School Rankings

[yoshiketchup]

Hi all,
As you may have heard, Klebb, AC (builderguy135) and I have been working on a spreadsheet that ranks all teams in the nation based on competition performance. We've decided to release the spreadsheet here!

How are competitions rated?
Competitions are rated based on a variety of factors such as date, team count, school count, event count, and other factors to determine a score (approximately varying between 0 and 100).

A team's score is then calculated for each competition by comparing their superscore to the top score. Then, a school's total score is calculated based on a combination of the competitions that they have attended.

Why are there only 100 schools?
It's kinda toxic ¯\_(ツ)_/¯

Why is copy/paste export disabled?
See below. We also understand that there's ways to bypass it: we can't stop you, but we hope "please don't" is enough.

Why is my school ranked lower than that other school?
No.

Is there a division B one?
No.


Disclaimer:
These rankings, while mathematically objective in nature, are by no means official. Some of the criterion and weights are subjective based on how we designed the algorithm. Do not take these rankings as fact, and don't get too caught up in them! This was more an interesting way to visualize data for us.

We also request that you do not use this data in any other manner, such as predictions.

[pepperonipi]

Will you share what the algorithm is?

I think this is super cool, but some of the numbers make no sense by simple observation. I'm also curious what factors you think are important in the design of an algorithm to rank schools throughout the nation.

[builderguy135]

Will you share what the algorithm is?

I think this is super cool, but some of the numbers make no sense by simple observation. I'm also curious what factors you think are important in the design of an algorithm to rank schools throughout the nation.

First, each competition is awarded a weight. The exact formula is extremely complicated, but it takes the following factors into consideration:

• Date - Invitationals that are later in the season are worth more.
• # of Teams - More teams = Harder competition.
• # of Schools - More schools = Harder competition.
• Non-HS - Non-HS invitationals generally attract more competitive teams.
• T25 "Bonus" - A bonus for the presence of teams that were in the top 25 in 2019 Nationals.
• In Person - In person competitions are given a small bonus.
• Events - A small bonus is given to teams with more events.
• In State - A penalty is applied if a competition restricts registrations to within a state, since this decreases competitiveness.

Then, the competition is superscored and each team is assigned a score. The weight of each competition is then given in its fullest to the team that won the superscore, and points are awarded to teams by dividing the number of points that the superscore winner obtained by the number of points the team scored, then multiplying by the weight of the competition.

Finally, a ranking score is given to each team by taking the maximum value of the following expressions:

• 80% of the highest competition score of the team
• 70% of the highest competition score of the team + 20% of the second highest
• 60% of the highest competition score of the team + 30% of the second highest + 10% of the third highest

There are still many issues with the formula, and we're working on improving it - for instance, a team that scores 30 points should not be awarded the same number of points as a team which wins a competition by scoring 300 points in two competitions of equal value.

Just out of curiosity, which numbers do you think need to be adjusted? We're still trying to improve the formula, and we're open to any suggestions.

[MorningCoffee]

I think in-person should get a penalty, since COVID restrictions reduce competitiveness when many strong teams can't attend.
UGA's score seems a bit high.

[knightmoves]

You're assuming a somewhat equivalent field at each competition when you do this. I suspect you might do better to view, for example, School A scoring 5% of the no-show score for a tournament, and School B scoring 15% of the no-show score as a vote that School A is 10 percentage points better than School B. So if A beats B narrowly 3 times, the algorithm will rank A above B, but not by far. If C blew B out of the water a couple of times, but has never seen A in competition, you'd rank C well above B, and so by inference above A.

You can probably adapt a score voting algorithm to do this.

[sciolyperson1]

You're assuming a somewhat equivalent field at each competition when you do this. I suspect you might do better to view, for example, School A scoring 5% of the no-show score for a tournament, and School B scoring 15% of the no-show score as a vote that School A is 10 percentage points better than School B. So if A beats B narrowly 3 times, the algorithm will rank A above B, but not by far. If C blew B out of the water a couple of times, but has never seen A in competition, you'd rank C well above B, and so by inference above A.

You can probably adapt a score voting algorithm to do this.

That's a great idea, I'll think about it further.

As for equal fields at competition, we score each competition, then assign points to each school - so weights are balanced to each individual competition, and no competition is weighted equally.

[builderguy135]

Agreed with everything sciolyperson1 said above, but I also wanted to add that it's really difficult to implement a true rating system - ELO and GLICKO wouldn't work, and multiplayer rating systems such as TrueSkill wouldn't work either, as that would not only not account for scores (only placements), but it also wouldn't account for teams that send junior varsity or non-main teams to competitions. The best way is to count only the competitions that a team places the best at, and giving each rating points depending on the competitiveness of the competition as well as the number of points scored.

[knightmoves]

but it also wouldn't account for teams that send junior varsity or non-main teams to competitions.

That's a good point, and without inside information, you don't know whether a particular school sent their better people to a particular competition. There are also schools that like to hide their performance on build events, saving their real build for the real competitions - but that's less important this year!

[BobbyJoe]

This is awesome! Will it be updated again soon?

[sciolyperson1]

This is awesome! Will it be updated again soon?

Been busy with invitational season these few weeks but hopefully it will! Looking like it'll be Sunday when I put in GGSO and the rest.