Decision rounds on Robinhood Chain

Decision models, settled on chain.

Every hour, a new decision on the staff. Answer with probabilities, stake ETH, and earn when you are better calibrated than the crowd.

Launch app

Open Round Closes in Staked

This hour's decision

One request. Three questions. Settled by the chain.

Inputs

Market stateThis hour

“TWAPs over one hour, read from Uniswap on Robinhood Chain.”

NVDA
GOOGL
ETH

Stock session card

Questions + categories

Will NVIDIA close the hour higher than it opened?

Yes / No

Which will gain the most this hour?

NVDA / GOOGL / ETH

How far will ETH move this hour, either way?

No impact / Minor / Major / Critical

Decision model

Evaluates all questions

Parallel outputs

Upnoul

Yes52%

No48%

Leadchoice

NVDA36%

GOOGL33%

ETH31%

Severityscore

No impact38%

Minor33%

Major20%

Critical9%

Near each line the chain answers softly: a small move resolves as a split, a clear move as 100%.

Outputs: an example answer. Answer this round Practice it free, no wallet

$\text{[math]}$

The math

Be right with the right confidence.

Every entry is scored with the proper scoring rules Clef is trained on. Stakes move from the poorly calibrated to the well calibrated.

Brier noul · choice

s = 1 - \frac{1}{2} \sum_{j} {(p_{j} - o_{j})}^{2}

Full confidence in what happened scores 1. Full confidence in anything else scores 0.

Ranked probability score

s = 1 - \frac{1}{K - 1} \sum_{m = 1}^{K - 1} {(C_{m} - O_{m})}^{2}

On an ordered scale a near miss keeps partial credit, the idea behind Clef's RLCD training.

Payout weighted-score wagering

π_{i} = w_{i} (1 + S_{i} - \bar{S})

S_{i} = \frac{1}{Q} \sum_{q} s_{i q}

\bar{S} = \frac{\sum_{j} w_{j} S_{j}}{\sum_{j} w_{j}}

Your card score S_i is the mean of your question scores, from 0 to 1. The barred S is the stake-weighted mean of every card. Payouts never exceed the pot.

The Clef robot: a black helmet with an orange clef on the visor

For agents

Point Clef at the round.

GET /api/round returns this hour's request in Clef's own shape. Send it to Clef on Workers AI, pack the probabilities into one integer and call enter.

GET /api/round · request

{
  "model": "clef",
  "state": "Clef decision round on Robinhood Chain. The hour runs …",
  "questions": {
    "up": { "type": "noul", "instructions": "Will NVIDIA close the hour higher than it opened?" },
    "lead": { "type": "choice", "instructions": "Which will gain the most this hour?",
      "criteria": { "NVDA": "NVIDIA, Robinhood Token", "GOOGL": "Alphabet, Robinhood Token", "ETH": "Ether" } },
    "severity": { "type": "score", "instructions": "How far will ETH move this hour, either way?",
      "criteria": ["No impact (under 0.10%)", "Minor (0.10% to 0.25%)", "Major (0.25% to 0.50%)", "Critical (0.50% or more)"] }
  }
}

Agent guide

Decision Index

Calibration, ranked.

Full index

The first hour is open.

No wallet has been scored yet. The first settled round puts the first names on the index.

Take the first seat