"Random" match Schedules

[Tristan Lall]

I was just looking at the St. Louis Regional thread, and after reading RoboPhantom's post, I think I see a problem in the way the surrogates are assigned. Would anyone care to check my reasoning?

Basically, it looks like the rules require surrogates that aren't necessarily needed, and the number of surrogate assignments in St. Louis was excessive.

[Dan-o]

People keep asking about FIRST's intent with the programming change. In 2005 (at smaller regionals), situations arose in which teams would have to play matches that would be only 2 apart from one another. If everyone looks at their match schedules from week 1, it is apparent that FIRST has done everything they could to maximize the time between matches where the same teams are involved (hence giving teams more time to repair and prepare). The repeated comments of people having match schedules of #9, #19, #29, #39.... etc. tells us that they have succeeded in this optimization.

Unfortunately, we have a second equally important constraint in which teams do not want to have to play with or against the same teams for every match. So now we have 2 separate constraints. First is the time between matches and second is the number of common partners. From this, we want to minimize the following:

Min: (time between matches) + (number of common teams from previous matches)

To successfully complete this, we have to apply weights to either side (to use the weighted sum method).

So the equation becomes:

Min: a*(time between matches) + b*(number of common teams from previous matches)

where "a" and "b" are weights and where a+b=1.

In 2005, it seems that a higher value was set for "b" than "a." Conversely, it seems that "a" was maximized for the week 1 2007 match schedules. The solution would involve finding appropriate values for "a" and "b" that bring the solution to somewhere more reasonable.

Unfortunately, this analysis is just that... and analysis. I have absolutely no understanding of how to set up team schedules based on this method, only how to evaluate how good a given match schedule is.

Also, Lucas. That idea sounds amazing. If generic schedules were made for every possible number of teams attending a regional (say from 24 to 80), then these could be used based on how many teams register. For people who are afraid that this would eliminate randomness and that you'd play the same teams every year at the same original... have no fear. At the drivers meeting, teams could have a representative pull a number from a hat held by the head referee to determine their seeding into the "optimized pairing matrix."

Just my 2 cents.

[Po-ser]

[This is long, so I'll bold the important parts.]

I think that the original idea was probably to space the matches apart as far as possible (something my team noted happily), but we didn't appreciate it when, after we complained to the organizers, they handed out "entirely new" match sheets, asking us to rip up our old ones in front of them first, in which the only change was that teams previously listed under "blue alliance" were now under "red alliance." I would rather have been told that there was nothing they could do rather than being tricked. I don't really know what the reasoning behind the new match sheets was, and I don't mean to whine, but it really wasn't very nice.

Also, I think the most important side effect of the new pairing system was stated at the top of page four by Cuog (http://www.chiefdelphi.com/forums/sh...t=55178&page=4): at Trenton 694 faced 637 for all seven of our matches. 637 was a purely defensive robot, and we formed our match strategies around the fact that we would be defending our strongest cappers against them every match. It didn't help that we were conspiring against them (and probably vice versa) and our pits were right next to each other. I thought that they were a great team with really nice people and I really hope nobody holds any grudges, because I know that there were people on my team that regarded them very aggressively during matches (and vice versa). I feel like the point of these competitions is to make people feel good about their robots, not bad about other robots.

****Most importantly, we should aim to win matches based on the strengths that we built into our machines. We shouldn't have to win them by picking on other robots' weaknesses. We didn't get to cap a tube or deploy our ramps successfully more than once because we had to play defense, and this was what we had built our robot to do.


637 and 694 will see each other in at least one other regional, and I really hope the algorithm for match selections changes by then. I think the best and quickest solution would be to revert to last year's algorithm. Perhaps there are pros to the new system, but the cons outweigh them by far. I'd rather have three minutes between my matches than have to spend a year raising money just to face one other team again and again.

[AdamHeard]

C'mon Dave, please comment on this thread and provide us some insight to the intent and/or if there will be any changes.

[deshirider430]

612 played 611 every match, we were so unhappy because team 611 was such a great team and didn't get a chance to show off there robot as much as they could of.

[Bharat Nain]

Originally Posted by Brian C

One other thing to consider is that FIRST likes to have rookie teams on an alliance with veteran teams so that they can gain experience. You also want to avoid having an alliance made up entirely of rookie teams.

While those are some great thoughts, it makes the regional unfair without an effort of being fair either. By following nuggetsyl's system, they are bound to end up with atleast one veteran team in atleast one of their matches. It is more preferable to the majority of the teams and makes the rankings at the end of the day much more fair. The alliance pairing was horrible but I can say one thing. If there is one team I have scouted very well, it is our fellow team 11. They were in the pits right next to us and against us on the field. I am sure they can say the same for us .

I'm really hoping this problem gets fixed. I do call it a problem, because it ruined the regional for me. I was very thankful to be chosen for the Elimination rounds, but the composition of teams selecting alliances was baffling. On there was a team whose arm scored a total of 4 tubes the entire Qualifiers, and whose drivetrain was able to be pushed around by a bot on Mechanums. Our team was paired every match with a defensive powerhouse, which made every match into a pushing game. It got to the point where we said as we went to the filed "Oh goody... time to wear down some more expensive tread for nothing."

As for the scoring system on the field that I've heard mentioned, it worked very well, so mad props to FIRST for that.

But I sincerely hope this issue gets resolved for later regionals...

[Dasistmeinmoped]

In NJ,
25 vs 11, 3 of 6 matches.
25 vs 41, 3 of 6 matches.

(im on 41)
ALSO the other 3 matches we played 75. Every other match we were paired AGAINST both our pit neighbors and 2 of the top 8 teams. We finished last, yes our robot failed this year, and our crab drive was a flop, but this was just a great way for first to kick sand in our faces.

0-6-1
'07 Judges Award Winners (hah)

[Guy Davidson]

Like most other voices on this thread, I think there is a big problem with the way the system is currently set. Teams should not be forced to compete against similarly numbered teams every time. This is not what we paid for, and not what we worked a month a half during build season for. Teams deserve to compete against all opponents, to get a feel for what many teams can do.

I also believe that writing a new algorithm before the next set of regionals is not only possible, but also a necesity. I am currently working on a new algorithm myself. I don't expect my code to be used - I have under a year of experience in Java (in which I'm writing the code) and only a few years of programming experience overall. My belief is that if I can write a code that will behave correctly in a few hours today and some testing tomorrow, then someone of the hunders of professional programmers, or the thousands of college students, or the dozens of thousands of high school students that are affiliated with FIRST can write something better than I did. And hence, there is no reason for the current algorithm to be used for another set of regionals.

I will post my code tomorrow, after I get around to testing it, if I believe it is worthwhile.

[Craig1989]

Has/Is anyone planning to contact first about this issue? We went up against 540 every single round in the VCU competition and while we only lost a couple of times we really disliked this schedule and would hate to be in the rough side of the deal. (It worked out ok for 540 in the end). Additionally I know for a fact that 611 was a good robot, but 612 was simply a huge ramp and there was very little 611 could do, hence they lost almost every round. Does anyone have any plans to try and do something about this before the next set regionals?

[Bongle]

Craig has a point. Everyone who was affected by this and think it should change should probably write their local FIRST coordinator and express their concerns. If enough people do it, perhaps they'll be able to adjust the algorithm by next week. Depending on how it is written, it may be as simple as adjusting some constants. It is entirely possible that the right people don't read Chiefdelphi on a regular basis and furthermore may not be reading this thread.

Contact info:
States A-L
States M-O
States P-Z
Outside the US

[AdamHeard]

Originally Posted by Bongle

Craig has a point. Everyone who was affected by this and think it should change should probably write their local FIRST coordinator and express their concerns. If enough people do it, perhaps they'll be able to adjust the algorithm by next week. Depending on how it is written, it may be as simple as adjusting some constants. It is entirely possible that the right people don't read Chiefdelphi on a regular basis and furthermore may not be reading this thread.

Contact info:
States A-L
States M-O
States P-Z
Outside the US

Just sent an email to the LA regional planner; I will post any insight I receive.

[AcesPease]

Originally Posted by Bongle

Craig has a point. Everyone who was affected by this and think it should change should probably write their local FIRST coordinator and express their concerns. If enough people do it, perhaps they'll be able to adjust the algorithm by next week. Depending on how it is written, it may be as simple as adjusting some constants. It is entirely possible that the right people don't read Chiefdelphi on a regular basis and furthermore may not be reading this thread.

Contact info:
States A-L
States M-O
States P-Z
Outside the US

I just sent a message to the CT contact. I will try to contact some other people I know on Monday.

[Guy Davidson]

I emailed the Northern California contact, and will post any response I get.

[AdamHeard]

Originally Posted by AdamHeard

Just sent an email to the LA regional planner; I will post any insight I receive.

Jim Beck promptly replied (for LA and San Diego) saying he would look into it. Hopefully they can resolve this.

I still wish someone high up like would give us some information on this.