PDA

View Full Version : Match Scheduler Issues


James Tonthat
03-25-2011, 09:45 PM
So I was watching Waterloo Regional and I was noticing a couple interesting matches. 1114 had played 2056 three times today, two of those matches were next to each other. Last week at Lone Star Regional, we had played 624 two times with two of those matches being next to each other.

I've talk to some of my friends on other teams and one of the reasonings is "small regional" but LSR was a 55 team regional, and it seems pretty interesting for 2056 to play 1114 three times despite it being a small regional. Are we seeing another 2007?

Am I crazy or is there something going on? Share some of your interesting experiences.

JackS
03-25-2011, 09:50 PM
We had 4 matches against the top 3 teams at the regional. We played 40 back to back at WPI. We also played 195 and 177. Both of those teams were finalists or champions of the event. It was an extremely difficult schedule.

At the same time, 358 only played 40 once with the help of 177, and 195 once.

Just the luck of the draw I guess...

Andrew Schreiber
03-25-2011, 10:12 PM
http://forums.usfirst.org/showthread.php?t=7805 May shed some light on your questions if you have the time/desire to read through it. It details the 2008 scheduling algorithm. To my knowledge nothing major has changed (ie no return to the 2007 algorithm) . http://www.idleloop.com/matchmaker/download.php may also provide some information if you are willing to dig.

James, as I mentioned to you, after finals I'll poke around a little more if no one else has by then. (Only 2 more papers to write... and some hw)

Edit: Criteria from Game Manual


5.3.2 Match Assignment
The Field Management System (FMS) will assign each TEAM two ALLIANCE partners for each
qualifying match played using a predefined algorithm. The algorithm employs the following criteria:
A. Maximum time (in number of matches) between each match played for all TEAMS
B. Minimum possible number of times a TEAM plays opposite any TEAM
C. Minimum possible number of times a TEAM is allied with any TEAM
D. Minimize the use of SURROGATESs.
E. Even distribution of matches played on Blue and Red ALLIANCE (without sacrificing A,
B, C and D)
All TEAMS will play the same number of qualification matches except if the number of TEAM
appearances (number of TEAMS multiplied by number of rounds) is not divisible by six; in that case
the FMS will randomly select some TEAMS to play an extra MATCH. For purposes of seeding
calculations, those TEAMS will be designated as SURROGATES for the extra MATCH. If TEAMS
play a MATCH as a SURROGATE, it will be indicated on the match schedule, and it will always be
their third match

flippy147852
03-25-2011, 10:26 PM
I remember 74 had to play a very strong 67 three times whereas 2054 (#1 seed) had a relatively weak schedule.

James Tonthat
03-25-2011, 10:26 PM
I looked at our Lone Star schedule, this is what I found.

Match Breakdown - 1477 - Lone Star Regional

Team # W = Played with, A = Played against

57
118 A
231 A
434
441 W
624 A A
653 W
1255 W
1429 A
1477
1480 W
1484 A
1513 W
1801 A
1817
1865
1898 W
2072 A
2276 W
2582 W A
2583
2585 A W
2587 A
2613 A
2737 A
2833
2881 A
2882 W
2936 A
2956 W
2969 W
2979
3012 A
3016 W A
3080
3103
3320 A
3335 A W
3353 A W
3363
3526 W
3696 A W
3713
3726 A
3727 A
3728 W A
3731
3735 W
3741 A
3758 A
3762 A
3823 A
3833 A
3834 A
3847 A

13 Teams - We never see on the field
8 Teams - We see on the field twice
1 Team - We play against twice

jcatt
03-25-2011, 10:27 PM
As far as I know First has used the same match scheduling algorithm since 2008. The software is simply called MatchMaker and was produced by Idle Loop Software. It was first used I think (correct me if I'm wrong) in 2007, and was later modified in 2008 after the scheduling issues discovered at the 2007 regionals. Since 2008 I don't believe it has changed at all, so what you are seeing are just flukes in the software, when it tries to balance the alliances Red/Blue.

You can find more detailed information on how the algorithm works here (http://www.idleloop.com/matchmaker/) (http://www.idleloop.com/matchmaker/ ). I know from experience that the algorithm generally creates generally fair matches, but don't forget that the software receives no information about the team other that the team number. It does not know the record of any team, or if they have a "good" robot this year. (Or any year)

Edit: Looks like someone beat me to it... What also might be happening, is the large increase in rookie and #3000+ teams, and that is messing with the "balancing" of alliances. Thoughts?

James Tonthat
03-25-2011, 10:43 PM
Actually, my guess would be the 'match separation' time. That usually takes priority over 'pairiing uniformity'. While I can't say I'm an expert on this but I remember scheduling for a couple FLL tournaments and the separation time affects the variety a lot. I also noticed we had really large separation time this year so maybe that allows for less variety in teams that you see on the field.

Thanks for the input all, I might just be getting a bit bored that our regionals are over and we're waiting for champs, something to keep my mind occupied.

PayneTrain
03-25-2011, 10:58 PM
At Chesapeake, we played with 88, then against 88 and 1895, then with 1895.

Why can't someone scan through that list and see enough near-identical match pairings to pull the hypothetical scheduling lever? I know we're talking about 5400 different numbers and all, but regardless of the team, I wouldn't want to play with/against them more than once in qualifiers.

I'd like to face as much of the field in qualifiers as I can.

JackS
03-25-2011, 11:00 PM
James and I were discussing this, and I too did a breakdown of our matches at WPI.

Match Breakdown - 2791- WPI

40 3
121 1
177 1
190 1
195 1
228 1
230 1
236 1
358 1
467 3
571 1
839 1
1100 2
1665 1
1687 1
1735 2
1772 2
1884 2
1991 2
2067 2
2104 1
2168 1
2370 1
2437 1
2648 2
2876 1
2877 2
3044 1
3125 2
3205 2
3273 2
3280 1
3499 2
3780 1

EricH
03-25-2011, 11:08 PM
Why can't someone scan through that list and see enough near-identical match pairings to pull the hypothetical scheduling lever? I know we're talking about 5400 different numbers and all, but regardless of the team, I wouldn't want to play with/against them more than once in qualifiers. They can, and often do. However, they can't get everything, and may not have enough time to run it again if they do spot something like that.

TEE
03-25-2011, 11:13 PM
lol, at the Detroit District Event there were so many odd qualification matches where two powerhouse robots would end up on an alliance against three average/less than average robots

also, we (201) played with 217 once and against 217 twice (they were beastly when they played against us, but did almost nothing when they played with us... if I remember correctly, it was due to a mechanical failure of some sort :eek: )

and we had to replay the same match (Match 60) 3 times in a row due to a field fault

Mr. Lim
03-26-2011, 06:29 AM
Some years you get an easy schedule, some years we get a tough one.

This year at Waterloo was probably the most difficult schedule I've encountered in all my FRC years:

Match 1: Against 2056
Match 2: Against 1114
Match 4: Against 1114
Match 6: Against 1114

610, 1114 and 2056 did get a chance to see a lot of each other on Friday.

Although there are plenty of years where I've benefitted from an easy schedule.

Allison K
03-26-2011, 07:59 AM
I think it should be noted that getting a "bad" schedule is not the same thing as getting a "hard" schedule. A team may have the misfortune of playing against every powerhouse robot at the competition, which would be a hard schedule, but since the match scheduler was still outputting a variety of opponents it's not a bad schedule. I think a bad schedule indicates a situation in which the match scheduler broke - it didn't generate a schedule that fulfilled the necessary requirements. Playing with or against the same team repeatedly would indicate a bad schedule, regardless of how well that other team performed.

I have nothing more than anecdotes but it seems that the scheduler has been doing it's job the past few years. The only anomaly that I noticed was that 67 and 74 played against each other three or four times at West Michigan, but considering the size of the event and all of the other constraints I don't think that's considered a failure.

I'll admit that my team's match schedule at Waterford did make me groan a little bit, but that wasn't the scheduler's fault :)