Skip to content

Chrono Score

Phase 0. Buildable now. Zero capital required.

Chrono Score is a forecaster reputation score computed from resolved prediction market positions across Kalshi, Polymarket, and Azuro. It is domain-partitioned, time-decayed, Wang-corrected, and published both as a soulbound attestation (EAS on Optimism) and as a queryable JSON API.

The literature is unambiguous: forecasting skill is real, persistent, concentrated, and measurable.

  • Persistence — Reichenbach 2025 (N=124.5M trades): month-to-month persistence Welch t=5.14, p<0.001. Previous winners have 59% probability of positive next-month PnL vs 40% for previous losers.
  • Out-of-sample — Yang 2026: top-5% wallets sustain 89.4% accuracy in hold-out vs 50.4% for experience-matched non-skilled.
  • Cross-platform — Atanasov 2024: test–retest reliability r=0.38 across IARPA seasons.
  • Selection dominates — Atanasov: top 2% produce 21% accuracy improvement over sub-elite crowds. 3× larger than the prediction-market-vs-poll mechanism effect.

Seven design decisions, each literature-grounded.

  1. Edge per trade. edge_i = d_i × (O_m − P_i), demeaned per market (Deleep 2026). This captures directional correctness weighted by distance from market-implied probability.
  2. Wang correction. Observed prices on real-money platforms embed a systematic pricing wedge (λ ≈ 0.18, Yang 2026). Raw accuracy against raw prices is biased. We evaluate against Wang-corrected benchmarks.
  3. BIN decomposition. Bias, information, and noise are separately measurable (Satopaa 2021). Chrono Score emphasizes discrimination over calibration — Atanasov shows elite forecasters’ advantage is telling events apart, not probability precision.
  4. Domain-partitioned. Skill is not transitive across domains (Bossaerts 2022: 75% of traders are price-sensitive in fewer markets than the mean). Scores are issued per domain (politics, crypto, macro, sports, etc.).
  5. Minimum threshold. 20 resolved markets before classification (Yang 2026 methodology). Below that, no score is issued.
  6. Time-weighted. Early and mid-lifecycle trades weighted more than near-resolution trades (pricing wedge decays with lifetime — Yang 2026).
  7. Reputation-only weighting architecture. Agustin 2026 proves multiplicative (w = r·s) is too weak to overcome 40× stake differentials. We use reputation-only (or log-scaled) weighting in any internal aggregation where reputation is meant to have authority.
  • Not a leaderboard. Ranking is a side effect, not the product. The product is a portable credential.
  • Not a wealth score. Capital size is anti-correlated with edge (Deleep). Large position size does not improve Chrono Score.
  • Not a single number. Domain partitions mean a forecaster may have a 94 in macro and a 51 in sports. We publish all partitions, not a single aggregate.
  • Not transferable or salable. Soulbound by design.

A second score, Execution Score, measures realized PnL net of timing / spread / fee drag. Motivation: Vedova 2026 (N=222M trades) shows bots earn +0.56% ROI at 49.91% accuracy while 51.27%-accuracy retail lose $79M. Execution and accuracy are orthogonal capabilities.

Execution Score is consumed internally by Fast-Forward (position sizing) and is not publicly displayed in v0. Publishing it creates a copycat risk for genuine alpha strategies. Revisit after Phase 1 data.

  • Spec: complete (this page + CONCEPTUAL_MODEL.md §4a)
  • Data sources: Polymarket subgraph, Kalshi API, Azuro subgraph — all read-only, public
  • Smart-contract surface: EAS schema registration + attestation writer. Lightweight. No custody.
  • Frontend: /score/[wallet] + search + domain filter
  • Timeline: 4–8 weeks from dedicated start
  • Team: Dan + contract frontend dev
CHRONO SCOREOn-chain reputation

A soulbound (non-transferable) on-chain record of your forecasting skill. It accumulates across every prediction you make — and unlocks better terms across all Chronomancy products.

Accuracy
Did your predictions resolve correctly?
Calibration
Were your confidence levels well-tuned?
Market volume
How much did you put at stake?
Time in system
Consistent track record over epochs
What it unlocks
CS 1–3
Developing
  • Fast-Forward: +5% exit premium
  • Rewind: Silver tier (70% recovery)
CS 4–6
Established
  • Fast-Forward: +10% exit premium
  • Rewind: Gold tier (100% recovery)
CS 7–10
Elite
  • Fast-Forward: +18% exit premium
  • Rewind: Gold tier + priority access

Full citations in the literature review. Summary pages: skill is real, reputation as infrastructure, two-score architecture.