Demo -- read-only snapshot of the live system as of 2026-07-16T01:45:31Z; no live data, paper units only.
Skip to main content

evidence / Frontier measurements

Novel Analytics -- Uniquely Auditable Measurements

Every number below is copied verbatim from a JSON artifact written by a showcase module that ran once against local data on disk. Nothing here is re-derived from memory. The single truth-source for any figure is docs/JOB_EVIDENCE_PACKET.md. The product is a calibrated predictor, not an edge product -- an honest REJECT or null is a success, and where a live market matches or beats us that is stated plainly, in numbers.

strongest single receipt

The honest claim on this page is not "no one else can measure these." Every method below already exists in the literature, and each analytic ships with a prior-art receipt that says so -- two verdicts of INCREMENTAL, one ALREADY_DONE_ON_CORE_METHOD, and one N/A (an internal hygiene artifact that borrows no novelty)

the claim

The honest claim on this page is not "no one else can measure these." Every method below already exists in the literature, and each analytic ships with a prior-art receipt that says so -- two verdicts of INCREMENTAL, one ALREADY_DONE_ON_CORE_METHOD, and one N/A (an internal hygiene artifact that borrows no novelty). The word "novel" here means one thing only: uniquely auditable.

What is uniquely auditable is the combination -- each measurement is (a) run across the multi-sport corpora this system actually holds (NBA, MLB, soccer, tennis), (b) preregistration- and mask-gated so thin buckets can never masquerade as findings, (c) provenance-stamped to a JSON you can open and a module you can re-run to the same number, and (d) shipped with its own honest prior-art verdict, including the ones that say "this is not our method." The market beats our in-game model at nearly every checkpoint below, and this page says so before it says anything else. That spread of honest verdicts -- not a novelty claim -- is the transferable thing. No entry claims "first ever"; each carries the receipt that would refute it.

All five modules live in scripts/platformkit/analytics_showcase/, each cross-referenced in docs/ANALYTICS_CATALOG.md. Every one ran once on 2026-07-22.


cited artifacts

committed artifact
scripts/platformkit/analytics_showcase/out/info_arrival_curve.json
scripts/platformkit/analytics_showcase/out/market_overreaction.json
scripts/platformkit/analytics_showcase/out/mechanism_survival.json
scripts/platformkit/analytics_showcase/out/comeback_atlas.json
scripts/platformkit/analytics_showcase/out/kernel_transfer.json
scripts/platformkit/analytics_showcase/out/murphy_decomposition.json
scripts/platformkit/analytics_showcase/out/soccer_calibration_pack.json
scripts/platformkit/analytics_showcase/out/tennis_showcase.json
Per-checkpoint model / market / naive Brier, MLB innings and soccer 5-min buckets
Moved-to-minus-outcome by price-move magnitude bucket, MLB and soccer
Hypothesis survival rate by sport and by mechanism category
NBA in-play reliability atlas: model vs market vs realized frequency by lead band x time remaining
Cross-sport Murphy reliability composition, with the two comparable sports and the reasons the other two are not

why this matters

Five measurements, five honest verdicts: two INCREMENTAL, one ALREADY_DONE_ON_CORE_METHOD, one N/A internal-hygiene artifact, and -- across all of them -- a market that is sharper than our in-game model at nearly every checkpoint. None of that is a novelty claim, and that is the whole point. The field has these instruments; what this system adds is that each one runs across the multi-sport corpora we actually hold, behind preregistered masks that keep thin buckets from becoming findings, writes a provenance-stamped JSON you can re-run to the same number, and ships with the prior-art receipt that would refute any "first-ever" boast. The transferable skill is not a new metric -- it is measurements built so the honest reading is the only available one, and so the ones that lose to the market say so in numbers. Full per-analytic caveats live in docs/ANALYTICS_CATALOG.md; the truth-source for every figure is docs/JOB_EVIDENCE_PACKET.md.


no edge claimed

This site reports calibration and sharpness only, never a dollar edge, ROI, or bankroll result. An honest null is a success.