|
|
|
![]() |
|
|||||||
|
||||||||
|
|
Thread Tools | Rate Thread | Display Modes |
|
#27
|
|||
|
|||
|
Joe (and friends;-)
No, this year, the match scheduling is definively not random. Unlike last year, secrecy is not needed, so we used a simple algorithm that yeilds a predictable but (usually) good schedule. I have always got my hand out looking for a better way of doing schedules. Let me lay out the schedule requirements first: (1) All teams must play the same number of rounds (with the fewest number possible "extra" matches). (2) A round (a set of matches where all teams play each other once) should basically be played to completion before the next round starts (this ensures that all teams play all day) (3) For each team, the inter-match spacing between successive matches should be as large and as regular as possible. There MUST BE at least 3 matches between succesive plays (if possible) (4) Teams should play with different partners as much as possible (5) Teams should play with different opponents. (6) Teams should play on different field positions through the day -- and probably some more constraints. This year, I create the schedule by splitting the field into 4 groups. Imagine them arranged as columns. Each row then forms a match. I then rotate the second column by 1, the third column by 2, and (Surprise!) the fourth column by 3. Form the next matches and repeat. There are a few more tricks in here. I have recently added an extra rotation step to help mix the lists when the number of teams (like 48) would yield a repeating schedule. Also, there is some trickyness on handling the end matches (when the number of teams are not evenly divisible by 4), but I think you get the idea here. This results in a schedule that is pretty good, but not always perfect. It is very good at getting an even number of plays, and most of the time, making the inter-match spacing very even. With some numbers of teams, it yields repeated partners and opponents (Sorry) but this has been improved in week 5 of regionals. It is also non-random and the patern is discernable by observation. After the schedule is formed, teams are randomly assigned into it. Sorry, conspiracy theorists, there is no checking to see if teams are vetrans or rookies -- just luck of the draw. My sincere apologies if the randomness went against you. It would be possible to implement this in a random fashion, or maybe as a genetic algorithm, but so far, a good solution in this form has eluded me. Mostly, it is difficult to get the inter-match spacing large for all teams. Maybe someone out there wants to contribute one? B. |
| Thread Tools | |
| Display Modes | Rate This Thread |
|
|
Similar Threads
|
||||
| Thread | Thread Starter | Forum | Replies | Last Post |
| Divisions are out!! | Joel J | Championship Event | 55 | 06-04-2003 01:34 |
| national competition divisions a little fishy | archiver | 2001 | 5 | 24-06-2002 03:06 |
| 4 divisions in florida... (was it my idea?) | archiver | 2001 | 7 | 23-06-2002 23:13 |
| See the DIVISIONS! | Digo | Championship Event | 2 | 16-04-2002 17:14 |
| Divisions | Tom | Championship Event | 2 | 10-04-2002 21:23 |