Polymarket

Founded in 2020, Polymarket is a decentralized prediction market platform that enables traders to speculate on event outcomes by buying and selling outcome tokens.

NautilusTrader provides a venue integration for data and execution via Polymarket's Central Limit Order Book (CLOB) API.

Today the repository exposes two Polymarket implementations:

• The Python adapter in nautilus_trader.adapters.polymarket, which uses the official Python CLOB V2 client library.
• The Rust-native adapter surface in nautilus_trader.polymarket, which NautilusTrader is consolidating toward.

:::warning The two implementations overlap heavily, but they do not behave identically in every area. This guide calls out the current differences where they matter. :::

NautilusTrader supports multiple Polymarket signature types for order signing, which gives flexibility for different wallet configurations while NautilusTrader handles signing and order preparation.

Installation

To install NautilusTrader with Polymarket support:

uv pip install "nautilus_trader[polymarket]"

To build from source with all extras (including Polymarket):

uv sync --all-extras

Examples

You can find live example scripts here.

Binary options

A binary option is a type of financial exotic option contract in which traders bet on the outcome of a yes-or-no proposition. If the prediction is correct, the trader receives a fixed payout; otherwise, they receive nothing. NautilusTrader represents Polymarket outcome tokens as BinaryOption instruments.

Polymarket uses pUSD as the collateral token for trading, see below for more information.

Polymarket documentation

Polymarket offers resources for different audiences:

• Polymarket Learn: Educational content and guides for users to understand the platform and how to engage with it.
• Polymarket CLOB API: Technical documentation for developers interacting with the Polymarket CLOB API.

Overview

This guide assumes a trader is setting up for both live market data feeds and trade execution. The Polymarket integration adapter includes multiple components, which can be used together or separately depending on the use case.

• PolymarketWebSocketClient: Low-level WebSocket API connectivity (built on top of the Nautilus WebSocketClient written in Rust).
• PolymarketInstrumentProvider: Instrument parsing and loading functionality for BinaryOption instruments.
• PolymarketDataClient: A market data feed manager.
• PolymarketExecutionClient: A trade execution gateway.
• PolymarketLiveDataClientFactory: Factory for Polymarket data clients (used by the trading node builder).
• PolymarketLiveExecClientFactory: Factory for Polymarket execution clients (used by the trading node builder).

:::note Most users will define a configuration for a live trading node (as below), and won't need to work with these lower-level components directly. :::

Python and Rust implementations

The current docs cover both the Python adapter and the Rust-native adapter surface. The table below shows the main differences that affect behavior today.

pUSD

pUSD is the collateral token used for trading on Polymarket. It is a standard ERC-20 token on Polygon, backed by USDC.

The proxy contract address is 0xC011a7E12a19f7B1f670d46F03B03f3342E82DFB on Polygon. Direct on-chain funding wraps Polygon USDC.e (bridged USDC) into pUSD through the CollateralOnramp. The Bridge API can also deposit supported assets from other chains and credit pUSD after conversion.

Wallets and accounts

To interact with Polymarket via NautilusTrader, you'll need a Polygon-compatible wallet (such as MetaMask).

Signature types

Polymarket supports multiple signature types for order signing and verification:

:::note See also: Proxy wallet in the Polymarket documentation for more details about signature types and proxy wallet infrastructure. :::

NautilusTrader defaults to signature type 0 (EOA) but can be configured to use any of the supported signature types via the signature_type configuration parameter.

A single wallet address is supported per trader instance when using environment variables, or multiple wallets could be configured with multiple PolymarketExecutionClient instances.

:::note Ensure your wallet is funded with pUSD, otherwise you will encounter the "not enough balance or allowance" API error when submitting orders. :::

Setting allowances for Polymarket contracts

Before you can start trading, you need to ensure that your wallet has allowances set for Polymarket's smart contracts. You can do this by running the provided script located at nautilus_trader/adapters/polymarket/scripts/set_allowances.py.

This script is adapted from a gist created by @poly-rodr.

:::note You only need to run this script once per EOA wallet that you intend to use for trading on Polymarket. :::

This script automates the process of approving the necessary allowances for the Polymarket contracts. It sets approvals for the pUSD collateral token and Conditional Token Framework (CTF) contract to allow the Polymarket CLOB Exchange to interact with your funds.

Before running the script, ensure the following prerequisites are met:

• Install the web3 Python package: uv pip install "web3==7.12.1".
• Have a Polygon-compatible wallet funded with some POL (used for gas fees).
• Set the following environment variables in your shell:
  • POLYGON_PRIVATE_KEY: Your private key for the Polygon-compatible wallet.
  • POLYGON_PUBLIC_KEY: Your public key for the Polygon-compatible wallet.

Once you have these in place, the script will:

• Approve the maximum possible amount of pUSD (using the MAX_INT value) for the Polymarket collateral token contract.
• Set the approval for the CTF contract, allowing it to interact with your account for trading purposes.

:::note You can also adjust the approval amount in the script instead of using MAX_INT, with the amount specified in fractional units of pUSD, though this has not been tested. :::

Ensure that your private key and public key are correctly stored in the environment variables before running the script. Here's an example of how to set the variables in your terminal session:

export POLYGON_PRIVATE_KEY="YOUR_PRIVATE_KEY"
export POLYGON_PUBLIC_KEY="YOUR_PUBLIC_KEY"

Run the script using:

python nautilus_trader/adapters/polymarket/scripts/set_allowances.py

Script breakdown

The script performs the following actions:

• Connects to the Polygon network via an RPC URL (https://polygon-rpc.com/).
• Signs and sends a transaction to approve the maximum pUSD allowance for Polymarket contracts.
• Sets approval for the CTF contract to manage Conditional Tokens on your behalf.
• Repeats the approval process for specific addresses like the Polymarket CLOB Exchange and Neg Risk adapter.

This allows Polymarket to interact with your funds when executing trades and ensures smooth integration with the CLOB Exchange.

API keys

To trade with Polymarket, you'll need to generate API credentials. Follow these steps:

1. Ensure the following environment variables are set:

   • POLYMARKET_PK: Your private key for signing transactions.
   • POLYMARKET_FUNDER: The wallet address (public key) on the Polygon network used for funding trades on Polymarket.

2. Run the script using:

   bashCopy

   python nautilus_trader/adapters/polymarket/scripts/create_api_key.py
   

The script will generate and print API credentials, which you should save to the following environment variables:

• POLYMARKET_API_KEY
• POLYMARKET_API_SECRET
• POLYMARKET_PASSPHRASE

These can then be used for Polymarket client configurations:

• PolymarketDataClientConfig
• PolymarketExecClientConfig

Configuration

When setting up NautilusTrader to work with Polymarket, it’s crucial to properly configure the necessary parameters, particularly the private key.

Key parameters:

• private_key: The private key for your wallet used to sign orders. The interpretation depends on your signature_type configuration. If not explicitly provided in the configuration, it will automatically source the POLYMARKET_PK environment variable.
• funder: The pUSD funding wallet address used for funding trades. If not provided, will source the POLYMARKET_FUNDER environment variable.
• API credentials: You will need to provide the following API credentials to interact with the Polymarket CLOB:
  • api_key: If not provided, will source the POLYMARKET_API_KEY environment variable.
  • api_secret: If not provided, will source the POLYMARKET_API_SECRET environment variable.
  • passphrase: If not provided, will source the POLYMARKET_PASSPHRASE environment variable.
• auto_load_missing_instruments (default True): Controls whether subscribe and request commands for an instrument that is not already in the cache trigger an ad-hoc load via the Gamma API. When disabled, subscribing to an uncached instrument returns an error. See Runtime instrument loading.
• auto_load_debounce_ms (default 100): The window (milliseconds) over which concurrent auto-load requests are coalesced into a single batched Gamma call.

:::tip We recommend using environment variables to manage your credentials. :::

Orders capability

Polymarket operates as a prediction market with a more limited set of order types and instructions compared to traditional exchanges.

Order types

Quantity semantics

Polymarket interprets order quantities differently depending on the order type and side:

• Limit orders interpret quantity as the number of conditional tokens (base units).
• Market SELL orders also use base-unit quantities.
• Market BUY orders interpret quantity as quote notional in pUSD.

As a result, a market buy order submitted with a base-denominated quantity will execute far more size than intended.

When submitting market BUY orders, set quote_quantity=True on the order. The Python SDK or Rust adapter converts the quote amount (pUSD) to the signed base-unit share amount before posting to the CLOB. The Polymarket execution client denies base-denominated market buys to prevent unintended fills.

# Market BUY with quote quantity (spend $10 pUSD)
order = strategy.order_factory.market(
    instrument_id=instrument_id,
    order_side=OrderSide.BUY,
    quantity=instrument.make_qty(10.0),
    time_in_force=TimeInForce.IOC,  # Maps to Polymarket FAK
    quote_quantity=True,  # Interpret as pUSD notional
)
strategy.submit_order(order)

Execution instructions

Time-in-force options

Polymarket calls the POST /order field orderType. In NautilusTrader, this maps to TimeInForce. The valid combinations depend on the Nautilus order type:

:::note Polymarket uses FAK (Fill-And-Kill) for the semantics NautilusTrader calls IOC (Immediate or Cancel). Polymarket docs classify FOK and FAK as market order types, while GTC and GTD are limit order types. For Nautilus MARKET orders, both adapters accept only IOC and FOK; GTC and GTD are valid for resting LIMIT orders only. :::

Advanced order features

Batch operations

Batch submit

SubmitOrderList commands are routed to Polymarket's POST /orders endpoint. The endpoint accepts at most 15 orders per request (BATCH_ORDER_LIMIT); larger lists are split into sequential 15‑order chunks.

• Only LIMIT orders are batched. MARKET orders inside the list are routed to the single-order path, which signs a marketable order and submits it with FAK or FOK based on Nautilus time_in_force.
• reduce_only orders, quote_quantity orders, and post_only with market TIF (IOC or FOK) are rejected before submission.
• A single eligible order falls through to POST /order so it keeps the single‑order retry semantics; the batch path deliberately disables retry because the venue does not expose an idempotency key.
• BatchCancelOrders is dispatched to DELETE /orders in one shot.

Submit error handling

Polymarket's public documentation describes successful POST /order responses with success, orderID, status, and errorMsg, and documents API errors as structured error responses. It does not document statusless py-clob-client exceptions or transport failures as venue rejections.

The adapter rejects only when the response proves the order was not accepted, such as success=false, a documented order processing error, or another non-retryable client/API error. Transport failures, timeouts, ambiguous retry exhaustion, statusless PolyApiException, malformed responses, and server-side failures keep the order submitted.

For unknown outcomes, both adapters derive the expected Polymarket order hash from the signed EIP-712 order when possible and cache it as the VenueOrderId. Later WebSocket or reconciliation reports then attach to the local ClientOrderId instead of becoming external orders.

Quote-quantity market BUY orders still apply the signed quote-to-base quantity update on the unknown path. Cancels requested while submit outcome is unknown are deferred until the expected venue order ID is known, and fill tracking is registered under that ID.

Position management

Order querying

Contingent orders

Precision limits

Polymarket enforces different precision constraints based on tick size and orderType.

Binary Option instruments typically support up to 6 decimal places for amounts (with 0.0001 tick size), but market orders (FAK and FOK) have stricter precision requirements:

• Market order types (FAK and FOK):

  • Sell orders: maker amount limited to 2 decimal places.
  • Taker amount: limited to 4 decimal places.
  • The product size × price must not exceed 2 decimal places.

• Resting limit order types (GTC and GTD): More flexible precision based on market tick size.

Tick size precision hierarchy

:::note

• The tick size precision hierarchy is defined in the py-clob-client-v2 ROUNDING_CONFIG.
• Market order precision limits (2 decimals for the size field, plus tick-derived bounds for the computed amount) come from the same ROUNDING_CONFIG and are enforced by OrderBuilder.get_market_order_amounts before signing.
• Tick sizes can change dynamically during market conditions, particularly when markets become one-sided.

:::

Tick size change handling

When a market's tick size changes (tick_size_change WebSocket event), old book levels can be invalid on the new grid (for example 0.505 fits a 0.001 tick but not a 0.01 tick). To keep old-grid prices out of the new epoch, the adapter treats the change as a book epoch transition:

1. Publish the updated BinaryOption with the new price_increment and price_precision.
2. Drop the local order book for the instrument.
3. Mark the instrument as awaiting a fresh snapshot.
4. Drop incremental price_change book deltas until the snapshot arrives.
5. Reseed the book from the snapshot and resume normal processing.

Trade ticks and the instrument update flow through unchanged. The Rust adapter keeps emitting QuoteTick events through the gap by reading best_bid and best_ask from each price_change. The Python adapter derives quotes from the local book, so quote subscribers see the same brief gap as the deltas (typically sub-second, until the venue snapshot arrives).

Trades

Trades on Polymarket can have the following statuses:

• MATCHED: Trade has been matched and sent to the executor service by the operator. The executor service submits the trade as a transaction to the Exchange contract.
• MINED: Trade is observed to be mined into the chain, and no finality threshold is established.
• CONFIRMED: Trade has achieved strong probabilistic finality and was successful.
• RETRYING: Trade transaction has failed (revert or reorg) and is being retried/resubmitted by the operator.
• FAILED: Trade has failed and is not being retried.

Once a trade is initially matched, subsequent trade status updates will be received via the WebSocket. NautilusTrader records the initial trade details in the info field of the OrderFilled event, with additional trade events stored in the cache as JSON under a custom key to retain this information.

Trade ID derivation

Polymarket does not publish a trade ID on last_trade_price market-data events. The adapter derives a deterministic TradeId by FNV-1a hashing the asset ID, side, price, size, and timestamp (determine_trade_id in both Rust and Python). For CLOB Data API trade history the adapter composes the TradeId from a hash suffix, an asset suffix, and a per-(transaction, asset) sequence number (format {transactionHash[-24:]}-{asset[-4:]}-{seq:06d}). A single Polygon transaction can settle multiple fills sharing the same transactionHash, so the older last-36-character form collapsed those fills to a single id and downstream catalogs silently dropped duplicates. The same venue event yields the same trade ID across replays, keeping downstream dedup intact.

Fees

Polymarket uses the formula fee = C * feeRate * p * (1 - p) where C is shares traded and p is the share price. Fees peak at p = 0.50 and decrease symmetrically toward the extremes. Only takers pay fees; makers pay zero.

Fees are calculated in USDC, rounded to 5 decimal places, and applied at match time by the protocol. The smallest fee charged is 0.00001 USDC; smaller fees round to zero.

:::note For the latest rates, see Polymarket's Fees documentation. :::

Backtest fee model

For backtests, the adapter ships PolymarketFeeModel (a nautilus_trader.backtest.models.FeeModel subclass) which applies the taker fee formula above and credits passive maker fills with a rebate inferred from the market category. Polymarket pays a 20% maker rebate on Crypto markets and 25% on other fee-enabled categories (Sports, Finance, Politics, Economics, Culture, Weather, Tech, Mentions, Other), distributed daily from each market's rebate pool. Geopolitics markets are fee-free with no rebates and the model returns zero for them.

from nautilus_trader.adapters.polymarket.fee_model import PolymarketFeeModel

# Default: maker rebates enabled
fee_model = PolymarketFeeModel()

# Or for taker-only strategies
fee_model = PolymarketFeeModel(maker_rebates_enabled=False)

The model can also be configured through BacktestVenueConfig.fee_model via ImportableFeeModelConfig and PolymarketFeeModelConfig. Maker rebate share inference uses the instrument's category labels first, then falls back to the documented per-category fee rate when labels are absent.

Reconciliation

The Polymarket API returns either all active (open) orders or specific orders when queried by the Polymarket order ID (venue_order_id). The execution reconciliation procedure for Polymarket is as follows:

• Generate order reports for all instruments with active (open) orders, as reported by Polymarket.
• Generate position reports from current user positions reported by Polymarket's Data API.
• Compare these reports with Nautilus execution state.
• Generate missing orders to bring Nautilus execution state in line with positions reported by Polymarket.

Note: Polymarket does not directly provide data for orders which are no longer active. The Python adapter exposes an experimental generate_order_history_from_trades option to fill some of this gap from trade history. The Rust adapter does not expose the same option today.

:::warning An optional execution client configuration, generate_order_history_from_trades, is currently under development. It is not recommended for production use at this time. :::

Single-order recovery from trades

/data/order/{id} only returns active orders, so a Filled or Canceled order returns an empty response. To avoid the engine resolving a local ACCEPTED order as REJECTED (which discards fills that already happened at the venue), generate_order_status_report falls back to /data/trades filtered by the venue order ID. The cached order is resolved via client_order_id, falling back to the cache's venue_order_id index when only the venue ID is known. Recovery is keyed on the cached order; without one the recovery defers to the engine rather than synthesizing an external order from trade history alone:

• Cached order + recovered fills covering the cached quantity (within DUST_SNAP_THRESHOLD for CLOB cent-tick truncation): returns Filled. The engine reconciles any delta over the cached filled_qty via inferred fill.
• Cached order + recovered fills that fall short of the cached quantity by more than dust: returns Canceled with the recovered filled_qty. The engine's CANCELED branch transitions the order at the cached filled_qty, so any newly recovered fills that arrived only via REST (not WS) are not applied in this rare partial-cancel case. Closing the order is preferred over leaving it stuck open; if exact fill metadata matters in this scenario the venue trade history can be reviewed manually.
• Cached order, no trades: returns Canceled with cancel_reason="ORDER_NOT_FOUND_AT_VENUE".
• No cached order (regardless of trades): returns None; the engine's not-found-at-venue path resolves the local entry.

open_check_interval_secs is recommended for Polymarket so the engine periodically drives this recovery path for orders whose terminal WS update was missed.

Fill quantity normalization

Polymarket reports fill quantities that drift slightly from the submitted order quantity due to protocol-level rounding: the CLOB rounds matched fills to integer cent ticks (underfill) and the V2 SDK truncates takerAmount to USDC scale on market-BUY quote-quantity orders (overfill, a few microshares). Both drift sources are fixed in absolute share terms, so the adapter normalizes them with a single threshold of DUST_SNAP_THRESHOLD = 0.01 shares. Anything beyond that surfaces to the engine as a real partial fill or overfill.

FillReport.commission always reflects the venue-reported size, not the snapped quantity. The few-ulp difference is sub-microcent in pUSD.

The fill tracker is keyed by venue_order_id and registered on order accept, so fill reports for orders placed in another session pass through unchanged. DUST_SNAP_THRESHOLD is not configurable per-strategy; it lives in nautilus_polymarket::common::consts.

WebSockets

The PolymarketWebSocketClient is built on top of the high-performance Nautilus WebSocketClient base class, written in Rust.

Data

The data adapter buffers the initial market subscriptions during the connection window and then subscribes dynamically as new instruments are requested. The client manages multiple WebSocket connections internally when the subscription count grows past the configured per-connection cap.

Runtime instrument loading

Polymarket lists thousands of active markets and new markets appear throughout the day, so preloading the full universe at startup is rarely practical. The data adapter auto-loads missing instruments on demand so that strategies can subscribe to markets that are not in the cache:

• When a strategy issues subscribe_quote_ticks, subscribe_trade_ticks, subscribe_order_book_deltas, or request_instrument for an instrument that is not cached, the adapter registers the request and waits auto_load_debounce_ms (default 100 ms) so that concurrent requests coalesce.
• It then issues a single batched Gamma API call. Batches larger than the Gamma condition_ids query ceiling (about 100) are split across multiple calls and merged.
• Once the instruments are loaded, they are published to the data engine (populating the cache) and the deferred subscriptions open their WebSocket subscriptions atomically. A strategy that unsubscribes while the auto-load is in flight does not see a spurious subscription opened.

The feature is enabled by default. Disable it by setting auto_load_missing_instruments=False on PolymarketDataClientConfig. To preload a known set of markets at startup instead, supply load_ids or event_slug_builder on PolymarketInstrumentProviderConfig.

Newly-minted markets pass through a CLOB hydration window of several minutes during which Gamma reports active=true but GET /markets/{cid} returns either a 404 or a 200 with empty token_id strings. The adapter classifies both states as transient and retries the auto-load with bounded exponential backoff. Tune the cadence with auto_load_max_retries (default 12), auto_load_retry_delay_initial_secs (default 5.0), and auto_load_retry_delay_max_secs (default 15.0); the defaults cap the retry window near 3 minutes. Set auto_load_max_retries=0 to disable retry. 5-minute markets (e.g. updown crypto) can expire before the venue finishes hydrating, so budget for that or raise the cap.

Purging instruments at runtime

Polymarket auto-loads instruments on demand, so a long-running session keeps growing the cache as markets resolve, new markets appear, and strategies cycle through events. Use cache.purge_instrument to drop markets the strategy no longer tracks. The call removes the instrument record and every cache-owned map keyed by it (order book, quotes, trades, bars).

class PolymarketHousekeeping(Strategy):
    def on_position_closed(self, event: PositionClosed) -> None:
        # Drop the market once the position is closed and you have no further interest.
        instrument_id = event.instrument_id
        self.unsubscribe_quote_ticks(instrument_id)
        self.unsubscribe_order_book_deltas(instrument_id)
        self.cache.purge_instrument(instrument_id)

Common triggers on Polymarket:

• A market resolves and produces no further trades.
• An event ends and the strategy rotates off its markets.
• The strategy rotates a fixed-size watchlist and drops the oldest entry.

The purge skips any instrument that still has non-terminal orders (initialized, submitted, accepted, emulated, released, or inflight) or non-closed positions, so it is safe to call without coordinating with the execution client. Active WebSocket subscriptions belong to the data engine. Unsubscribe before purging if you no longer want updates.

The cache also exposes purge_order, purge_position, purge_closed_orders, purge_closed_positions, and purge_account_events for trimming closed execution state. For long-running Polymarket nodes, schedule the bulk purges from LiveExecEngineConfig (15 min interval, 60 min buffer is a sensible default). See Cache: purging cached data for the full set.

:::warning The caller decides when an instrument is no longer needed. Purging an instrument that another actor, strategy, or engine still relies on causes missing instrument lookups and loses market-data history. :::

Execution

The execution adapter keeps a user channel connection for order and trade events and manages market subscriptions as needed for instruments seen during trading.

Both the Python and Rust adapters support dynamic WebSocket subscribe and unsubscribe operations.

Subscription limits

Polymarket enforces a maximum of 500 instruments per WebSocket connection (undocumented limitation).

When you attempt to subscribe to 501 or more instruments on a single WebSocket connection:

• You will not receive the initial order book snapshot for each instrument.
• You will only receive subsequent order book updates.

NautilusTrader automatically manages WebSocket connections to handle this limitation:

• The adapter defaults to 200 instrument subscriptions per connection (configurable via ws_max_subscriptions_per_connection).
• When the subscription count exceeds this limit, additional WebSocket connections are created automatically.
• This ensures you receive complete order book data (including initial snapshots) for all subscribed instruments.

:::tip If you need to subscribe to a large number of instruments (e.g., 5000+), the adapter will automatically distribute these subscriptions across multiple WebSocket connections. You can tune the per-connection limit up to 500 via ws_max_subscriptions_per_connection. :::

Rate limiting

Polymarket enforces rate limits via Cloudflare throttling. When limits are exceeded, requests are throttled on sliding windows. Sustained overshoot can still surface as HTTP 429 responses or temporary blocking.

REST limits

Polymarket changes these quotas over time. As of 2026-05-06, the official limits are:

WebSocket limits

The WebSocket quotas are not part of the published REST rate-limits table. The adapter ships a configurable per-connection subscription cap (ws_max_subscriptions_per_connection) defaulting to 200; Polymarket previously documented an upper bound of 500 per connection.

:::warning Exceeding Polymarket rate limits triggers Cloudflare throttling. Requests are queued using sliding windows rather than rejected immediately, but sustained overshoot can result in HTTP 429 responses or temporary blocking. :::

Data loader rate limiting

The PolymarketDataLoader includes built-in rate limiting when using the default HTTP client. Requests are automatically throttled to 100 requests per minute by default. That is a NautilusTrader default, not Polymarket's current published limit. The current Rust HTTP clients also ship with conservative 100 requests per minute quotas.

When fetching large date ranges across multiple markets:

• Multiple loaders sharing the same http_client instance will coordinate rate limiting automatically.
• For higher throughput, pass a custom http_client with adjusted quotas.
• The loader does not implement automatic retry on 429 errors, so implement backoff if needed.

:::info For the latest rate limit details, see the official Polymarket documentation: https://docs.polymarket.com/api-reference/rate-limits :::

Limitations and considerations

The following limitations are currently known:

• Python order signing via py-clob-client-v2 is slow and can take around one second per order.
• Reduce-only orders are not supported.
• Batch submit (POST /orders) accepts at most 15 orders per request; the adapter splits larger SubmitOrderList commands into sequential 15-order chunks.

Configuration

The Python adapter (nautilus_trader.adapters.polymarket) and the Rust-native adapter (nautilus_trader.polymarket) expose different config surfaces. The tables below document both adapters in full.

Data client options (Python v2)

Class: PolymarketDataClientConfig in nautilus_trader.adapters.polymarket.config.

Execution client options (Python v2)

Class: PolymarketExecClientConfig in nautilus_trader.adapters.polymarket.config.

Data client options (Rust v2)

Struct: PolymarketDataClientConfig in crates/adapters/polymarket/src/config.rs.

The Rust data client config does not accept account credentials; authentication is handled by the execution client. Subscription buffering (ws_connection_initial_delay_secs) and quote handling (compute_effective_deltas, drop_quotes_missing_side) are Python-only today.

Execution client options (Rust v2)

Struct: PolymarketExecClientConfig in crates/adapters/polymarket/src/config.rs.

The Rust execution client does not expose generate_order_history_from_trades, log_raw_ws_messages, ws_max_subscriptions_per_connection, or instrument_config. Batch submissions via POST /orders deliberately skip retry regardless of max_retries; the single-order path still retries on transient failures.

Instrument provider configuration options

The instrument provider config is passed via the instrument_config parameter on the data client config.

Event slug builder

The event_slug_builder feature enables efficient loading of niche markets without downloading the full venue catalogue. Instead of loading everything, you provide a function that returns event slugs for the specific markets you need.

from nautilus_trader.adapters.polymarket.providers import PolymarketInstrumentProviderConfig

# Configure with a slug builder function
instrument_config = PolymarketInstrumentProviderConfig(
    event_slug_builder="myproject.slugs:build_temperature_slugs",
)

The callable must have signature () -> list[str] and return a list of event slugs:

# myproject/slugs.py
from datetime import UTC, datetime, timedelta

def build_temperature_slugs() -> list[str]:
    """Build slugs for NYC temperature markets."""
    slugs = []
    today = datetime.now(tz=UTC).date()

    for i in range(7):
        date = today + timedelta(days=i)
        slug = f"highest-temperature-in-nyc-on-{date.strftime('%B-%d').lower()}"
        slugs.append(slug)

    return slugs

See examples/live/polymarket/slug_builders.py for more examples including crypto UpDown markets.

Historical data loading

The PolymarketDataLoader provides methods for fetching and parsing historical market data for research and backtesting purposes. The loader integrates with multiple Polymarket APIs to provide the required data.

:::note All data fetching methods are asynchronous and must be called with await. The loader can optionally accept an http_client parameter for dependency injection (useful for testing). :::

Data sources

The loader fetches data from three primary sources:

1. Polymarket Gamma API - Market metadata, instrument details, and active market listings.
2. Polymarket CLOB API - Market details for instrument construction.
3. Polymarket Data API - Historical trades and current user positions.

The current loader does not expose helpers for CLOB price history timeseries or order book history snapshots.

Method naming conventions

The loader provides two ways to access the Polymarket APIs:

Use query_* when you want to explore markets, discover events, or fetch metadata before committing to a specific instrument:

# No loader needed: query the API directly
market = await PolymarketDataLoader.query_market_by_slug("some-market")
event = await PolymarketDataLoader.query_event_by_slug("some-event")

Use fetch_* when you have a loader instance and want to fetch data using its configured HTTP client (for coordinated rate limiting across multiple calls):

loader = await PolymarketDataLoader.from_market_slug("some-market")

# All fetch calls share the loader's HTTP client
markets = await loader.fetch_markets(active=True, limit=100)
events = await loader.fetch_events(active=True)
details = await loader.fetch_market_details(condition_id)

Finding markets

Use the provided utility scripts to discover active markets:

# List all active markets
python nautilus_trader/adapters/polymarket/scripts/active_markets.py

# List BTC and ETH UpDown markets specifically
python nautilus_trader/adapters/polymarket/scripts/list_updown_markets.py

Basic usage

The recommended way to create a loader is using the factory classmethods, which handle all the API calls and instrument creation automatically:

import asyncio

from nautilus_trader.adapters.polymarket import PolymarketDataLoader

async def main():
    # Create loader from market slug (recommended)
    loader = await PolymarketDataLoader.from_market_slug("gta-vi-released-before-june-2026")

    # Loader is ready to use with instrument and token_id set
    print(loader.instrument)
    print(loader.token_id)

asyncio.run(main())

For events with multiple markets (e.g., temperature buckets), use from_event_slug:

# Returns a list of loaders, one per market in the event
loaders = await PolymarketDataLoader.from_event_slug("highest-temperature-in-nyc-on-january-26")

Look-ahead protection for resolved markets

When constructing a loader for a market that has already resolved at backtest build time, the venue payload includes the answer (closed, closedTime, umaResolutionStatus, per-token winner). A strategy that reads cache.instrument(...).info from on_start can therefore see the outcome before the simulation runs.

Pass sanitize_info=True to either factory to redact those fields from instrument.info before the instrument is constructed. The redacted slice is stashed on the loader as resolution_metadata for post-hoc analytics (settlement PnL, Brier scoring) without leaking it into the simulation:

loader = await PolymarketDataLoader.from_market_slug(
    "some-resolved-market",
    sanitize_info=True,
)

assert "closed" not in loader.instrument.info
assert loader.resolution_metadata["closed"] is True

Discovering markets and events

Use fetch_markets() and fetch_events() to discover available markets programmatically:

loader = await PolymarketDataLoader.from_market_slug("any-market")

# List active markets
markets = await loader.fetch_markets(active=True, closed=False, limit=100)
for market in markets:
    print(f"{market['slug']}: {market['question']}")

# List active events
events = await loader.fetch_events(active=True, limit=50)
for event in events:
    print(f"{event['slug']}: {event['title']}")

# Get all markets within a specific event
event_markets = await loader.get_event_markets("highest-temperature-in-nyc-on-january-26")

For quick exploration without creating a loader, use the static query_* methods (see Method naming conventions above).

Fetching trade history

The load_trades() convenience method fetches and parses historical trades in one step:

import pandas as pd

# Load all available trades
trades = await loader.load_trades()

# Or filter by time range (client-side filtering)
end = pd.Timestamp.now(tz="UTC")
start = end - pd.Timedelta(hours=24)

trades = await loader.load_trades(
    start=start,
    end=end,
)

Alternatively, you can fetch and parse separately using the lower-level methods:

condition_id = loader.condition_id

# Fetch raw trades from the Polymarket Data API
raw_trades = await loader.fetch_trades(condition_id=condition_id)

# Parse to NautilusTrader TradeTicks
trades = loader.parse_trades(raw_trades)

Trade data is sourced from the Polymarket Data API, which provides real execution data including price, size, side, and on-chain transaction hash.

:::note The public Data API caps offset-based pagination on high-activity markets. When this ceiling is hit the loader emits a RuntimeWarning and returns the trades fetched up to the cap rather than aborting the load. Use another historical data source if you need full coverage of a heavily traded market. :::

Complete backtest example

See examples/backtest/polymarket_simple_quoter.py for a full example:

import asyncio
from decimal import Decimal

from nautilus_trader.adapters.polymarket import POLYMARKET_VENUE
from nautilus_trader.adapters.polymarket import PolymarketDataLoader
from nautilus_trader.backtest.config import BacktestEngineConfig
from nautilus_trader.backtest.engine import BacktestEngine
from nautilus_trader.examples.strategies.ema_cross_long_only import EMACrossLongOnly
from nautilus_trader.examples.strategies.ema_cross_long_only import EMACrossLongOnlyConfig
from nautilus_trader.model.currencies import pUSD
from nautilus_trader.model.data import BarType
from nautilus_trader.model.enums import AccountType
from nautilus_trader.model.enums import OmsType
from nautilus_trader.model.identifiers import TraderId
from nautilus_trader.model.objects import Money

async def run_backtest():
    # Initialize loader and fetch market data
    loader = await PolymarketDataLoader.from_market_slug("gta-vi-released-before-june-2026")
    instrument = loader.instrument

    # Load historical trades from the Polymarket Data API
    trades = await loader.load_trades()

    # Configure and run backtest
    config = BacktestEngineConfig(trader_id=TraderId("BACKTESTER-001"))
    engine = BacktestEngine(config=config)

    engine.add_venue(
        venue=POLYMARKET_VENUE,
        oms_type=OmsType.NETTING,
        account_type=AccountType.CASH,
        base_currency=pUSD,
        starting_balances=[Money(10_000, pUSD)],
    )

    engine.add_instrument(instrument)
    engine.add_data(trades)

    bar_type = BarType.from_str(f"{instrument.id}-100-TICK-LAST-INTERNAL")
    strategy_config = EMACrossLongOnlyConfig(
        instrument_id=instrument.id,
        bar_type=bar_type,
        trade_size=Decimal("20"),
    )

    strategy = EMACrossLongOnly(config=strategy_config)
    engine.add_strategy(strategy=strategy)
    engine.run()

    # Display results
    print(engine.trader.generate_account_report(POLYMARKET_VENUE))

# Run the backtest
asyncio.run(run_backtest())

Run the complete example:

python examples/backtest/polymarket_simple_quoter.py

Helper functions

The adapter provides utility functions for working with Polymarket identifiers:

from nautilus_trader.adapters.polymarket import get_polymarket_instrument_id

# Create NautilusTrader InstrumentId from Polymarket identifiers
instrument_id = get_polymarket_instrument_id(
    condition_id="0xcccb7e7613a087c132b69cbf3a02bece3fdcb824c1da54ae79acc8d4a562d902",
    token_id="8441400852834915183759801017793514978104486628517653995211751018945988243154"
)

Contributing

:::info For additional features or to contribute to the Polymarket adapter, please see our contributing guide. :::