> For the complete documentation index, see [llms.txt](https://docs.uniocean.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.uniocean.io/forecast-platform/rewards/reward-distribution.md).

# Reward Distribution

UCPRA ranks all forecasts by accuracy and distributes the reward pool to the top 51%.

## Reward Groups

| Total Forecasts | Rewarded (Top 51%) |
| --------------: | -----------------: |
|              15 |              Top 8 |
|             100 |             Top 51 |
|             500 |            Top 255 |
|           1,000 |            Top 510 |

## Distribution Groups

### Rankers

Rankers are the highest-accuracy forecasters. They receive **power-law distributed** rewards from a dedicated portion of the reward pool. This means:

* Rank #1 receives significantly more than Rank #2
* Each lower rank receives progressively less
* The distribution curve rewards exceptional accuracy

### Bucket Winners

Bucket Winners are all remaining winners in the top 51% who are not Rankers. They receive guaranteed multiplier rewards between **1.4x** and **1.6x** their entry fee.

Rewards ramp smoothly from 1.4x (for the last Bucket Winner) to 1.6x (for the best Bucket Winner), based on rank position within the bucket.

## Distribution Flow

```mermaid
flowchart TD
    A[Reward Pool] --> B{Split}
    B --> C[Ranker Pool]
    B --> D[Bucket Pool]
    C --> E[Power-Law Distribution]
    D --> F[Linear Ramp 1.4x–1.6x]
    E --> G[Rank 1 gets most]
    E --> H[Each rank gets less]
    F --> I[Best bucket gets 1.6x]
    F --> J[Last bucket gets 1.4x]
```

## Guarantees

The UCPRA algorithm enforces these invariants:

| Guarantee                  | Description                                                    |
| -------------------------- | -------------------------------------------------------------- |
| Total payout = Reward pool | Every token in the pool is distributed                         |
| Ranker minimum             | Lowest Ranker earns at least as much as the best Bucket Winner |
| Bucket minimum             | Every Bucket Winner receives at least 1.4x entry fee           |
| Bucket maximum             | Best Bucket Winner receives up to 1.6x entry fee               |
| Determinism                | Same inputs always produce the same output                     |
| Verifiability              | Results can be independently recomputed                        |

## Dynamic Top-K

The number of Rankers scales with market size to maintain meaningful reward differentiation:

| Market Size       |               Rankers |
| ----------------- | --------------------: |
| 15–99 forecasts   |                     1 |
| 100–499 forecasts |                     3 |
| 500+ forecasts    | 1% of total forecasts |

This ensures that in small markets, one person earns the top reward, while large markets have multiple top earners creating a competitive gradient.

***

## Related Pages

* [Reward Overview](/forecast-platform/rewards/overview.md)
* [Rankers](/forecast-platform/rewards/rankers.md)
* [Bucket Winners](/forecast-platform/rewards/bucket-winners.md)
* [Reward Examples](/forecast-platform/rewards/reward-examples.md)