Calibr

Calibr introduces a new DeFi primitive: Proof of Calibration — a way to turn long-term decision quality into an owned, composable on-chain asset.

Today, most on-chain systems only measure capital. A new wallet, a skilled decision-maker, and a malicious actor all look identical. This breaks governance, incentives, and any system that depends on judgment rather than money.

Calibr fixes this by making accuracy and confidence measurable over time. Users participate in prediction markets with objectively verifiable outcomes and explicitly state how confident they are in each decision. When markets resolve, outcomes are finalized via an automated oracle pipeline, and each user’s calibration history is updated using proper scoring rules that reward well-calibrated conviction and penalize overconfidence. Because this signal compounds gradually, it cannot be instantly bought, farmed, or reset with a new address.

This calibration history is not a badge or an off-chain profile. On Sui, it is implemented as an owned on-chain object that the user controls. Prediction markets exist as shared objects, allowing parallel execution across markets, while user reputation updates happen independently and safely at scale. This object-centric model makes reputation a first-class asset rather than an application-specific score.

Other protocols can directly compose on top of Calibr — to gate access, cap or unlock confidence and leverage, weight governance input, or price risk based on demonstrated skill instead of raw capital.

Calibr shifts DeFi from capital-dominant systems to skill-aware systems, creating a foundation where long-term decision quality becomes a durable, portable economic signal across markets, governance, and future credit primitives.

Calibr is built to measure decision quality over time, which requires two things to work correctly: long-lived user state that cannot be reset, and many independent markets that can update in parallel without bottlenecks.

Each prediction market is implemented as its own on-chain object, allowing markets to exist and resolve independently. Users predicting on different events never contend on a single global contract. Prediction positions are attached using dynamic fields, keeping access constant-time and preventing state growth as markets scale.

User reputation — “Proof of Calibration” — is implemented as an owned on-chain object bound to a wallet. It stores a rolling calibration score updated after every resolved market using proper scoring rules. Because this object is owned, non-transferable, and persistent, reputation cannot be sold, reset, or recreated by switching accounts. Skill must be earned over time.

When a market resolves, the protocol executes a single atomic transaction that finalizes the outcome, settles confidence-weighted gains and losses, and updates each participant’s reputation object. This guarantees that payouts and reputation updates always occur together, with no intermediate or exploitable states.

Economically, Calibr uses a risk-weighted parimutuel system rather than an order book or AMM. Every prediction has a fixed base stake, but users explicitly choose confidence (50–90%). Higher confidence increases the portion of stake at risk. When a market resolves, all losing risk forms a pool that is distributed to winners proportional to their risk, creating a zero-sum system where being more correct matters, not just being on the right side.

Market resolution is handled by an automated oracle service that listens to on-chain events, gathers objective off-chain data, and submits a final YES/NO outcome on-chain at resolution time. This removes manual intervention while keeping outcomes verifiable.

Because reputation lives as an owned on-chain object rather than an off-chain score or token balance, other applications can directly read and use it — to gate access, adjust limits, weight governance, or price risk based on demonstrated skill instead of capital alone.