Worth of Dynamic Match Schedule

One of the ideas to come out of Competition Alliances (before it was locked) was the idea of dynamic match schedule.

@Robo_Eng_13 said

I like this idea a lot, but it has some drawbacks.

Pros:

  • More accurately rates teams
  • Match schedule no longer dependent on luck
  • Less likely to for teams to (a) be carried, or (b) carry another team in qualifications

Cons:

  • Unknown match schedule
  • Teams will not be guaranteed time between matches, making it harder to do skills, interviews, scouting, etc.
  • The above means that this system will give an advantage to larger teams (in team members)

For the implementation itself, you could use a variation on an Elo, TrueSkill, or Glicko. I’d actually be interested in writing such an algorithm (it might be a summer project)

What are your thoughts? Do you like the idea? Do you think it’s terrible? Why?

Discuss!

A dynamic match schedule would be awesome! Would it be hard to create a schedule where you saw different teams every round if the system is designed to match you up with equal levels of competition?

I suppose the system could be designed to make alliances equal (rather than all 4 teams), so the following might be good matches, for example:

Rank2 and rank 20 vs rank 10 and rank 12
Rank 7 and rank 9 vs rank 6 and rank 8

I think this would be great for teams of all levels.

The problem with that solution is that you are assuming that the skill difference between 1 and 2 is the same as 12 and 13, which it likely isn’t. A better system would attempt to construct a scaling function based on ranking differences, and use that to pair teams.

This seems promising, but it would ruin the concept of VEX for learning teams. They would have not as good robots which would result in the best teams being ranked as the best, but the bad robots would not be given a chance to compete at a higher level, which means they wouldn’t be given elimination experience ever. I do agree with dynamic rankings, but it wouldn’t be implemented because it wouldn’t be good for newer teams.

Using the dynamic schedule on a limited basis (for instance, using the dynamics to shift teams between four tiers) would allow the benefits to teams of all levels of always competing against teams of similar ability, and then do elimination within each tier, allowing weaker teams to still experience eliminations.

I am not sure how scaling would work to do alliance selections, but i think a tiered system would allow some benefits to strong and weak teams alike.

At this point, there have been many suggestions for how to overhaul the ranking process. I particularly like this one. I think that Vex should designate a few tournaments world wide to be “pilot” tournaments in which new strategies like this are implemented. It’s tough to just pick one new method and implement it for a whole year world wide, do you think anything like this would happen in the near future. I’m sure this issue has been brought up in the higher ranks of Vex.

Part of the benefit of having the third alliance is the ability to advance a weaker team. It allows them the exposure that helps VEX grow. My first year, i was part of the terrible team that got carried to Worlds as a second pick. We came in the mid 90’s of our 100 ish team division. However, the experience drove us to improve and the next year we qualified on our own merit and competed at an average level (40s or 50s). Had we not gotten that exposure to the next level, we probably would not have bothered having a second year.

But if a bad robot wins one tournament, it will encourage them to continue, and most likely in the future they will build better robots. If not, lighting doesn’t always strike twice.

With multiple fields I’m sure it would be possible to give teams an idea of when they will have their match, if that’s not enough / or if they only have one field maybe a time range could be given? If match times will be decided on the spot, I find it highly unlikely for this to ever take off. However, I can imagine solutions to make this slightly less variable.

The dynamic scheduling would only take place incrementally after a “set” of rounds. For example, with 32 teams, it takes 8 matches for every team to play one match. I would consider that a “set”. After 8 matches, then the next 8 would be scheduled. Allowing a little soace between sets would be reasonable without drastically slowing down the tournament.

I think that a good robot losing because of bad pairings is worse that a bad robot winning because of good pairings. The good teams that have prepared should not have their day shot by a random decision made by a computer before the whole thing begins.

Would this really be better than just scheduling an extra two matches?

I think this is the real question here. Is all of this actually better than just scheduling a couple extra matches.

We were so bad at out first Starstruck tournament in February and got carried to tournament champions. A week later we went to another tournament and after making a few adjustments we went unbeaten and won the tournament … Getting carried gave us a lot of experience and encourage

You should read this article https://en.m.wikipedia.org/wiki/Swiss-system_tournament about how other sports do it with this style. Key issue that I see is this works for single players, but not alliances.

The other issue has to do with there are variables around the robot, a broken motor may dump you into misdemeanor pack where in a 100 team division you wouldn’t be able to recover.

I think that the way the match schedule is formed should stay the same just there should be an amendment to the ranking system. Such as some WP may be worth more than others, or the SP system should be turned into TRSP’s. If you don’t know what TRSP’s are, you can find their definition on vexdb.

Whatever the tiebreakers are, they need to reward teams for winning matches and scoring lots of points. We reward the former but not the latter. TSRP’s reward wins, not points. Is that good enough for you? How should points work into it? I hate it when teams just stop playing because they have their points because of the SP tiebreaker.

Matches are also scheduled to give you maximum time between matches, and I would like at least 4 matches before you start getting dynamic and locking teams into the “bad” bracket.

Also, if your first couple of matches go sour for any number of reasons you might as well pack up and go home. And some teams will. If you have a bad first match for a number of reasons then you will be shot for the rest of the day as you will be stuck in the lower bracket with no hope. You will be paired with poor alliances and your opposition will more likely be a pair of great robots. By the final round you’re going to be seeing really good robots scoring just enough to beat two poor teams because of the SP tie breaker. I’m bored already.

Simply adding another field set would double your number of matches at most events. This means we all need to volunteer, bring fields, help set up and tear down events, etc. If we all pull together, every tournament can have 10+ qualification matches. and a whole lot more action in general. If you have ever been bored at a tournament, then they don’t have the help they need to keep things exciting. Get your aunts, uncles, cousins, friends, co-workers, professors, whoever out there and make it happen!

No amount of volunteers can magically come up with the space necessary for more fields. Some events are quite confined in terms of space, and additional fields would be virtually impossible.

All this talk about having 10 or 16 matches and banding together is great and we should definitely try for it, but we’ve also been trying for it for quite a few years now so I don’t expect much improvement. I think the best idea is to have the 6 matches that most of the az competitions have be broken in half with the first 3 running random and regular, then the last three running where the algorithm tries to make every robot on the field as closely ranked as possible. So if you bomb the first two matches, you’re going to have easier opponents and you can still fight your way to the middle of the tournament. If you’re last place with a field of last place robots, you shouldn’t be ranked anything but last place if you still can’t win. The ranking system could be identical or even a bit fine tuned. Worst case scenario is the first three matches you get strong enemies and bad allies and then in the second three matches you win and you’re at least ranked averagely. You can also go up to a alliance captain and explain you were undefeated when the matches were fair. There is also the sub debate here that needs to run its course: should a new ranking system have ‘carried’ teams? I believe the best team should be ranked like they are the best team and we should implement other strategies for strong teams and weak teams to interact, for example in alliance selection we could make the first pics play like normal and the second picks start from the eighth alliance and end at the first alliance.

They could form elaborate human scaffolding allowing additional fields to be placed on a second level formed out of the volunteers’ bodies (assuming the event is taking place in a gym or other area with a tall ceiling).

On a more serious note, it seems apparent to me that the current ranking system was not even designed with the intention of finding the best teams (which seems to be a tacit assumption in this thread). If it were, APs wouldn’t be the first tie breaker. Rather RECF chooses to reward the things they want to emphasize (ie winning matches and then winning autonomous).
So the trouble isn’t merely a matter of coming up with a better system for finding the best teams, but rather convincing RECF that the point of qualifications should be to find the teams most likely to win the tournament and to rank them at the top.