PNW, 2046 sticks out to me, with their awesome autonomous mode that worked extremely well.

If FIRST was really trying to make this whole qualifying match thing fair, in order to rank the teams, they should go around the pits, look at every teams robot and see how they preform in their practice rounds. (Oh wait, isn't that what the scouts already do?)

I really don't see any legitimate reason for the qualifying match algorithm that FIRST has in place, and until they give concrete reasoning for this dramatic change, I will continue to try and bring back truly random qualifying matches. My team still has another $8,000 worth of regionals to attend this year, and I don't want to waste all that money on a bum schedule that limits what teams I will be able compete with and against.

I guess we'll all just have to wait until Update #17.

Mike C.

EDIT: So I guess truly random isn't quite what I'm looking for, more like random, but with a reasonable amount of space between matches (Mr. Saxton, very nice explanation of the yours and FIRST's algorithms, even I understood it . I really like the constraints you have in yours and I would love it if FIRST implemented it. Let us know what they say!). I don't see what's wrong with going back to the 2006 algorithm though. I understand that Hatch was in charge of it back then, but can't we just mimic that algorithm's results?

[huskyrobotics]

I agree that a wonderful program like first can be improved by looking at match scheduling. The issue that seems very fixable to me that has not been addressed is to insure that every team has at least two matches on any Saturday during a regional. I find it very unfair to families and friends of first teams that some schools have only one match on Saturday, while others have 3 matches. People do work and quite often Saturday is their only day to attend the regional events and cheer for their teams. If we want more support, we should enable the people of the community to support us.

[Jacob Plicque]

At the Flordia Regional Team 86 had several repeat pairings as follows:
Partners
1065 - three times
1592 - two times
Foes
1345 - two times
1649 - two times
There were 51 robots at this competition. I can believe that one repeat pairing is random, but this many repeat pairings points to a problem with the program.

[waialua359]

according to the algorithm, we will be playing 254, the cheesy poofs, almost every round in vegas!

[Travis Hoffman]

I'm thinking teams like 968, 1038, 1126, 1114, etc. are loving the fact FIRST considers them to be "mid" (or even high, depending on the event) level teams.....

The likelihood of them and any other quad digit team partnering with a solid low numbered veteran team is a lot greater than the likelihood of a 2 or 3 digit team being paired with a solid quad digit team.

So how is this algorithm benefitting all teams again?

[Jack Jones]

Originally Posted by T. Hoffman

I'm thinking teams like 968, 1038, 1126, 1114, etc. are loving the fact FIRST considers them to be "mid" (or even high, depending on the event) level teams.....

The likelihood of them and any other quad digit team partnering with a solid low numbered veteran team is a lot greater than the likelihood of a 2 or 3 digit team being paired with a solid quad digit team.

So how is this algorithm benefitting all teams again?

We loved it all the way into third seed at GLR.
Don't know about ALL teams, but sixth seed sure got a bonus from that.

[jagman2882]

Originally Posted by T. Hoffman

I'm thinking teams like 968, 1038, 1126, 1114, etc. are loving the fact FIRST considers them to be "mid" (or even high, depending on the event) level teams.....

The likelihood of them and any other quad digit team partnering with a solid low numbered veteran team is a lot greater than the likelihood of a 2 or 3 digit team being paired with a solid quad digit team.

So how is this algorithm benefitting all teams again?

maybe we will realize that advantage at buckeye....i didnt really notice it that much at FLR cause the teams we were continually paired with failed to show up to our matches.

[Alan Anderson]

Originally Posted by T. Hoffman

So how is this algorithm benefitting all teams again?

A match schedule should not benefit "all teams". It should definitely not benefit any team or category of teams in particular. It should benefit the competition.

The seeding matches are undeniably intended to produce a ranking of teams. It seems reasonable for that rank to reflect how well each team performs, with the better-playing robots at the top of the list. The "perpetual opponent" schedule certainly doesn't give that result. A very good team can be beaten by a marginally better team every time, placing the very good team near the bottom of the list.

Assuming that veterans are "more good" and rookies are "less good" is a shaky thing to do, but let's do it for the sake of argument. Any schedule which intentionally pits veterans against veterans along with rookies against rookies can skew the final rankings to carry a bunch of "less good" robots to the top of the list. And if the assumption isn't correct, such a schedule serves only to deny teams the opportunity to play against a wide variety of opponents.

I believe that qualification match alliances should be assigned entirely without regard for team age, team number, or prior performance. The goals should be primarily to maximize the number of different teams each team plays against and with, and secondarily to make the time between matches consistently long. There still can't be enough matches to establish a very high correlation between rank and robot "goodness", but it seems to have worked well enough in the past few years.