A benchmark-driven probability subnet that identifies and rewards true forecasting skill, producing a decentralized superforecaster oracle for prediction markets.
Probity is a benchmark-driven forecasting subnet on Bittensor that transforms probabilistic accuracy into a competitive digital commodity.
It creates a decentralized marketplace for calibrated intelligence, where miners are rewarded not for capital allocation, popularity, or narrative momentum, but for consistently producing better probability forecasts than the market itself.
The Problem
Prediction markets are powerful sources of information, but they are capital-weighted, not skill-weighted.
Market prices reflect positioning, liquidity, and crowd sentiment. They do not directly identify who is persistently well-calibrated, who actually improves on consensus, or how forecasting ability can be converted into a reusable intelligence layer.
There is still no decentralized system that can continuously:
identify high-skill forecasters
reward probabilistic accuracy over time
turn forecasting performance into programmable probability infrastructure
The Solution
Probity introduces a decentralized forecasting marketplace where miners compete to submit probability estimates for objectively verifiable binary events.
Validators score every forecast deterministically using:
strictly proper log-loss
relative performance versus the market baseline
rolling skill aggregation over time
exponential weighting for miner ranking and influence
The result is a Skill-Weighted Probability Ensemble (SWPE): a probability output shaped by persistent forecasting skill rather than raw capital.
This gives Bittensor a new kind of digital commodity: benchmarked probabilistic intelligence.
Why It Matters
Probity is designed to answer a simple question:
Who can consistently forecast better than the market?
Instead of treating market price as the final output, Probity treats it as the benchmark. Miners are rewarded for generating calibrated probabilities that outperform consensus over time. Repeated outperformance increases influence. Poor calibration and short-term luck wash out.
That structure creates an open competitive environment for superforecasting, where performance is measurable, auditable, and economically enforced.
Scoring and Incentives
Probity measures skill as the difference between market log-loss and miner log-loss:
Skill = LogLoss_market - LogLoss_miner
Positive skill means a miner outperformed the market.
Rolling skill smooths noise and emphasizes consistency across many events.
Miner influence grows according to:
Weight = exp(beta * RollingSkill)
This creates a continuous incentive surface where sustained accuracy matters more than one-off wins, and simply mirroring the market offers little to no long-term edge.
Core Features
permissionless miner participation
deterministic validator scoring
commit-reveal anti-copy mechanism
market-relative forecasting evaluation
skill-weighted probability ensemble generation
auditable score and weight routing
live oracle and leaderboard API outputs
Use Cases
Probity can serve as a probability intelligence layer for:
DeFi derivatives and structured products
parametric insurance
DAO governance and decision support
risk monitoring and analytics
prediction platforms that need machine-readable probability signals
Why Bittensor
Probity is naturally aligned with Bittensor because Bittensor is built to evaluate intelligence competitively and route rewards based on measurable performance.
Bittensor provides:
on-chain weight setting tied to validator scoring
decentralized competition between miners
permissionless participation
transparent network-level incentive mechanics
Probity uses those primitives to make forecasting skill legible, comparable, and economically valuable.
In centralized systems, forecasting reputation is opaque and platform-controlled.
In Probity, probabilistic performance is continuously evaluated in the open and translated into network influence.
Current Stack
Current implementation includes:
Python
Bittensor SDK
Polymarket market data ingestion
deterministic validator scoring
REST API and dashboard for oracle and leaderboard outputs
Planned extensions include:
expanded probability API infrastructure
EVM-compatible oracle integrations
Resources
During Round I, we focused on designing Probity’s core incentive system and validating its economic foundations. This included formalizing the subnet’s scoring model, defining strictly proper log-loss evaluation, specifying market-relative benchmarking, rolling skill aggregation, exponential weight routing, and the commit-reveal anti-copy mechanism. We also published the public repository, documented miner-validator interactions, and analyzed key edge cases such as market mirroring, identity splitting, short-term variance dominance, and collusion.
During Round II, Probity advanced from architecture and prototype into a working subnet implementation. We implemented live miner and validator loops on Bittensor, the full commit-reveal forecasting workflow, deterministic validator-side scoring, rolling skill tracking, exponential weighting, and SWPE oracle generation. We also integrated live Polymarket event ingestion, dual-source market resolution checks, persistent state handling, and a REST API plus dashboard for oracle and leaderboard visibility. The project now reflects a functional testnet-ready system rather than just a scoring prototype.
At this stage, Probity is no longer only a mechanism design concept. It is an operational benchmark-driven forecasting subnet with working miner-validator coordination, live event handling, deterministic scoring, and testnet deployment progress. The next phase is focused on improving forecasting quality, strengthening operational robustness, and extending Probity into a production-grade programmable probability layer.
Probity is currently pre-seed and self-funded.
We have not raised external capital and are focused on mechanism design, subnet implementation, and testnet validation during the hackathon phase.
We are open to strategic ecosystem support and partnerships aligned with long-term subnet sustainability, particularly within the Bittensor ecosystem.