Dataset Viewer
Auto-converted to Parquet Duplicate
by_price_band
dict
changelog
list
citation
dict
contains_holdings
bool
contains_pnl
bool
contest
string
decomposition
dict
field_dictionary
dict
generated_at
timestamp[s]
how_to_replay
dict
method
dict
sample
dict
schema
string
three_party_close
dict
window
dict
{ "0.00-0.05": { "fee_pct": 4.8728, "gross_pct": -54.5928, "n": 113930, "notional_usd": 2366267.59, "taker_net_pct": -59.4655 }, "0.05-0.10": { "fee_pct": 4.6208, "gross_pct": -47.8334, "n": 60933, "notional_usd": 2402103.83, "taker_net_pct": -52.4542 }, "0.10-0.20": { ...
[]
{ "dataset": "prediction-market-zero-sum-ledger", "how_to_cite": "Ailin Sun. \"Polymarket zero-sum flow ledger.\" Version 858973c2494c (2026-08-13), window 2025-12-11 to 2026-08-13. Retrieved from https://github.com/ailinsun/MarketFlow/blob/v0.1.2/data/zero_sum_ledger/2026-08-13-858973c2494c.json", "immutable_url...
false
false
think a row is wrong? https://github.com/ailinsun/MarketFlow/issues — corrections are published in the snapshot changelog
{ "fee_over_gross_x": 10.423, "protocol_fee_pct": 1.4879, "taker_gross_selection_pct": 0.1427 }
{ "fee_over_gross_x": "protocol take divided by the crowd's entire gross edge", "maker_net_pct": "the counterparty side of the same trades; it pays no taker fee, so this is the mirror of gross selection", "protocol_pct": "taker fees collected by the venue over the same notional", "sum_pct": "the three rows abov...
2026-08-13T07:24:56
{ "caveats": [ "This is a census of the top 500 markets by large-print notional — the venue's active core, not the venue. Long-tail markets are outside the frame.", "The collector caps each market at 6,000 trades. The venue API returns newest-first, so a truncated market keeps its most recent 6,000 fills and ...
{ "closure": "taker_net + maker_net + protocol == 0 by construction", "denominator": "taker BUY notional = sum(px * size)", "fee_per_share": "rate * p * (1-p); maker pays zero", "fee_rate": "per-market, read from Gamma feeSchedule; markets without a readable rate are dropped rather than defaulted", "sample_fr...
{ "dropped_trades": { "bad_price": 0, "no_fee_rate": 44138, "no_meta": 0, "sell_leg": 559246, "unsettled": 107170 }, "markets_dropped": 61, "markets_reconciled": 439, "max_trades_per_market": 6000, "share_of_large_print_notional": 0.5256457381151874, "taker_buy_notional_usd": 419438465...
marketflow-zero-sum-ledger-public-v0.1
{ "closes": true, "maker_net_pct": -0.1427, "protocol_pct": 1.4879, "sum_pct": 0, "taker_net_pct": -1.3451 }
{ "basis": "trade timestamps of every reconciled fill", "from": "2025-12-11T20:33:31", "to": "2026-08-13T04:55:43" }

Polymarket zero-sum flow ledger

Three-party aggregate accounting for 1,338,933 taker BUY fills across 439 settled markets, window 2025-12-11 to 2026-08-13. Buy notional: $419.4M. The buying crowd, counterparties and modelled venue fees close to zero by construction.

Files and viewer

The viewer loads the canonical aggregate snapshot only. The dated file is an archival copy of the same observations.

Method and source

See the embedded metadata and repository inventory. Files are frozen aggregate exports from experimental research by Ailin Sun, an independent researcher. Research numeric values are unchanged.

Limitations

Use the embedded method, sampling frame and citation blocks. This is the active core selected by large-print notional, not a venue census. Raw tape and wallet identities are not supplied. Fees are modelled; these are flow accounts, not observed wallet PnL.

Citation and licence

CC BY 4.0 for the original compilation and annotations. Cite Ailin Sun, the exact dataset file and its embedded citation block where present, together with the repository release. Upstream venue material retains its rights. Archived repository release v0.1.2: DOI 10.5281/zenodo.22734802.

Downloads last month
27