Live D1 ledger · Five benchmark contracts · Public run receipts

Football benchmark reproducibility you can inspect, run by run

Direct answer: every accepted automation run records which declared benchmark artifacts reproduced, the exact SHA-256 digests compared, the Git commit and the originating public JSON receipt. Empty or unavailable data stays visibly empty—this page never invents a successful run.

Published by Football Proof AI · Schema version research-reproducibility-ledger/1.0.0
Football benchmark reproducibility ledger with five public run checks
APPEND-ONLY RESEARCH LEDGERREPRODUCTION IS NOT ACCURACY · FAILED RUNS REMAIN VISIBLE

Canonical publication record

Abstract

A public, append-only operations ledger for automated benchmark reproduction runs. Each record exposes the declared checks, exact artifact and source digests, workflow producer, Git commit, source-workflow reference that may require repository access and payload SHA-256 without claiming independent review, historical timestamping or model accuracy.

Author and publisher
Football Proof AI
Publication version
research-reproducibility-ledger/1.0.0
Published
Last modified
Release status
Current release
Review status
Editorial technical note; not externally peer reviewed
Production method
Editorial dataset maintained from the declared method, linked sources and any listed reproducible artifacts.
Software and AI assistance
Software, including generative AI, may assist drafting, transformation or quality checks. It is not treated as a source, author or independent reviewer; factual and quantitative claims must remain bound to cited sources, declared methods or reproducible artifacts.
Why this exists
Enable readers to inspect assumptions, reproduce calculations and reject claims that exceed the published evidence.
Commercial boundary
No bookmaker, prediction provider or paid placement controls the evidence rules, calculations or conclusions on this page.
Version history
  1. research-reproducibility-ledger/1.0.0 : Initial public release.
Release resources
  1. Public benchmark reproducibility run API
References
  1. https://footballproofai.com/research/football-1x2-scoring-benchmark
  2. https://footballproofai.com/research/football-1x2-margin-removal-benchmark
  3. https://footballproofai.com/research/football-1x2-probability-calibration-benchmark
  4. https://footballproofai.com/research/poisson-vs-dixon-coles-football-benchmark
  5. https://footballproofai.com/tools/xg-poisson-football-calculator

Latest accepted run · Exact hashes · Public Git receipt

Did the declared benchmark artifacts reproduce?

The status below is read from the public D1 operations ledger. “Reproduced” means the workflow regenerated the declared bytes and matched the recorded digest contract for that check.

Run 29479983533 · attempt 1

All declared checks reproduced

5/5 reproduced
Checked at
Accepted by the ledger
Repository / workflow
YAN555999/proofxi-pipeline · Football Proof AI benchmark reproducibility ledger
Git commit SHA
73dc78ded6f10552d8adf34ff637a8e8ad92aa1e
Public receipt commit SHA
19006a43be2dfe61f7fc03ec03466126b51cbdb1
Public receipt document SHA-256
e3f59110ff550e6cb9b4a6cb77097921c478c785959fcec0070c8bac2c499feb
Canonical payload SHA-256
3a005903bff06bba11c6ba7e760448ad840808630bb99e6aafed2a76a786441f
  1. football-1x2-empirical-benchmark/1.0.0

    football-1x2-scoring-benchmark

    Reproduced
    Manifest path
    data/football-1x2-empirical-benchmark-1.0.0-manifest.json
    Manifest SHA-256
    801f8c6c23e30f945dcba38e4c256fd4e9f3652f7e2ea965e4788f0e7cf9014a
    Generator path
    scripts/generate-football-1x2-empirical-benchmark.mjs
    Generator SHA-256
    7b7d728e4445d92fe90b40c866ae2344e444a5bfb1fdecb504db1f24c3ac07e4
    Source digest SHA-256
    eb44dfa83ebc3e99bff305ddbb7ee0bfcd79ee51496c15953a8607e210a81389
    Artifact digest SHA-256
    6385fa68c6f79504ac29c8ab604f2949d0a33d1b6b86bef77438687f57da4e59
  2. football-1x2-margin-removal-benchmark/1.0.0

    football-1x2-margin-removal-benchmark

    Reproduced
    Manifest path
    data/football-1x2-margin-removal-benchmark-1.0.0-manifest.json
    Manifest SHA-256
    033b8c5deb3b2141214631027d8445469a48dea8a4eb9e56d2536be85e47a1eb
    Generator path
    scripts/generate-football-1x2-margin-removal-benchmark.mjs
    Generator SHA-256
    60968a40b62331f2bdc7987579d9520dd6863e3136de0714aec34dcfe5ccf986
    Source digest SHA-256
    35b4059e73f7157ed8a8565dc89aa29218cd17c106a8e06d329d0e586b76297f
    Artifact digest SHA-256
    91aea76c6521ea9d39635329299373e8c4678258913499fc2c962547582c39ab
  3. football-1x2-probability-calibration-benchmark/1.0.0

    football-1x2-probability-calibration-benchmark

    Reproduced
    Manifest path
    data/football-1x2-probability-calibration-benchmark-1.0.0-manifest.json
    Manifest SHA-256
    1a16037252ff5ccd2cecb405894a4605627770ea77030096f2bc5b4d17d392cc
    Generator path
    scripts/generate-football-1x2-probability-calibration-benchmark.mjs
    Generator SHA-256
    7a995e72cccc438c4574877999863abebed399a3350bf00315cbee347875a0ca
    Source digest SHA-256
    0bcf11b5792d2c6b4e251e96d2443dfbd5ebb867928e70abe42ffb6fee2e97c9
    Artifact digest SHA-256
    fcd92a4df3c582c4019f2cae00adef275bb3b990f8cf5e35edf5007ba48f8617
  4. football-independent-poisson-scenario-grid/1.0.0

    football-poisson-scenario-grid

    Reproduced
    Manifest path
    data/football-poisson-scenario-grid-1.0.0-manifest.json
    Manifest SHA-256
    d020ead55eb34871780c2d440c9b45cd43f140b8e98ebd512e0e20a04e0999fa
    Generator path
    scripts/generate-football-poisson-scenario-grid.mjs
    Generator SHA-256
    0f8240cd3c2bc693058683e04df36e6eb83dcf56853f719bd8da1072a6446b3f
    Source digest SHA-256
    368fbc5a8ad8ae6d5249427d8ce38ce06d6b8d10aad1c6e872b0f1c965c47238
    Artifact digest SHA-256
    5aafda0ee94e44b5d038b75cff335c91d6b24820f820650ac55aee260c12d4a1
  5. poisson-vs-dixon-coles-football-benchmark/1.0.0

    poisson-vs-dixon-coles-football-benchmark

    Reproduced
    Manifest path
    data/poisson-vs-dixon-coles-football-benchmark-1.0.0-manifest.json
    Manifest SHA-256
    79314f2517b6e940aa3edb770afe91485af90ec97be4f8df498bc8d8ddf02601
    Generator path
    scripts/generate-poisson-vs-dixon-coles-benchmark.mjs
    Generator SHA-256
    320c647988c25908312beb7f411aa8d9231ed97de7e5f53ca964cbcafb862b75
    Source digest SHA-256
    af2adb402e9ef6aa87668a9c92d0a85da5fc7b8adb5b03c58e9fbe4c8362dbe7
    Artifact digest SHA-256
    d72c4a6d900a06afbbd7cfa980de308b31407b5251b14ce4ee43842f65255ced

Recent history · Up to five accepted entries

Reproducibility run history

Each row preserves its own result, payload digest and public receipt link. The highlighted latest receipt is kept separate; this window contains up to five earlier accepted runs.

History is empty

No accepted run receipts are available to list

This is an honest absence, not a zero-failure result. The API remains the machine-readable source of truth.

Coverage contract · Five public benchmark surfaces

Which football benchmarks does the workflow check?

These are the five canonical pages represented by the declared benchmark IDs. Open a page to inspect its method, versioned distribution and benchmark-specific evidence boundary.

  1. 01

    Football 1X2 empirical and scoring benchmark

    football-1x2-scoring-benchmark

  2. 02

    Football 1X2 proportional versus Shin margin-removal benchmark

    football-1x2-margin-removal-benchmark

  3. 03

    Premier League 1X2 probability calibration benchmark

    football-1x2-probability-calibration-benchmark

  4. 04

    Football xG Poisson scenario grid

    football-poisson-scenario-grid

  5. 05

    Poisson versus sequential Dixon–Coles EPL benchmark

    poisson-vs-dixon-coles-football-benchmark

Three-layer verification · Byte contract, public receipt, source reference

How to verify a reproducibility run yourself

Do not rely on the green label alone. Follow the links and compare the same fields exposed in the JSON dataset.

  1. 01

    Read the declared byte contract

    Match the benchmark version, manifest path, generator path and every published SHA-256 value.

  2. 02

    Inspect the public receipt

    Open the public Git receipt and compare its exact checked time, repository, workflow, run attempt, Git commit and five check rows with this ledger.

  3. 03

    Bind the public receipt

    Compare the canonical payload SHA-256 and check rows in the public API with the visible run record above.

Benchmark reproduction result
Whether the declared regenerated output matched the check contract.
Manifest SHA-256
Digest of the benchmark's declared reproducibility manifest bytes.
Generator SHA-256
Digest of the generator bytes used by the declared check.
Source digest SHA-256
Digest binding the manifest-declared source, derivation and formula records, not a licence or provenance verdict.
Artifact digest SHA-256
Digest of the committed artifact set targeted by the check; on success, regenerated outputs matched it.
Error digest SHA-256
Hash of normalized internal failure metadata. It does not reveal or independently verify the failure reason; detailed source logs may require private repository access.
Canonical payload SHA-256
Parse the exact public receipt, recursively sort object keys lexicographically, preserve array order, serialize compact UTF-8 JSON and apply SHA-256. The result must equal this field.
Public receipt document SHA-256
SHA-256 of the exact pretty-printed JSON file bytes read back anonymously from the commit-pinned public URL before D1 acceptance.
Unavailable digest sentinel
On a failed check, an unreadable file or source set is represented deterministically as SHA-256 of the UTF-8 string footballproofai-research-reproducibility:unavailable:<kind>:<path>, directly or inside the artifact-set digest. It is not evidence that those bytes were read.
Public receipt and source reference
The exact payload is public in a separate Git history. The source workflow URL is also retained, but its private repository may require access.

Evidence boundary · Reproduction is not independent validation

What a successful check proves—and what it cannot prove

A successful check supports one narrow claim: in the recorded workflow, the declared generator and inputs reproduced the declared artifact digest. A SHA-256 binds bytes; it is not an independent timestamp, audit opinion or accuracy certificate.

  • It does not prove source rights. A matching digest does not grant a licence to redistribute third-party data.
  • It does not prove preregistration. The ledger records an accepted run; it cannot independently establish when a method was first conceived.
  • It does not prove predictive accuracy. Rebuilding an artifact is separate from whether a model forecasts future matches well.
  • It does not prove historical publication time. GitHub and D1 timestamps are operational metadata, not an independent trusted timestamping service.
  • It does not prove live production health. A benchmark run is not an uptime, data-freshness or prediction-pipeline status check.
  • It does not guarantee future reproduction. Toolchains, external services and source availability can change after a successful run.
  • It does not make private source public. The JSON receipt is public; inspecting the originating workflow or repository may still require access.
  • Licence boundary. CC BY 4.0 covers only Football Proof AI's original run-ledger selection, schema and editorial metadata. It does not license benchmark source inputs, GitHub artifacts, third-party datasets or trademarks.