![]() |
Match Scheduling
I was thinking...
At the events, teams are randomly paired with other teams to create alliances and score as many points as they can. This means that some teams will get put with other teams multiple times, and other teams will never see that team. This puts some teams at a disadvantage. If they never get matched with the best robot, there score will probably never be that good. So for the teams that were paired multiple times with that robot, their scores will be so much better, and if they are continuously paired with robots better than them, their ranking will actually be better than they deserve. What I am proposing is to make the competitions so that each team gets paired with each of the other teams once No more, no less. Everyone is on the same alliance with all of the other teams at one point or another. This would show the true value of the robot of the team instead of the possibility of a robot being carried into a high seed simply because of their alliance partners in Qualifications. I realize this would result in longer competitions and scheduling would be more difficult, but that would be a true and fair representation of the robot instead of the robot's alliance partners. |
Re: Match Scheduling
Quote:
|
Re: Match Scheduling
At District events, where the standard is 40 teams, we would need to play 20 matches per team to ensure that every team gets to play with every other team. That would add 67% to the qualifying schedule, requiring an additional day.
Probably not going to happen. |
Re: Match Scheduling
While I would have loved to play the required 32 qualifications to make this possible at the Lake Superior Regional, we only had time to play 9 in our normal 3 day event. A week of qualification matches would sure be fun, but I doubt many schools would like how many extra missed days this would add up to :p
|
Re: Match Scheduling
Quote:
Definitely not going to happen. |
Re: Match Scheduling
Quote:
1. The scheduling algorithm already does the best it can in this regard, but there are other constraints that are even more important (available time, times between matches.) 2. Life isn't fair, FRC isn't perfectly fair either 3. Good scouting should cause the best teams to get to play in eliminations, regardless of the shortcomings of scheduling. |
Re: Match Scheduling
If you're relying on being partnered with the best robot at the regional for your average to be high, you're doing it wrong.
|
Re: Match Scheduling
Quote:
I don't think your description of the algorithm is correct. The cost function at the core of the optimization procedure prioritizes minimizing repeated alliance partners over repeated opponents. Time between matches is not part of the cost function at all, but the algorithm only considers solutions that satisfy a minimum match separation. After generating the schedule the FMS provides overall statistics, which in my experience at District sized events does typically show the maximum number of distinct partners for most teams. |
Re: Match Scheduling
Quote:
Quote:
|
Re: Match Scheduling
At WVROX, we played just about every single iteration of the 24 teams on the field. Granted, running 24 hours will do that for you. :P We had a long cycle time, but still ran so that every team played against and with every team, multiple times.
As far as my experience, at most events, you will not see repeats. I think the only repeats we have seen on our team that I can think of were due to surrogate matches. At most events, we don't even get to see every team on the field (whether for or against), so we actually see no repeats. |
Re: Match Scheduling
At the five events I worked at I can't recall any repeats.
In fact the last time I recall something like that happening was 2007 and I believe it caused FIRST to fix the scheduling algorithm so that wouldn't happen anymore. |
Re: Match Scheduling
Good memory, Ed. 2007 brought us the Match Schedule Algorithm of Death. Paul and John were very quick to bring the issue to my attention as a Week 1 FTA that year. It caused quite a stir, and everyone I have talked with was very glad to see FIRST HQ correct it.
Quote:
Jim Zondag will gladly say more about this, if asked. He has a lot of data. |
Re: Match Scheduling
Quote:
Rankings only make a difference if you are one of the top 4 (who gets to do the picking from a wide open field). Aside from that, it is all scouting. |
Re: Match Scheduling
Quote:
|
Re: Match Scheduling
While I generally agree with the new algorithm and wistfully consider the possibility of playing against every team and with being allied with every team, I have seen personally a number of times when it appears the current algorithm breaks.
Last year we had a schedule where we were paired against the same team 3 times and never with them in our alliance. This year it happened again with the two highest seeded teams (by unlucky coincidence), we were only paired with one of the top two teams once, but played against them multiple times. I understand the algorithm and the concept of the optimization methods. However in this case it appears that the cost function must have really been skewed to generate these schedules. I need to build an analytic model of the cost function and see what exactly happened. Maybe when I get some time to breath again :) Enjoy! |
| All times are GMT -5. The time now is 05:51. |
Powered by vBulletin® Version 3.6.4
Copyright ©2000 - 2017, Jelsoft Enterprises Ltd.
Copyright © Chief Delphi