Starting ratings are based on the closest-known rank according to osu!track, or the most recent global rank if none is known. The initial placement is based on the closest point in time relative to when the player started playing tournament matches.
These initial ratings roughly follow a bell curve and are piecewise linearly dependent on
This rank z-score will be a real number typically between
Info
Instead of constantly recomputing the values of
and for every rating recalculation, a snapshot of those values was stored during the beta phase of o!TR and is used for all calculations. The constants are found in the processor code and are also listed in the table below.
| Game mode | ||
|---|---|---|
| osu! | ||
| osu!taiko | ||
| osu!catch | ||
| osu!mania 4K | ||
| osu!mania 7K |
A player’s initial rating is then determined via the formula
with an enforced minimum and maximum rating of
Info
For reference, players ranked near
in osu! will have a rank z-score around and receive an initial rating of approximately .
While this formula does not account for historical rank drift 1, this approximation is still good enough for our purposes especially because rating decay applies to inactive players.
Any players whose rank data cannot be recovered from either osu!track or the osu! API are given a default initial rating of
Note
Regardless of initial assignment, the accuracy of one’s rating will improve after playing a few tournaments.
Initial Ratings for Filtering and Seeding
Tournaments may use our built-in filtering tool to filter or seed players. However, players who have never played in a verified match will not have a rating in the system. In some cases, these players may be filtered by the tool, such as when enforcing a minimum rating. Tournament organizers should follow these steps to ensure these players are accounted for.
- Using the formula above, an estimated initial rating can be calculated from a player’s current rank. If players’ ranks and o!TR ratings are both being displayed on a tournament’s main sheet, we recommend calculating estimated initial ratings using those displayed ranks for consistency.
- If ratings are used solely for filtering, then it is best for rank-eligible players to have estimated initial ratings within the filtering range. For example, players ranked between
and would have an initial rating roughly between and . Therefore, a 5-digit tournament should typically have an upper rating limit higher than . If there is no minimum rating required to play, the filtering tool will not exclude players without a rating. - If ratings are also used for seeding, keep in mind that on average, new players lose rating in their first matches. To avoid overestimating the performance of new players, consider using an adjusted initial rating. For example, if a tournament is seeding the top
players by rating into a bracket, the formula could be used for players without a rating.2
Important
As a reminder, any use of o!TR for an officially supported tournament must have approval from the Tournament Committee before play begins. Tournaments which deviate from the filtering tool’s results, including to follow these recommendations, must clearly document exactly what is different. To ensure fairness, organizers should determine how they will deviate from the filtering tool’s outputs before advertising their tournament.
Footnotes
-
This term refers to how a player at a certain rank today typically displays a higher skill level than a player at that same rank years ago. ↩
-
The choice of
comes from empirical rating data. Specifically, osu! players currently ranked or worse lose roughly TR on average in their first matches and TR in their first matches, so subtracting off from their estimated initial rating is a modest correction to avoid overestimating first-tournament performance. ↩