MSR Decode Discuss a matter
Case WorkMSR Decode

Public forensic investigations, documented from question to finding.

Each case below started as a bounded question about a wallet, a market or a claim, and was tested against the underlying public record. A negative or unresolved result is not a failed investigation — it is published the same way a confirmed one is. Systematic, cross-market testing of the mechanisms these cases surface lives separately at Integrity Research.

  1. Trading Conduct · Market Data CASE FILE 06 · Published 2026-08-19, updated 2026-08-27

    The cartel-arrest markets: no pre-public trading found in two tested events, but $117,153 of volume generated for $171.71.

    Question
    Did informed trading precede two law-enforcement announcements, and is the market's own displayed volume a reliable signal of independent directional participation?
    What MSR reconstructed
    1,849 fills across 8 child markets, full lifetime, reconciled on both sides of every match, cross-checked against Polygon's own OrderFilled events.
    Finding
    No pre-public trading found around either event. Two contracts' displayed volume was generated at minimal net economic cost by the same counterparties trading against each other — not broad independent directional participation.
    TRADE RECONSTRUCTION MARKET MICROSTRUCTURE DATA PROVENANCE
    Read the case →
  2. Market Data Integrity · PnL Accounting CASE FILE 05 · Published 2026-07-28

    The “weather bot” proof wallets: net −$2.35M under reconciliation; the flagship exhibit shown as +$80,749 has lost $2.42M.

    Question
    Do the wallets cited as proof of a viral trading strategy actually show the profit claimed for them?
    What MSR reconstructed
    228,106 fills across 10,111 markets, Nov 2024–Jul 2026, for all 4 cited proof wallets, from their complete public fill history.
    Finding
    Combined, the cited wallets are net negative. The flagship exhibit's cash flows contradict its displayed profit metric. One of five cited exhibits does reconcile.
    PNL ACCOUNTING TRADE RECONSTRUCTION
    Read the case →
  3. Market Data · API Integrity CASE FILE 04 · Method Report

    A trade-endpoint completeness defect behind apparent phantom positions: 150 of 150 flagged redemptions reconciled once the default view was replaced.

    Question
    Does Polymarket's default trades-endpoint view reliably represent a wallet's complete trade history?
    What MSR reconstructed
    Re-fetched the 50 largest flagged markets per wallet — 150 of 4,725 flagged in total — using takerOnly=false and /activity, scoped to the redeemed outcome token.
    Finding
    All 150 reconciled: every flagged redemption was paid for by the wallet's own purchases. The default view undercounted fills 1-to-80 in the flagship market.
    API SEMANTICS DATA PROVENANCE
    Read the case →
  4. Trading Conduct · PnL Accounting CASE FILE 03 · Published 2026-07-24

    The BNB 5-minute wallet: reconstructed to the cent. $917K realized profit confirmed; underlying timing mechanism reconstructed.

    Question
    Does a viral ~$917K profit claim, and its “23 of 24” win-rate claim, survive complete fill reconstruction?
    What MSR reconstructed
    139 buy fills across 24 of 39 consecutive five-minute markets, one 3.1-hour session, full per-market ledger.
    Finding
    The profit claim is confirmed. The win-rate claim is not (21 of 24, 87.5%). The mechanism is a late-window timing strategy against a non-repricing quote plus one arbitrage — not directional market-reading.
    TRADE RECONSTRUCTION MARKET MICROSTRUCTURE
    Read the case →
  5. PnL Accounting CASE FILE 02 · Sample decode report

    myzbsq — leaderboard PnL reconstructed to one concentrated resolved event. A real $1.22M profit, and n = 1.

    Question
    Is a leaderboard-topping profit the product of a repeatable strategy, or a single concentrated bet?
    What MSR reconstructed
    One wallet, one market, one side, over roughly 14 hours, from the complete public fill log.
    Finding
    The $1.22M profit is genuine — reconciles to 99.6% of the displayed figure. It is also a single resolved event, not a repeatable mechanism.
    PNL ACCOUNTING TRADE RECONSTRUCTION
    Read the case →
  6. Trading Conduct · Historical CASE FILE 01 · Archived — superseded research

    The Leaderboard Autopsy: displayed PnL can't tell you which leaderboard wallets you could copy. The fills can.

    Question
    Can leaderboard wallets be classified by strategy signature well enough to know which are copyable?
    What MSR reconstructed
    Full on-chain fill reconstruction for six classified wallets plus two contrast wallets, pulled 2026-07-11.
    Finding
    The per-wallet strategy-signature classification survives. Displayed PnL alone cannot distinguish a market-maker, an arbitrageur or genuine directional conviction. MSR Decode's first case file — the methodological point survives; most fill counts on the page lack a committed source artifact and should not be cited as current measurements.
    TRADE RECONSTRUCTION
    Read the case →

Have a similar question?

A commissioned investigation runs the same standard against a question that isn't yet public — a wallet, a market, an account cohort, or a claim.

Discuss a matter

← Back to Research