Loading
Loading
Notch Protocol is experimental software. Predictions are not financial advice. Use at your own risk. Read full disclaimer
Qais Alassa · Osama Alashqar
Research Paper
Strictly Proper Scoring Rules as Trustless Merit Filters in Decentralized Prediction Systems
We introduce calibration-gated reputation: a mechanism in which agents commit cryptographic hashes of probabilistic predictions, reveal them after a mandatory delay, and accumulate reputation through a strictly proper scoring rule. We prove that strict propriety is both sufficient and necessary for this construction. Five theorems validated through 10,000 Monte Carlo trials.
16 pages · 5 theorems · 30 references · Targeting AFT 2026
Technical Specification
A Cryptographic Framework for Verifiable Prediction Scoring and Alpha Commoditization
The companion technical specification covering the commitment scheme, Brier score calibration, difficulty adjustment, the instrument layer (Alpha Passes, Futures, Indices), dual oracle architecture (Chainlink + Pyth), and $NOTCH token economics.
DOI: 10.5281/zenodo.19118356
Key Results
Strict propriety is necessary
Under any proper but not strictly proper scoring rule, Sybil adversaries can produce wallets indistinguishable from skilled predictors.
Accuracy-only scoring is insufficient
Reputation based solely on directional accuracy admits polynomial-cost Sybil attacks regardless of prediction count.
Merit-gating
Under the Brier Score with commit-reveal, no Sybil strategy produces high-calibration wallets without genuine skill. Validated: 0/10,000 trials.
Sybil cost amplification
The cost of faking reputation scales at least linearly in Sybil identities. Empirical: 264x cost ratio at K=1,000.
Stationary convergence
Under temporal decay, each predictor's score converges to a unique value monotonically increasing in skill.