I was playing around with the scheduler last night and I noticed the same thing. It think Sean from 25 posted it in the Galileo thread as well.
Quote:
Originally Posted by Joe Ross
One very easy solution would be to to randomize the team list before it is fed into the MatchMaker software.
|
Looking at regional schedules, I would guess the software already randomize the team list (don't you agree?) Remember we are playing with an evaluation version of the scheduler algorithm with no GUI or anything, not the actual software used by the scorekeeper. Hopefully they don't put them in OPR order (which looks like what you did) since that would lead to
conspiracy theories.
Quote:
Originally Posted by Joe Ross
My hypothesis was that the algorithm was not able to stray far from the initial set of teams because the minimum match separation was too high to allow that. If the minimum match separation was lowered, so should the correlation between the position and the initial list and the times played.
MMS = Minimum Match Separation
Code:
MMS Correlation
12 -0.88
10 -0.88
8 -0.60
6 -0.17
2 -0.11
When looking at the MatchRater results, MMS of 12 gives good results, and MMS of 10 gives almost perfect results. As a result, I have a feeling that FIRST would chose 10 for the championship. However, that value still has a strong correlation to placement in the team list and times played. It means that team 25 will very likely not play either with or against any rookie teams, but very likely they would play teams 8 and 40. Each individual match schedule is very good at making sure that you don't play the same team twice, it just limits the number of teams you have a chance of playing. |
I think this is the real prob is
the default MMS is too high and that is what is typically used (I think it was used at FLR and Philly). If you have to play a single team (out of 86) twice (in 7 matches) that is unacceptable and can heavily skew the seedings (which are very imperfect to begin with).
Hopefully the scorekeepers will tweak the MMS until there are no repeats.
Quote:
Originally Posted by Joe Ross
Note that in the grand scheme of things, this is a very small problem. The MatchMaker software is far better then any other scheduler I've seen in FIRST. I just want to make it better.
|
I like this
algorithm and how it approaches the problem. I love how you can tweak all the variables and they released it for us to evaluate (fun and educational). It may not be perfect, but it is a thousand times better than the Algorithm of Doom last year. By this time last year, we had released the first round of matches for each division.
