DV-PMI Surveillance · Wash trading
Wash Trading (Tier 1)
Cumulative through week 2023-W17
Tier 1 (self-matched trades, where the same wallet is the maker and the taker on a single match) finds 905 matches across the full 712,317,074-trade panel (0.0001% of trades, 0.0008% of dollar volume), all confined to the 2022-2023 launch window. The CLOB matching engine does not permit a wallet to fill its own resting order under normal operation; the residual count is consistent with pre-launch test transactions and contract-deployment artifacts. Tier 1 rules out the simplest, most mechanical regime of wash trading. It does not rule out the harder regimes (round-trip via different counterparties, linked-wallet clusters). For those, see the limitations and superseding-method notes below.
| Share by count | Share by volume | |
|---|---|---|
| Self-matched (Tier 1) | 0.0001% | 0.0008% |
| All other trades | 99.9999% | 99.9992% |
The denominator is the full on-chain panel: every resolved Polymarket match from launch (Nov 2022) to the most recent refresh. The self-matched fraction is the share of matches where the maker and taker address are equal.
Top affected markets
Markets ranked by number of self-matched trades. Most are early bootstrap markets where the venue operator or a deployment-test wallet seeded liquidity by trading against itself before public launch.
| Market | Market ID | Self-matched trades | Volume |
|---|---|---|---|
| Will Donald J. Trump win the 2024 Republican presidential nomination? | 240380 | 100 | $4K |
| Will the Fed increase interest rates by 0 bps after its May meeting? | 249497 | 77 | $569 |
| Will 'Creed III' gross more than $35 million domestically on its opening weekend? | 248974 | 47 | $147 |
| Will U.S. inflation be greater than 0.1% from March to April 2023? | 250113 | 23 | $272 |
| Will Donald J. Trump be indicted by March 31, 2023? | 247266 | 17 | $873 |
| Will the price of $ETH be above $1700 on February 10? | 248578 | 17 | $111 |
| Will 'Creed III' gross more than $40 million domestically on its opening weekend? | 248975 | 16 | $183 |
| Will Arbitrum airdrop happen by March 31st? | 248890 | 16 | $10K |
| Will Andrew Tate be released by April 30? | 248820 | 14 | $3K |
| 2023 FIDE World Chess Championship: Ian Nepomniachtchi vs. Ding Liren | 249524 | 13 | $13K |
By year
| Year | Self-matched trades | Volume |
|---|---|---|
| 2022 | 109 | $18K |
| 2023 | 796 | $216K |
Methodology
A trade is counted as Tier 1 wash if maker_address == taker_address (case-insensitive). FINRA Rule 6140 economic analog: a trade between two accounts under common beneficial control with no change in ownership.
The test is implemented as a single pass through the full processed-trades file via duckdb: { maker_address.lower() == taker_address.lower() }. Pre-filter to the small self-matched subset, then aggregate counts, volumes, top markets, and weekly time series. No sampling; the denominator is the entire panel.
Scope of this test
Tier 1 catches one specific economic form of wash trading: a single wallet on both sides of a single match. A wallet that splits a coordinated wash across two addresses it controls is not detected here. Neither is a single wallet that buys and sells the same outcome through different counterparties in rapid succession. The near-zero Tier 1 result rules out volume inflation from the simplest mechanism; it does not rule out wash trading more broadly.
Superseding method (Sirolly, Ma, Kanoria, Sethi 2025)
Sirolly, Ma, Kanoria, and Sethi (Columbia Business School, 2025) propose an iterative network-clustering approach that identifies approximately-closed clusters of counterparties who trade primarily with each other and rarely with the broader market. Applied to Polymarket, they estimate that transaction patterns indicative of wash trading peaked at roughly 60 percent of weekly volume in December 2024, persisted through April 2025, and reached approximately 20 percent again in October 2025.
Their network-based method is the right tool for the wash-detection question. The structural counterparty-persistence screen we previously published as Tiers 2 and 3 captures the precondition for cluster wash but does not check closed-cluster topology, and is consequently dominated by competing algorithmic market makers rather than likely-manipulative activity. We have removed those threshold tables from this page pending implementation of a network-cluster screen consistent with Sirolly et al's method. The raw structural-screen output remains available via the data downloads below for reproducibility.
Data
Tier 1: surveillance_wash_latest.json · surveillance_wash_t1_weekly.csv
Structural-screen output (Tiers 2 and 3 v1, not currently surfaced on the page): surveillance_wash_tier2_latest.json · surveillance_wash_tier2_top_wallets.csv · surveillance_matched_latest.json · surveillance_matched_top_pairs.csv
Other surveillance indices
← All surveillance indices · Informed Trading (PII) · Insider Timing · Adverse Selection · Resolution Surprise · Concentration
Cite this index
@misc{dellavedova2026wash,
title = {Wash Trading Index (Tier 1)},
author = {Della Vedova, Joshua},
year = {2026},
url = {https://jdellavedova.com/surveillance/wash-trading}
}