LEAN Algorithmic Trading Engine

LEAN supports FOREX data provided by Oanda and FXCM. The data consists exclusively of Quote data.

Supported Resolutions:

• Tick
• Second
• Minute
• Hour
• Daily

Data is stored in compressed .zip files containing a single .csv file. You can explore and download this data at https://www.quantconnect.com/data/tree/forex/.

QuantConnect provides Contracts for Difference (CFD) data sourced from Oanda. This data consists exclusively of Quote data (Bid/Ask) and is stored in ZIP files containing a single CSV.

Supported Resolutions:

• Tick
• Second
• Minute
• Hour
• Daily

Supported Market: Oanda.

Note: CFD data operates across multiple timezones. Users should verify the specific timezone for their asset using the market hours database.

QuantConnect provides US Equity data (market 'USA') with various resolutions: Tick, Second, Minute, Hour, and Daily.

Key Data Characteristics

• Timezone: US equity data is in the New York timezone. Specific market hours can be found in MarketHoursDatabase.json.
• Price Format: Prices are stored in deci-cents. To convert deci-cents to dollars, divide prices by 10,000.
• Data Sources: Post-2007 data is provided by AlgoSeek (trade and quote); pre-2007 trade data is provided by QuantQuote.
• Filtering: While raw ticks are stored unfiltered, TradeBars and QuoteBars have suspicious ticks filtered out and consolidated.

LEAN futures data is provided by AlgoSeek and includes trade, quote, and openinterest data. Data is stored in compressed ZIP files containing one or more CSV files.

Supported resolutions:

• Tick
• Second
• Minute

Supported markets:

• CBOT, CME, NYMEX, COMEX, CBOE, ICE

tickType values used in file naming and schemas:

• trade
• quote
• openinterest

Python research notebooks use a setup script that automatically loads QuantBook libraries into the Python kernel.

Setup:

• Docker Environment: The script runs automatically.
• Local Environment: You must manually call the setup script in the first cell of your notebook.

Important Note on start.py: Python research is heavily dependent on the start.py script. It is responsible for assigning the core clr as the runtime for PythonNet and clr-loader. If you are running locally, you must call %run "start.py". Depending on your directory structure, you may need to use a relative path like %run "../start.py".

For a practical implementation, refer to the KitchenSinkQuantBookTemplate.ipynb reference notebook.

To use C# for research, you must load the Initialize.csx setup script into your notebook. This script loads the QuantConnect libraries into your C# Kernel.

Setup: In the first cell of your notebook, use the #load directive. In Docker environments, the file is typically one directory above the notebooks directory.

Important Note on Namespaces: The latest C# research kernel does not support global using statements outside the notebook context. You must add all required Lean namespaces directly via using statements within your notebook cells.

For a practical implementation, refer to the KitchenSinkCSharpQuantBookTemplate.ipynb reference notebook.

#load "../Initialize.csx"

using QuantConnect;
using QuantConnect.Data;
// ... other namespaces

A PositionGroupKey is a deterministic identifier for a position group. It is constructed from:

1. The Symbol and UnitQuantity of all contained positions.
2. The IPositionGroupBuyingPowerModel used by the group.

Because the UnitQuantities are stored in an ImmutableSortedSet, the order of positions does not change the key (e.g., +100 GOOG; -1 GOOG CALL is identical to -1 GOOG CALL; +100 GOOG). This key is used to index into the PositionManager and other collections. It also acts as a template, as the key contains enough information to reconstruct an IPositionGroup.

public bool IsDefaultGroup { get; }
public IPositionGroupBuyingPowerModel BuyingPowerModel { get; }
public IReadOnlyList<Tuple<Symbol, decimal>> UnitQuantities { get; }

Position Groups allow algorithms to submit orders for a logical grouping of securities as a single action. This is primarily used to manage margin requirements and risk modeling for hedged or market-neutral strategies.

By grouping securities (e.g., a covered call consisting of 100 shares of equity and 1 short option contract), the margin required for the group is often significantly lower than the sum of the individual parts. This feature enables LEAN to:

1. Accurately model reduced margin requirements for hedged positions.
2. Submit multi-leg orders so brokerages can process them as a single unit, which is often necessary when individual legs would otherwise fail margin checks.

The PositionManager is responsible for managing both individual positions and IPositionGroup collections. It functions similarly to SecurityPortfolioManager.

Key behaviors:

• Automatic Grouping: The manager ensures that all holdings are always grouped. Any holdings that do not match specific criteria are moved into the default SecurityPositionGroup (the 'group of last resort').
• Event-Driven Updates: The manager listens to SecurityHolding.QuantityChanged events. After every fill event, it invokes the configured IPositionGroupResolver to recalculate the current set of groups.
• Consistency: To ensure margin calculations and buying power models remain accurate, groups must be resolved immediately following any change in security holdings.