FAQ

How does the ranking algorithm work?
Imagine that every qualifying player is given a coordinate on a vertical line. The goal is to figure out where to place each player on this line to create the most accurate possible ranking. To do this, we use the set history of wins and losses between each pair of qualifying players who have played each other in tournament. If player A beats player B in a set, then that corresponds to a linear force acting between the two which attempts to push player A to a higher position on the line than player B. In other words, winning pushes a player higher, and losing pulls a player lower. Using the sum total of all of these forces from every set played between qualifying players in qualifying tournaments, it is possible to calculate the positions for every single player on the line where all of these forces achieve perfect balance with one another. These positions constitute our ranking.

Which players qualify for the ranking?
In order to guarantee a modicum of interactivity with other skilled players, only players who have attended at least two regional tournaments (200-399 players) and/or one major tournament (400+ players) and/or one invitational can qualify for the ranking.

Those attendance requirements seem very lenient. Is there a penalty for low tournament attendance?
There is a very light attendance penalty factored into the ranking algorithm. It corresponds to a small "gravitational force" that pulls you lower on the rankings if you don't have very many wins pushing you higher. However, there is a problem with penalizing attendance too harshly: it unfairly drags down anyone who loses to a "hidden boss." To illustrate this, let's take this example: say a hidden boss attends only one tournament throughout the year, and defeats many top players. If this player is penalized for low attendance, then everyone they defeated is also penalized, because they have lost to a "worse" player, even if the player isn't really worse! Another reason the attendance penalty is light is that I specifically wanted 1000rank to highlight these "hidden bosses" with untapped potential, because I want it to be as accurate a ranking of players' skill as possible, even for players with relatively little tournament data. Therefore, instead of instituting a harsh penalty for low tournament attendance, I opted to include an "Uncertainty" value for every player in the ranking, which is roughly inversely proportional to how much interaction they have had with other players somewhat near their skill in tournaments. See below for more information.

Where did you get your tournament data?
https://smashdata.gg/ is an absolutely incredible resource. Not only does it have results for every Ultimate tournament on smash.gg with at least 32 entrants, it also has most Japanese tournaments, even if the brackets were done on Challonge. And they provide all their data for free in the form of an sqlite database you can download from github! Big thanks to them, I could never have done this without them. By the way, clicking on the ranking number of any player on a 1000rank ranking table will take you to that player's profile on Smashdata, where you'll be able to see a summary of their tournament results.

Which tournaments are included in the ranking calculation?
Somewhat controversially, I have decided to use the results from every offline tournament in the database with more than 50 entrants, plus invitationals. This includes locals and weeklies (as long as they meet the entrant requirement). There are two major reasons for this: (a) overall, more data will result in higher accuracy in the ratings, and I believe this is worth the trade-offs that come with including locals, and (b) as mentioned previously, I want to be able to highlight the skill of "hidden bosses," who may have absolutely terrific results at locals, but haven't attended too many major tournaments. In order to rank these players with any degree of accuracy, I need to include their local results in the ranking calculation.

There is a column in your ranking table called "Uncertainty." What is that?
Uncertainty measures the lack of tournament interaction that a player has had with other players who are somewhat close to them in skill. In other words, if a player has a low uncertainty, their position in the rankings is relatively assured, but if they have a high uncertainty, then they are a "hidden boss" whose true level of skill could actually be significantly lower or higher than their current ranking. Including uncertainty in my rankings allows me to circumvent the problems that come with penalizing low attendance (see above), because this way you can just refer to uncertainty to see which rankings should be taken as most accurate, and which are less assured due to lack of tournament data.

The actual value of uncertainty is calculated by taking the derivative of the "forces" between players from set wins and losses that I described earlier. This value is roughly equivalent to how much your score will decrease if you go 0-2 at your next tournament, or how much your score will increase if you take a set off of the two best players in the world. Use this as a general rule of thumb: (a) if the uncertainty is less than 10, then the ranking is pretty accurate, (b) if the uncertainty is between 10 and 20, then the player's true level of skill is more nebulous, (c) if the uncertainty is greater than 20, then the player is a true "hidden boss" who needs to attend more events if they want to prove their position.

And what about the "Volatility" column?
Some players tend to cause upsets on players above them in the ranking. Or get upset by players below them. In fact, pretty much any player who's prone to one of these two is also prone to the other. These players are volatile ⁠— they're more likely than other players to go on crazy runs in bracket, or buster out and lose to players worse than them. I measure volatility by calculating the number of upsets both made and taken by a player relative to the number of non-upsets within the same radius of skill as the upsets. Because the resulting value only has meaning relative to other players' volatility scores, I then sort the volatility scores into ten qualitative categories. First, if a player has never made or taken an upset in a qualifying tournament, their volatility is Zero. Then, in increasing order, their volatility may be Minimal, Very Low, Low, Medium-Low, Medium, Medium-High, High, Very High, or, in certain rare cases, Extreme. Please note that most players who made it into the top 1,000 got there by not taking many bad losses (in other words, by demonstrating consistency, the opposite of volatility). Therefore, most of the players in the rankings have a volatility level on the lower end of the scale.

I have more questions. Where can I contact you?
DM me on Twitter.