Mecone Why backtests
A settlement oracle is a claim: that the index tracks reality and cannot be cheaply shoved. Claims need receipts. Every backtest here is a walk-forward reconstruction on real, public Polymarket price history: at each timestamp the index uses only what was knowable then, and where the market has resolved, the actual outcome scores it.
No look-aheadEach point is computed from prices published at that time, never later.
Real crowd dataActual order-flow prices from Polymarket, including resolved markets.
Ground truthResolved outcomes (like SpaceX's IPO) score the index, not a model.
ReproducibleEvery chart links its raw JSON artifact; the construction is the one documented in the docs.
Why Polymarket only? Its market data is public and permissively licensed. Kalshi-sourced work is research-only under Kalshi's data terms, so those backtests are shared under NDA rather than published here.
SpaceX valuation (resolved)
Terminal error 6.1% vs the resolved $2T–$2.2T outcome
crash day ~0%terminal err 6.1%6mo
Kraken valuation
Full market life, walked forward
7mo
Openai valuation
Moved ~0% on the day an EMA oracle crashed −45%
crash day ~0%9mo
Anthropic valuation
Moved ~0% on the day an EMA oracle crashed −45%
crash day ~0%62d
Fed rate cuts 2025 (resolved)
Terminal error 0.7% vs the resolved 3 outcome
terminal err 0.7%12mo
Stripe valuation (no-IPO close)
Full market life, walked forward
6mo
US home value rollover
6 real monthly rolls · seam 0.915% mean / 1.83% max
6 rollsmax seam 1.83%6mo
Miami home value rollover
6 real monthly rolls · seam 2.554% mean / 6.267% max
6 rollsmax seam 6.267%6mo
New York City home value rollover
6 real monthly rolls · seam 1.02% mean / 3.577% max
6 rollsmax seam 3.577%6mo
Chicago home value rollover
4 real monthly rolls · seam 2.428% mean / 3.921% max
4 rollsmax seam 3.921%5mo