Leaderboard Rebalancing - March 2024

Change to how we count resources

Previously we counted models.
Now we count versions. This addresses a balance issue with models that had many versions

Change to how resources are valued over time

Previously a resource was counted at it's full value until it hit the 30 days old and then it fell off completely.
Now a resource loses score gradually over time following this curve:

• 7 days: 94%

• 14 days: 78%

• 21 days: 50%

• 28 days: 12%

Change to how quantity vs quality is balanced

To help balance users that produce high volumes of models against those that prefer the one-hit wonder approach, we rank entries by their score and then apply a quantity multiplier.
Previously we had a stepped discount to the value a resource could have:

• Top 1-5: 100%

• Top 5-10: 75%

• Top 10-15: 50%

• Top 15-30: 25%

• Top 30-60: 10%

Now we follow this curve and have extended the range to 120. Happy to take suggestions on this formula (check it out by clicking on the link to the curve).