Skip to content

Martingale42/nautilus-dev / skills/nt-trading

nt-trading

Verified

Use when working with strategy logic, order execution, risk management, position/portfolio tracking, or exec algorithms in NautilusTrader.

nt-trading

What This Skill Covers

NautilusTrader trading domain — strategy logic, order execution, risk management, and portfolio/position tracking.

Python modules: trading/, execution/, risk/, accounting/, portfolio/ Rust crates: nautilus_trading, nautilus_execution, nautilus_risk, nautilus_portfolio

When To Use

  • Writing or modifying a Strategy or Actor
  • Configuring order execution (order types, time-in-force, exec algorithms)
  • Setting up risk management (risk engine config, margin models)
  • Working with positions, portfolio queries, or accounting
  • Building custom execution algorithms (ExecAlgorithm)
  • Order lifecycle handling (submit_order, cancel_order, modify_order)

When NOT To Use

  • Custom indicators or signal generation → use nt-signals
  • Fill models or simulated exchange → use nt-backtest
  • Adapter configuration → use nt-adapters
  • Domain model types (instruments, identifiers) → use nt-model
  • Data persistence or catalog → use nt-data
  • Live node system setup → use nt-live

Python Usage

Strategy

Subclass Strategy from nautilus_trader.trading.strategy:

from nautilus_trader.trading.strategy import Strategy
from nautilus_trader.config import StrategyConfig

class MyStrategyConfig(StrategyConfig, frozen=True):
    instrument_id: str = None
    bar_type: str = None

class MyStrategy(Strategy):
    def __init__(self, config: MyStrategyConfig):
        super().__init__(config)
        # Store config, initialize state

    def on_start(self):
        # Subscribe to data, register indicators
        self.subscribe_bars(self.bar_type)

    def on_bar(self, bar):
        # Core trading logic
        pass

    def on_quote_tick(self, tick):
        pass

    def on_trade_tick(self, tick):
        pass

    def on_order_filled(self, event):
        # Post-fill logic
        pass

    def on_position_changed(self, event):
        pass

    def on_stop(self):
        # Cleanup
        pass

Key order methods (inherited from Actor → Strategy):

  • self.submit_order(order) — submit to execution
  • self.cancel_order(order) — cancel open order
  • self.modify_order(order, quantity=None, price=None, trigger_price=None) — modify open order
  • self.cancel_all_orders(instrument_id) — cancel all for instrument
  • self.close_position(position) — close position with market order
  • self.close_all_positions(instrument_id) — close all for instrument

Order creation (via OrderFactory on self.order_factory):

  • self.order_factory.market(instrument_id, side, quantity)
  • self.order_factory.limit(instrument_id, side, quantity, price)
  • self.order_factory.stop_market(instrument_id, side, quantity, trigger_price)
  • self.order_factory.stop_limit(instrument_id, side, quantity, price, trigger_price)
  • self.order_factory.trailing_stop_market(instrument_id, side, quantity, trailing_offset, ...)

Actor

Subclass Actor from nautilus_trader.trading.actor for non-trading components (data processing, signal publishing, monitoring):

from nautilus_trader.trading.actor import Actor
from nautilus_trader.config import ActorConfig

class MyActor(Actor):
    def __init__(self, config: ActorConfig):
        super().__init__(config)

    def on_start(self):
        self.subscribe_data(...)

    def on_bar(self, bar):
        # Process data, publish signals via msgbus
        self.publish_signal(name="my_signal", value=signal_value)

Risk & Execution Configuration

from nautilus_trader.config import ExecEngineConfig, RiskEngineConfig

exec_config = ExecEngineConfig(
    load_cache=True,
    allow_cash_positions=True,
)

risk_config = RiskEngineConfig(
    bypass=False,
    max_order_submit_rate="100/00:00:01",  # 100 per second
    max_order_modify_rate="100/00:00:01",
    max_notional_per_order={"GBP/USD.SIM": 1_000_000},
)

Python Extension

Custom ExecAlgorithm

Subclass ExecAlgorithm from nautilus_trader.execution.algorithm:

from nautilus_trader.execution.algorithm import ExecAlgorithm

class MyExecAlgorithm(ExecAlgorithm):
    def on_start(self):
        pass

    def on_order(self, order):
        # Custom execution logic — split, time-slice, etc.
        self.submit_order(order)

    def on_stop(self):
        pass

Register in config:

exec_algorithms=[MyExecAlgorithm.fully_qualified_name()]

Custom Margin/Position Sizing

Extend risk calculations by subclassing margin models or implementing custom position sizing logic in your Strategy.

Rust Usage

NautilusTrader has a complete Rust implementation (v2 Rust) that runs without Python. Strategies and actors are written in pure Rust, compiled to standalone binaries.

Rust Strategy

A strategy owns a StrategyCore and implements DataActor for data handling. The nautilus_strategy! macro wires up Deref/DerefMut and the Strategy trait.

use nautilus_common::actor::DataActor;
use nautilus_model::{
    data::QuoteTick,
    enums::OrderSide,
    identifiers::{InstrumentId, StrategyId},
    types::Quantity,
};
use nautilus_trading::{nautilus_strategy, strategy::{Strategy, StrategyConfig, StrategyCore}};

pub struct MyStrategy {
    core: StrategyCore,
    instrument_id: InstrumentId,
    trade_size: Quantity,
}

impl MyStrategy {
    pub fn new(instrument_id: InstrumentId) -> Self {
        let config = StrategyConfig {
            strategy_id: Some(StrategyId::from("MY_STRAT-001")),
            order_id_tag: Some("001".to_string()),
            ..Default::default()
        };
        Self {
            core: StrategyCore::new(config),
            instrument_id,
            trade_size: Quantity::from("1.0"),
        }
    }
}

// Generates Deref<Target = DataActorCore>, DerefMut, and Strategy trait impl
nautilus_strategy!(MyStrategy);

impl std::fmt::Debug for MyStrategy {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        f.debug_struct("MyStrategy").finish()
    }
}

impl DataActor for MyStrategy {
    fn on_start(&mut self) -> anyhow::Result<()> {
        self.subscribe_quotes(self.instrument_id, None, None);
        Ok(())
    }

    fn on_quote(&mut self, quote: &QuoteTick) -> anyhow::Result<()> {
        let order = self.core.order_factory().market(
            self.instrument_id,
            OrderSide::Buy,
            self.trade_size,
            None, None, None, None, None, None, None,
        );
        self.submit_order(order, None, None)?;
        Ok(())
    }
}

Override Strategy hooks (order/position event handlers) by passing a block to the macro. The block can override multiple handlers — keep heavy logic in methods on the impl block, not inside the macro:

use nautilus_model::events::{OrderRejected, OrderFilled, OrderCanceled};

nautilus_strategy!(MyStrategy, {
    fn on_order_rejected(&mut self, event: OrderRejected) {
        let reason = event.reason.to_string();
        if reason.contains("POST_ONLY") {
            log::info!("post-only reject (expected): {reason}");
        } else {
            log::warn!("order rejected: {reason}");
        }
    }

    fn on_order_filled(&mut self, event: OrderFilled) {
        if let Err(e) = self.handle_fill(&event) {
            log::error!("handle_fill failed: {e}");
        }
    }

    fn on_order_canceled(&mut self, event: OrderCanceled) {
        // ...
    }
});

Inside the block: &mut self and the event by value. Do not redefine core() / core_mut() — the macro already generates them.

Warmup pattern (request_quotes + on_historical_*)

When a strategy needs historical data before quoting (e.g. for indicator seeding or calibration), use the request/historical-handler pair:

fn on_start(&mut self) -> anyhow::Result<()> {
    self.subscribe_quotes(self.cfg.instrument_id, None, None);
    self.subscribe_trades(self.cfg.instrument_id, None, None);

    let now_ns: u64 = self.core.clock().timestamp_ns().into();
    let start = UnixNanos::from(now_ns - 24 * 3_600_000_000_000); // 24h
    let end = UnixNanos::from(now_ns);

    self.warmup_quotes = Some(Vec::new());
    self.warmup_trades = Some(Vec::new());
    self.request_quotes(self.cfg.instrument_id, Some(start), Some(end), None, None, None)?;
    self.request_trades(self.cfg.instrument_id, Some(start), Some(end), None, None, None)?;
    Ok(())
}

fn on_historical_quotes(&mut self, quotes: &[QuoteTick]) -> anyhow::Result<()> {
    if let Some(buf) = self.warmup_quotes.as_mut() { buf.extend_from_slice(quotes); }
    self.maybe_run_warmup();
    Ok(())
}

fn on_historical_trades(&mut self, trades: &[TradeTick]) -> anyhow::Result<()> {
    if let Some(buf) = self.warmup_trades.as_mut() { buf.extend_from_slice(trades); }
    self.maybe_run_warmup();
    Ok(())
}

maybe_run_warmup() runs the seeding logic once both buffers have arrived. Backtest gotcha: if the run starts at the catalog's earliest data, request_* returns nothing — keep a rolling buffer in on_quote / on_trade as a fallback.

Daily timer for re-calibration

Use clock().set_timer_ns() to schedule a repeating timer at UTC midnight (or any anchor). It fires deterministically in backtest:

pub const DAILY_TIMER: &str = "my_daily_timer";

fn on_start(&mut self) -> anyhow::Result<()> {
    let now_ns: u64 = self.core.clock().timestamp_ns().into();
    let next_midnight = next_utc_midnight_ns(now_ns);
    self.core.clock_mut().set_timer_ns(
        DAILY_TIMER,
        24 * 3_600_000_000_000,                  // interval
        UnixNanos::from(next_midnight),          // start
        None, None,                              // stop_time, callback
    )?;
    Ok(())
}

// Override on_time_event in DataActor impl, or in nautilus_strategy! block.
fn on_time_event(&mut self, event: &TimeEvent) -> anyhow::Result<()> {
    if event.name.as_str() == DAILY_TIMER {
        // run daily work
    }
    Ok(())
}

Instrument resolution

Hardcoded tick_size / size_precision is a bug magnet — every instrument differs. Resolve from the cache in on_start:

fn on_start(&mut self) -> anyhow::Result<()> {
    let inst = self.cache().instrument(&self.cfg.instrument_id)
        .ok_or_else(|| anyhow::anyhow!("instrument not in cache"))?;
    self.tick_size = inst.price_increment().as_f64();
    self.instrument = Some(inst.clone());
    Ok(())
}

In backtest: add_instrument(...) populates the cache. In live: the adapter's InstrumentProvider::initialize() does it. Either way, on_start is the right hook — it runs after both.

Order creation (via self.core.order_factory()):

  • market(instrument_id, side, quantity, ...)
  • limit(instrument_id, side, quantity, price, ...)
  • stop_market(instrument_id, side, quantity, trigger_price, ...)
  • stop_limit(instrument_id, side, quantity, price, trigger_price, ...)
  • market_if_touched(...), limit_if_touched(...), trailing_stop_market(...)

Order management (via Strategy trait, available on self):

MethodAction
submit_orderSubmit a new order to the venue
submit_order_listSubmit a list of contingent orders
modify_orderModify price, quantity, or trigger price
cancel_orderCancel a specific order
cancel_ordersCancel a filtered set of orders
cancel_all_ordersCancel all orders for an instrument
close_positionClose a position with a market order
close_all_positionsClose all open positions

Rust Actor

An actor owns a DataActorCore and receives data/events without order management. The nautilus_actor! macro wires up Deref/DerefMut.

use nautilus_common::{nautilus_actor, actor::{DataActor, DataActorConfig, DataActorCore}};
use nautilus_model::{data::QuoteTick, identifiers::{ActorId, InstrumentId}};

pub struct SpreadMonitor {
    core: DataActorCore,
    instrument_id: InstrumentId,
}

impl SpreadMonitor {
    pub fn new(instrument_id: InstrumentId) -> Self {
        let config = DataActorConfig {
            actor_id: Some(ActorId::from("SPREAD_MON-001")),
            ..Default::default()
        };
        Self {
            core: DataActorCore::new(config),
            instrument_id,
        }
    }
}

nautilus_actor!(SpreadMonitor);

impl std::fmt::Debug for SpreadMonitor {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        f.debug_struct("SpreadMonitor").finish()
    }
}

impl DataActor for SpreadMonitor {
    fn on_start(&mut self) -> anyhow::Result<()> {
        self.subscribe_quotes(self.instrument_id, None, None);
        Ok(())
    }

    fn on_quote(&mut self, quote: &QuoteTick) -> anyhow::Result<()> {
        let spread = quote.ask_price.as_f64() - quote.bid_price.as_f64();
        log::info!("Spread: {spread:.5}");
        Ok(())
    }
}

DataActor Handler Table

All handlers have default no-op implementations. Override only what you need:

HandlerReceives
on_startActor started
on_stopActor stopped
on_quoteQuoteTick
on_tradeTradeTick
on_barBar
on_book_deltasOrderBookDeltas
on_bookOrderBook (at interval)
on_instrumentInstrumentAny
on_mark_priceMarkPriceUpdate
on_index_priceIndexPriceUpdate
on_funding_rateFundingRateUpdate
on_option_greeksOptionGreeks
on_option_chainOptionChainSlice
on_instrument_statusInstrumentStatus
on_order_filledOrderFilled
on_order_canceledOrderCanceled
on_time_eventTimeEvent

Running Rust Components

Three paths to run Rust strategies/actors:

  1. Pure Rust — standalone binary, no Python runtime:

    let strategy = MyStrategy::new(instrument_id);
    node.add_strategy(strategy)?;
    node.run().await?;
    
  2. Native config from Python — register built-in Rust strategies from Python:

    from nautilus_trader.core.nautilus_pyo3.trading import GridMarketMakerConfig
    config = GridMarketMakerConfig(instrument_id=..., trade_size=...)
    node.add_native_strategy(config)
    
  3. Plugin loading (planned) — load compiled cdylib crates at runtime.

Guard Safety

When accessing other actors in callbacks:

  • Look up actors by ID each time; do not cache an ActorRef
  • Drop the guard before scope ends; never store in a field
  • Never hold a guard across an .await point

Rust Extension (PyO3 Path)

Exposing Rust Strategies to Python

Use #[pyclass] configs with add_native_strategy/add_native_actor dispatch:

ConfigStrategy
EmaCrossConfigEmaCross
GridMarketMakerConfigGridMarketMaker
DeltaNeutralVolConfigDeltaNeutralVol
ConfigActor
BookImbalanceActorConfigBookImbalanceActor

To add custom components to this path: add a #[pyclass] config and dispatch arm in add_native_strategy or add_native_actor.

PyO3 Binding Conventions

  • Use #[pyclass] and #[pymethods] for Python-visible types
  • Register new modules in crates/pyo3/src/lib.rs
  • Use Python::attach(|py| { ... }) for callbacks that need the GIL
  • Wrap FFI functions in abort_on_panic(|| { ... }) — Rust panics must never unwind across FFI

Key Conventions

Order Lifecycle

Orders follow a strict state machine: INITIALIZED → SUBMITTED → ACCEPTED → [PARTIALLY_FILLED] → FILLED | CANCELED | EXPIRED | REJECTED. Never assume order state — always check via events.

Position State

Positions transition: OPENING → OPEN → CLOSING → CLOSED. A position's side is determined by its net quantity. Hedge mode vs netting mode affects how fills create/modify positions.

Live-Readiness Checklist

  • Handle all order events (not just fills — also rejects, cancels, expirations)
  • Implement on_stop() cleanup (cancel open orders, optionally close positions)
  • Use self.clock.timer_names for timer management
  • Check self.portfolio.is_net_long() / is_net_short() for position awareness
  • Log state transitions at appropriate levels

Testing

  • Use BacktestEngine for strategy unit tests
  • Verify order submission counts, fill events, position states
  • Test edge cases: partial fills, rejects, disconnections
  • See references/guides/testing.md for patterns

References

  • references/concepts/ — strategies, actors, execution, orders, positions, portfolio, rust
  • references/api/ — trading, execution, risk, portfolio, accounting, orders, position, events
  • references/architecture_patterns.md — Cross-skill production reference. Full strategy lifecycle (ingest → backtest → live), three-binary layout, onion architecture, atomic state, fail-fast philosophy. Load this first for any non-trivial Rust strategy work.
  • references/guides/ — testing patterns, write_rust_strategy, write_rust_actor, rust_patterns (architectural patterns from production Rust strategies), troubleshooting_rust (common Rust failure modes + diagnostic workflow)
  • references/examples/ — backtest examples (EMA cross, actor data/signals, msgbus), Rust strategies (ema_cross, grid_mm), Rust actors (imbalance)
  • templates/ — strategy.py, actor.py, exec_algorithm.py

When to load which guide

  • Architectural decisions (new strategy, refactor, lifecycle question) → architecture_patterns.md first.
  • Building a new Rust strategy → architecture_patterns.md §10 build sequence, then write_rust_strategy.md, then rust_patterns.md for production patterns.
  • Build/runtime errors, unexpected behavior → troubleshooting_rust.md (read the diagnostic workflow before guessing).
  • Need to handle order events (rejects, fills, cancels) → rust_patterns.md §5 (override block).
  • Need historical warmup or daily re-calibration → rust_patterns.md §2 and §3.
  • Splitting a strategy into ingest / backtest / live binaries → rust_patterns.md §6 or architecture_patterns.md §1.
  • Cross-skill troubleshooting (which skill / file owns this symptom) → architecture_patterns.md §8.
slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack."Stars: 197kUpdated Claude CodeCodexVerifiedInstall
ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt.Stars: 194kUpdated Claude CodeCodexVerifiedInstall