Fills, payouts and results in Polymarket's crypto Up or Down markets since September 2025, tagged with the coin, window length and window times in UTC.
We decode Polymarket's crypto Up or Down markets from Polygon ourselves and keep only those: Bitcoin, Ethereum, Solana, XRP, Dogecoin, BNB, Hyperliquid and Zcash, in 5-minute, 15-minute, hourly, 4-hour and daily windows. That comes to about 1.19 billion fills, 689 million settlement rows and 754,000 markets since 1 September 2025. Every row names the coin, the window length and the window's start and end in UTC, worked out from the market title, which gives the time in ET and no year.
Most buyers start with one coin and one window length. Filter on UpDown_Asset and UpDown_Interval, sum fills where Trade_IsTakerSide is false for volume, and join payouts on the condition id. The Resolved row in the markets table names the winner.
Use cases
Backtest 5-minute, 15-minute and hourly Up or Down strategies per coin
Odds path inside each window, fill by fill, up to the result
Wallet profit and loss on short crypto markets from fills and payouts
Market-maker and bot activity by coin, window length and hour
3 tables
132 columns · free sample for each table
prediction_trades
53 columns
One row per order fill in a crypto Up or Down market, with both addresses, shares, price, fee, USD value, the market question and outcome, and the coin and window.
matic/polymarket_up_or_down/prediction_trades<start_block>_<end_block>.parquet50 blocks per file
One row when a crypto Up or Down market is created and one when it is resolved, with the question, rules, outcomes, oracle, the winner on results, and the coin and window.
matic/polymarket_up_or_down/prediction_managements<start_block>_<end_block>.parquet50 blocks per file
prediction_trades53 columnsOne row per order fill in a crypto Up or Down market, with both addresses, shares, price, fee, USD value, the market question and outcome, and the coin and window.
ColumnTypeDescription
Block3
Block_NumberUInt64Polygon block of the fill.
Block_TimeDateTimeTime of the fill's block, UTC.
Block_DateDateUTC date of the fill.
Call2
Call_Signature_NameStringExchange function that made the fill: matchOrders.
Call_IndexUInt32Index of the call inside the transaction. With Transaction_Hash and Log_Index it gives a unique row key.
Log3
Log_Signature_NameStringEvent name: OrderFilled.
Log_SmartContractStringExchange contract that emitted the event.
Log_IndexUInt32Index of the event inside its call; it restarts in each call.
Trade_OutcomeTrade13
Trade_OutcomeTrade_BuyerStringAddress that got the outcome shares. When it is the exchange contract, the row is the taker's side of a match (see Trade_IsTakerSide).
Trade_OutcomeTrade_SellerStringAddress that gave up the outcome shares.
Trade_OutcomeTrade_AmountDecimalShares filled, to 6 decimal places.
Trade_OutcomeTrade_CollateralAmountDecimalCash paid or received for those shares: USDC.e before April 2026, pUSD after.
Trade_OutcomeTrade_CollateralAmountInUSDDecimalThe cash amount times the hourly USDC.e price.
Trade_OutcomeTrade_OrderIdStringOrder hash, stored as its hex text hex-encoded again, as in the Polymarket dataset.
Trade_OutcomeTrade_PriceDecimalPrice per share in cash: CollateralAmount / Amount, between 0 and 1.
Trade_OutcomeTrade_PriceInUSDDecimalPrice times the hourly USDC.e price.
Trade_OutcomeTrade_IsOutcomeBuyBoolTrue when the order bought outcome shares.
Trade_OutcomeTrade_Fee_AmountDecimalFee charged on this fill, in the cash token; 0 on every fill before January 2026.
Trade_OutcomeTrade_Fee_Currency_SymbolStringFee token symbol; empty when there is no fee.
Trade_OutcomeTrade_Fee_PayerStringAddress that paid the fee.
Trade_OutcomeTrade_Fee_RecipientStringAddress that got the fee.
Trade_Prediction24
Trade_Prediction_CollateralToken_NameStringName of the cash token.
Trade_Prediction_CollateralToken_SymbolStringCash token symbol: USDC (the bridged USDC.e) or pUSD.
Trade_Prediction_Question_TitleStringThe market question as shown on the site. Filled from the markets table when the fill lacks it.
Trade_Prediction_Question_ResolutionSourceStringSource named to decide the market.
Trade_Prediction_Question_ImageStringImage URL shown with the market.
Trade_Prediction_Question_MarketIdStringNumeric market id on the site.
Trade_Prediction_Question_IdStringId of the question behind the condition.
Trade_Prediction_Question_CreatedAtDateTimeWhen the market was created, UTC.
Trade_Prediction_Question_GroupIdStringEvent id that groups related markets; 0x when the market stands alone.
Trade_Prediction_Outcome_IdStringOutcome position id, the same as OutcomeToken_AssetId.
Trade_Prediction_Outcome_IndexInt32The outcome's index in its condition, from 0.
Trade_Prediction_Outcome_LabelStringOutcome name, such as Yes, No, Up or Down; taken from the market's outcome list when the fill has none.
Transaction2
Transaction_FromStringAddress that sent the transaction. Fills arrive through matchOrders calls sent by Polymarket's relayers (about 2,000 addresses on a sampled day), so this is not the person on either side.
Transaction_HashStringHash of the fill's transaction.
TransactionStatus1
TransactionStatus_SuccessBoolFalse when the transaction failed.
Trade1
Trade_IsTakerSideBoolTrue when the exchange contract is the buyer or seller: the row restates the taker's side of a match whose maker rows are already there. Sum rows where it is false for market volume.
UpDown4
UpDown_AssetStringCoin the fill's market is on: BTC, ETH, SOL, XRP, DOGE, BNB, HYPE or ZEC.
UpDown_IntervalStringLength of that market's window: 5m, 15m, 1h, 4h or 1d.
UpDown_WindowStartDateTimeStart of the price window the market settles on, UTC.
UpDown_WindowEndDateTimeEnd of that window, UTC. Daily windows run from noon ET to noon ET.
prediction_settlements46 columnsOne row per split, merge or payout of positions in a crypto Up or Down market, with the holder, shares, cash, USD value, and the coin and window.
ColumnTypeDescription
Block3
Block_NumberUInt64Polygon block of the event.
Block_TimeDateTimeTime of the block, UTC.
Block_DateDateUTC date of the event.
Call2
Call_Signature_NameStringFunction called: splitPosition, mergePositions or redeemPositions.
Call_IndexUInt32Call position in the transaction; part of the row key with Transaction_Hash and Log_Index.
Log3
Log_Signature_NameStringEvent name: PositionSplit, PositionsMerge or PayoutRedemption.
Log_SmartContractStringContract that emitted the event: Conditional Tokens or the NegRisk adapter.
Log_IndexUInt32Event position within its call, counted from 0 in each call.
Settlement3
Settlement_EventTypeStringSplit, Merge or Redemption.
Settlement_HolderStringAddress whose cash or positions moved.
Settlement_OutcomeTokenIdsStringERC-1155 position ids involved, as text in the form ['id1', 'id2'].
Settlement_Amounts4
Settlement_Amounts_AmountDecimalShares split, merged or redeemed, to 6 decimal places.
Settlement_Amounts_AmountInUSDDecimalUSD value of those shares at the time, equal to CollateralAmountInUSD.
Settlement_Amounts_CollateralAmountDecimalCash locked by a split, freed by a merge, or paid out by a redemption (0 for a losing position).
Settlement_Amounts_CollateralAmountInUSDDecimalThe cash amount times the hourly USDC.e price; 0 for cash tokens outside Polymarket's own (under 0.1% of rows).
Settlement_Prediction25
Settlement_Prediction_CollateralToken_AssetIdStringAsset id of the cash token; 0 for ERC-20.
Settlement_Prediction_CollateralToken_BalanceSlotIndexStringStorage slot of the cash token's balances; empty when 0.
Settlement_Prediction_CollateralToken_DecimalsInt32Cash token decimals: 6 for Polymarket's cash tokens.
Settlement_Prediction_CollateralToken_SmartContractStringContract of the cash token.
Settlement_Prediction_CollateralToken_SymbolStringCash symbol: USDC (the bridged USDC.e), WCOL (the NegRisk adapter's wrapped USDC.e) or pUSD.
Settlement_Prediction_ConditionIdStringCondition id of the market, used to join fills and market events.
Settlement_Prediction_Marketplace_SmartContractStringConditional Tokens contract or the NegRisk adapter.
Settlement_Prediction_Marketplace_ProtocolNameStringName of the protocol: polymarket.
Settlement_Prediction_Marketplace_ProtocolFamilyStringFamily of the protocol: Gnosis_CTF.
Settlement_Prediction_Outcome_LabelStringOutcome name; empty on splits and merges, which cover every outcome of the market.
Settlement_Prediction_Outcome_IndexInt32Index of the outcome, from 0.
Settlement_Prediction_Outcome_IdStringOutcome position id.
Settlement_Prediction_OutcomeToken_SmartContractStringERC-1155 contract of the position.
Settlement_Prediction_OutcomeToken_SymbolStringSymbol of the position token; usually empty.
Settlement_Prediction_OutcomeToken_AssetIdStringERC-1155 token id of the position.
Settlement_Prediction_OutcomeToken_BalanceSlotIndexStringStorage slot of the position balances; empty when 0.
prediction_managements33 columnsOne row when a crypto Up or Down market is created and one when it is resolved, with the question, rules, outcomes, oracle, the winner on results, and the coin and window.
ColumnTypeDescription
Block3
Block_NumberUInt64Polygon block of the market event.
Block_TimeDateTimeBlock time in UTC.
Block_DateDateUTC date of the market event.
Call2
Call_Signature_NameStringFunction that made the event.
Call_IndexUInt32Call number within the transaction, part of the row key.
Log3
Log_Signature_NameStringEvent name: TokenRegistered for a new market, ConditionResolution for a decided one.
Log_SmartContractStringContract that emitted the event.
Log_IndexUInt32Event number within its call; it restarts per call.
Management2
Management_EventTypeStringCreated (the market is set up) or Resolved (the market is decided).
Management_DescriptionStringThe market's full rules text.
Management_Group3
Management_Group_IdStringEvent id that groups related markets; 0x when none.
Management_Group_TitleStringTitle of that event.
Management_Group_DescriptionStringText that explains that event.
Management_Prediction14
Management_Prediction_Question_IdStringQuestion id, as in the other two tables.
Management_Prediction_Question_MarketIdStringThe site's numeric id for the market.
Management_Prediction_Question_TitleStringMarket question. Filled from the market's other rows when this one lacks it.
Management_Prediction_Question_CreatedAtDateTimeMarket creation time, UTC.
Management_Prediction_Question_ResolutionSourceStringSource named to decide it.
Management_Prediction_Condition_IdStringCondition id of the market; fills and settlements join on it.
Management_Prediction_Condition_OracleStringOracle contract that may decide the market.
Management_Prediction_Condition_Outcomes_LabelStringAll outcome names in index order, as text in the form ['Up', 'Down'].
Management_Prediction_Outcome_IdStringOutcome position id. On Resolved rows, the winning outcome.
Management_Prediction_Outcome_IndexInt32Outcome index, from 0. On Resolved rows, the winner.
Management_Prediction_Outcome_LabelStringOutcome name: the winner on Resolved rows, empty on Created rows.
Management_Prediction_CollateralToken_SmartContractStringCash token contract of the market.
Management_Prediction_CollateralToken_SymbolStringSymbol of the market's cash token.
Management_Prediction_Marketplace_SmartContractStringExchange or adapter contract the market is listed on.
Transaction2
Transaction_FromStringSender address of the transaction.
Transaction_HashStringHash of the event's transaction.
UpDown4
UpDown_AssetStringCoin the market tracks.
UpDown_IntervalStringHow long the market's window runs: 5m, 15m, 1h, 4h or 1d.
UpDown_WindowStartDateTimeOpening time of the market's price window, UTC.
UpDown_WindowEndDateTimeClosing time of the window, UTC. On Resolved rows the result comes after it.
Send this dataset’s full schema to an assistant and ask it anything. It reads the plain-text brief ↗ first, so the answer comes from the real column list rather than a guess.
The same three Parquet tables as our Polymarket dataset, cut to crypto Up or Down markets and joined by one key, the market's condition id. The trades table has one row per order fill, the settlements table one row per split, merge or payout, and the markets table one row when a market is created and one when it is resolved. Every table adds four columns: UpDown_Asset, UpDown_Interval, UpDown_WindowStart and UpDown_WindowEnd.
Which markets are covered?
Up or Down markets on eight coins: BTC, ETH, SOL, XRP, DOGE, BNB, HYPE (titled Hyperliquid or HYPE) and ZEC. By count since September 2025: about 488,000 5-minute markets, 200,000 15-minute, 52,000 hourly, 12,000 4-hour and 2,200 daily. Up or Down markets on stocks, indices, currencies, gold and oil, and on polls, are left out. The Polymarket dataset has them.
How are the window times worked out?
From the market title, which gives the date and time in ET with no year. We convert to UTC, allowing for clock changes and windows that cross midnight. Where a title fits more than one moment, such as the same title a year apart or the repeated 1 AM hour on 2 November 2025, we take the reading that ends nearest the market's resolution. The 2 AM hour that the clock skips on 8 March 2026 is read as 3 AM EDT, the same instant. Hourly windows are the hour starting at the named time. Daily windows run from noon ET the day before to noon ET on the named date.
How do these markets resolve?
Each market's rules are in Management_Description in the markets table. Markets from 5 minutes to 4 hours settle on Chainlink: Up if the end price is at or above the start price, and newer ones use a time-weighted average over the window in place of the end price. Hourly markets use the Binance 1-hour candle: Up if the close is at or above the open. Daily markets compare Binance closes at noon ET a day apart; if the two are equal, the market resolves 50-50.
How big is it?
About 1.19 billion fills, 689 million settlement rows and 1.4 million market events since 1 September 2025, roughly 200 GB as Parquet, going by the sample files. The last 6 months hold about 668 million fills and the latest month about 67 million. Exact figures arrive with the delivery manifest.
How far back does it go?
From 1 September 2025, the first day of steady coverage in our index, to yesterday. The first 15-minute window opens on 13 September 2025, the first 4-hour window on 15 October 2025 and the first 5-minute window on 18 December 2025, all UTC.
How do I get market volume without counting fills twice?
Use the rows where Trade_IsTakerSide is false. Each match also writes one row for the taker's side, with the exchange contract as buyer or seller; on 15 August 2026 those rows were 36% of Up or Down fills. Summing every row nearly doubles volume.
What should I watch out for?
Fees start in January 2026: every fill before then shows 0. Transaction_From is a Polymarket relayer, not either person in the fill; use buyer and seller. Log_Index restarts inside each call, so key rows on hash, call index and log index. A few fills land after the window ends and before the result is posted, when the price sits near 0 or 1.
How are fills valued in USD?
Cash amount times the hourly price of USDC.e from our own price series. Polymarket settles in USDC.e until April 2026 and in pUSD after, and the NegRisk adapter wraps USDC.e as WCOL; all three are priced as USDC.e. Price is cash per share, between 0 and 1. The few settlements in any other token carry a USD value of 0.
How was it checked?
We ran the export queries for 15 August 2026: 1,938,358 fills, 981,792 settlement rows and 6,208 market events, each equal to a direct count of the source with the same filter. No key repeated, every price was between 0 and 1, every fill was labelled Up or Down, and every row got a coin and a window. On the sample blocks, 94,364,050 to 94,364,149, each row equals the matching row of the Polymarket dataset.
How do I load it?
One line in DuckDB: SELECT UpDown_Asset, UpDown_Interval, count(*) FROM read_parquet('prediction_trades/*.parquet') WHERE NOT Trade_IsTakerSide GROUP BY ALL. Pandas works too: pandas.read_parquet(path). The sample files open without a login.
How current is it, and how is it delivered?
Refreshed daily with T+1 latency, so a purchase made today includes everything to yesterday. Delivered as Parquet named by block range, 50 blocks per file, under a stable S3 layout with a JSON manifest listing every file and its sha256. Signed HTTPS links are emailed once the files are ready, or on request we deliver into your own S3, GCS or R2 bucket.
How does it differ from the Polymarket dataset?
It is the same data cut to crypto Up or Down markets, about two-thirds of all Polymarket fills, with four columns added for the coin and window. Buy the Polymarket dataset if you also need sports, politics or other markets.
PolymarketTradesEvents
SolanaLast 12 months