For the complete documentation index, see llms.txt. This page is also available as Markdown.

Global Leaderboard

Global Leaderboard on Crunch.

The Global Leaderboard allows Crunchers to compare their performance across all competitions. Each ranking matters, but depreciates over time to ensure that new Crunchers have the opportunity to rank up as well.

There are two rankings:

They are updated every Monday night.

How does it work?

The actual implementation is available on GitHub.

Events

When you appear on a Competition's leaderboard, it creates an event that will become active once the Crunch is over.

An event may be ineffective if:

  • Your rank is lower than 500.

  • You are not eligible for a reward on the leaderboard, e.g. the system detected a duplicate submission.

Raw Points

Points are based on the size of the competition prize pool and your rank on the leaderboard. For continuous competitions, the annualized prize number is used.

Then, multiple weights are applied:

  1. The weight of the best rank of the user's group is based on a power law distribution.

    • If you participate alone, your highest model rank is used.

    • If you participate as a team, the highest rank of each team member is used. Members who did not submitted are also awarded points.

    • For real-time competitions such as Falcon or Synth, your rank on the payouts is used.

  2. The weight of each leaderboard's target as each target may have a different prize pool. For example, DataCrunch (v1) has the following distribution:

    Target
    Prize Pool
    Weight

    white

    10,000

    ~0.09

    red

    20,000

    ~0.18

    green

    20,000

    ~0.18

    blue

    60,000

    ~0.55

  3. The Phase's per-crunch weight:

    These weights are also divided by the competition's duration in crunches. For continuous competitions, the weight is divided by the equivalent in a single year.

The following formula is executed for each event, i.e., one per user and leaderboard:

raw_points=prize_pool×rank_weight×target_weight×phase_weight\text{raw\_points} = \text{prize\_pool} \times \text{rank\_weight} \times \text{target\_weight} \times \text{phase\_weight}

Legacy Competitions

Submissions and rankings from the DataCrunch (Legacy) competition are also taken into account. Since the system was quite different, the weights had to be chosen carefully in order to continue rewarding those who have been here since the beginning.

Decayed Points

Over time, the event will have a smaller and smaller impact. The following decay formula is used for this purpose:

decay_factor=e−today−event_startdecay_constant\text{decay\_factor} = e^{-\frac{\text{today} - \text{event\_start}}{\text{decay\_constant}}}

Then, the raw points are multiplied by this factor to determine the number of points for a given date.

The decay constant currently used is around 500 days, but it may change in the future.

Participations & Institutions

Participation is considered when a user registers for a competition.

It's easier for an institution with more members to appear at the top, but it's not guaranteed (quality versus quantity).

In order to join an institution, you must select the correct university in your profile. You will then be automatically added to the next Global Leaderboard release.

If you are the first, an institution will be created with your university's name. The "About" page will feature a reworded version of your university's website description, crafted by an LLM for a more polished look. If you would like to contribute information to your institution's page, such as links, the "About" section, an avatar, or a banner, please contact us on Discord.

Positions

After completing all the previous steps, the leaderboard positions are finally computed and ranked.

A history is kept so you can see your progress over the years.

Last updated

Was this helpful?