Re: "New" 2nd Week Scheduling Algorithm
 

One thing that hasn't changed - you could still accurately predict the 1st round of qualifying matches at Buckeye in Week 4.

Re: "New" 2nd Week Scheduling Algorithm
 

I've posted a white paper with an analysis of both scheduling algorithms, proposed metrics for measuring the desirable qualities of a match schedule, explained a superior solution to the program, and posted a program that generates much better schedules.

Re: "New" 2nd Week Scheduling Algorithm
 

It looks like they tuned the system up for Week 5. They still have the teams divided into 3 parts, but you don't play with/against everyone a whole lot often. In fact, if the regional is big enough, I think it tries to pair you up only once with/against teams. Anyone have a match list to confirm this?

Re: "New" 2nd Week Scheduling Algorithm
 

Well, the analysis from my stat class is not done yet (I decided to wait until data from all of the regional matches is available) but a preliminary look seems to confirm that FIRST did succeed in "levelling the playing field" in terms of rookies and veteran teams. Using this algorithm there is basically no correlation between number of years in FIRST and record after qualifying. It is a completely different question as to whether this is "better" or "worse" than in the past.

The data from the past two years shows that veteran teams do have a significant advantage over non-veterans, but that advantage largely disappears once you get to about 3-4 years in FIRST. This is about what you would expect because by that time you are turning over team members. The really big advantage goes for teams competing in multiple regional competitions. The more competitions, the bigger the advantage. (But the biggest jump comes from the second regional.) I am going to try to compile some data from the past couple of years and find out if the "veteran" correlation gets stronger by comparing the total number of competitions attended to qualifying record. My suspicion is that it will.

A couple of other things jump out at first glance. The algorithm seems to work quite differently for larger regionals (over 50 teams) and smaller regionals. At a small regional there are many more repeat partners and opponents. The teams in the top 8 tend to have played multiple matches with other teams in the top 8. It would have been very interesting to keep track of number of points scored (or at least ringers placed, ramps successfully deployed or climbed) by each team. I know we kept data at Pittsburgh and the correlation between points scored by a team and its record in qualifying is a lot weaker than one would expect.

Tom Saxton's algorithm looks good to me. I am thinking that something like Arrow's Impossibility Theorem will be in play here in that there can't be a system that does the best under all criteria. So there is a need to look at the problems perceived in the current system and try to find a way to change them without causing other problems. I am also wondering if it might not be advisable for FIRST to incorporate how many total competitions a team has attended rather than just number of years in FIRST and perhaps how many competitions a team has/will attend that year.

The real question becomes two part: Should FIRST change the algorithm and will FIRST change the algorithm? Neither answer is really obvious to me.

Re: "New" 2nd Week Scheduling Algorithm
 

We were against 1114 TWICE IN A ROW and with them once during qualifying at GTR, a very large regional.

We had several other repeating partners/opponents.

Draw your own conclusions.

Re: "New" 2nd Week Scheduling Algorithm
 

Teams went ten rounds at Palmetto, which had 45 teams. Of 1618's rounds:

Played With:
Three times each: 665 and 1251
Two times each: 34, 343, 1102, 1293
Once each: 337, 538, 587, 900, 1249, 1390

Played Against:
Three times each: 1051, 1522, 2237
Two times each: 337, 538, 1436
Once each: 135, 281, 342, 386, 804, 832, 845, 900, 1026, 1225, 1251, 1287, 1390, 1553, 2187

Take from that what you will.

Re: "New" 2nd Week Scheduling Algorithm
 

We were with 1255 3 times at LSR, and they didn't have a prayer of dribing up our ramps. In fact, we only had about 3 matches where teams could actually drive on our ramps. If we had known we'd be consistently paired with rookies and never with veterans, it would've seriously affected our strategy and robot design at the beginning of th season. The very least FIRST can do next year is let us know what the match algorithm will be so we can plan accordingly.

Re: "New" 2nd Week Scheduling Algorithm
 

Re: "New" 2nd Week Scheduling Algorithm
 

They also improved the rankings display. It goes from page to page instead of poorly scrolling. It also has both RS and MP now so you can actually see how they break ties

Re: "New" 2nd Week Scheduling Algorithm
 

111-2 with 1 against
71-1 with 3 against
703-2 against (Both with Hammond)
1918-2 with
288-2 with
27-1 with 1 against
2188- 1 with 1 against
1783-1 with 1 against
2054-2 against

The system really bugs me. Take 1718 for example, Yes they have on of the best machines this year, but with the scheduling system they also have easier matches. We love 71 alot, but playing them three times is ludicrous. Admit it FIRST, the system is partially broken, go back to last years system or risk having several teams really upset with you