Accuracy.FunDocs

Introduction

Score Mechanism

Score Mechanism

The Score Mechanism is the core algorithm that calculates a content creator's performance score based on changes in their metrics over a specific betting period. This score is used to determine winners and distribute rewards from the betting pool.

Optimal Equation for Score Calculation

To calculate a score for a content creator based on changes in views, likes, and subscriber count over a specific period, we use a weighted equation that normalizes metrics across creators and different platforms.

Variables

Let:

  • ΔV = Percentage change in views over the betting period
  • ΔL = Percentage change in likes over the betting period
  • ΔS = Percentage change in subscribers over the betting period

Step 1: Normalize Each Metric Individually

Since views, likes, and subscribers can grow by varying degrees, we first cap each metric's percentage change so that none of them can contribute an unlimited amount to the score.

We'll assume that any change beyond 100% should be capped at 100%, and negative changes will contribute negatively. The simplest approach would be to cap values between -100 and +100 for each metric.

Capping Function:

cap(x, -100, 100) = {
  -100 if x < -100
  x    if -100 ≤ x ≤ 100
  100  if x > 100
}

This ensures that:

  • Each metric's contribution is bounded
  • Extreme outliers don't skew the final score
  • Negative changes (like losing subscribers) are properly accounted for

Step 2: Weighted Score Formula

We assign weights α, β, γ to views, likes, and subscribers, respectively, and normalize the sum of the weights to 1. These weights reflect how important each metric is relative to the final score.

Normalized Score Formula:

Score = α × cap(ΔV, -100, 100) + β × cap(ΔL, -100, 100) + γ × cap(ΔS, -100, 100)

Or in mathematical notation:

Score = α·cap(ΔV, -100, 100) + β·cap(ΔL, -100, 100) + γ·cap(ΔS, -100, 100)

Where:

  • α + β + γ = 1
  • The function cap(x, -100, 100) ensures that x is restricted to the range [-100, 100]

Weight Configuration

The α, β, γ factors can be given a balanced approach where each will be ⅓, but as per analysis of the platforms we can define these factors by priority:

Recommended Weights:

  • α = 0.5 (for views) - Views are given the highest weight as they represent overall reach
  • β = 0.3 (for likes) - Likes indicate engagement quality
  • γ = 0.2 (for subscribers/followers) - Subscriber changes reflect long-term growth

Example Calculation:

If a creator has:

  • Views increased by 150% → capped to 100%
  • Likes increased by 80%
  • Subscribers increased by 20%

Then:

Score = 0.5 × 100 + 0.3 × 80 + 0.2 × 20
     = 50 + 24 + 4
     = 78

Ensuring Score Ranges from 0 to 100

To ensure that scores always range from 0 to 100, we:

  1. Cap the values for each metric between -100 and 100 (as done above)
  2. Ensure the sum of the weights equals 1 (α + β + γ = 1)

With these constraints:

  • Minimum possible score: -100 (if all metrics decrease by 100% or more)
  • Maximum possible score: 100 (if all metrics increase by 100% or more)
  • To normalize to 0-100 range, we can use: normalized_score = (score + 100) / 2

Alternative Normalization:

If you want scores strictly between 0 and 100:

Normalized Score = 50 + (Score / 2)

This maps:

  • Score of -100 → Normalized Score of 0
  • Score of 0 → Normalized Score of 50
  • Score of 100 → Normalized Score of 100

Reward Distribution

Once the score is calculated for a content creator, it is used to determine the reward distribution from the pool to respective users according to their relative prediction closeness.

The closer a user's predicted score is to the actual calculated score, the higher their share of the reward pool. The distribution mechanism ensures that 100% of the pool is allocated based on accuracy scores, with perfect predictions receiving the largest share.

Key Considerations

  1. Time Period: The score is calculated based on changes over the betting period (1 day, 7 days, 1 month, etc.)

  2. Negative Changes: Subscriber count can negatively affect the score if it decreases, which is important for maintaining accuracy

  3. Platform Differences: The same formula works across different platforms (YouTube, Farcaster, Twitter, Instagram) by using percentage changes rather than absolute numbers

  4. Fairness: The capping mechanism ensures that creators with massive follower bases don't have an unfair advantage over smaller creators

Previous
Getting started
Next
Reward Pool Distribution