Writing Algorithms

Initialization

Introduction

The Initializeinitialize method is the entry point of your algorithm where you define a series of settings, including security subscriptions, starting cash balances, and warm-up periods. LEAN only calls the Initializeinitialize method one time, at the start of your algorithm.

Set Dates

To set the date range of backtests, call the SetStartDateset_start_date and SetEndDateset_end_date methods. The dates you provide are based in the algorithm time zone. By default, the end date is yesterday, one millisecond before midnight. In live trading, LEAN ignores the start and end dates.

// Set the start and end dates for the algorithm, defining the backtesting period.
SetStartDate(2013, 1, 5);                  // Set the start date to January 5, 2013.
SetEndDate(2015, 1, 5);                    // Set the end date to January 5, 2015.
SetEndDate(DateTime.Now.Date.AddDays(-7)); // Set the end date to last week.
# Set the start and end dates for the algorithm, defining the backtesting period.
self.set_start_date(2013, 1, 5)                  # Set the start date to January 5, 2013.
self.set_end_date(2015, 1, 5)                    # Set the end date to January 5, 2015.
self.set_end_date(datetime.now() - timedelta(7)) # Set the end date to last week.

Set Account Currency

The algorithm equity curve, benchmark, and performance statistics are denominated in the account currency. To set the account currency and your starting cash, call the SetAccountCurrencyset_account_currency method. By default, the account currency is USD and your starting cash is $100,000. If you call the SetAccountCurrencyset_account_currency method, you must call it before you call the SetCashset_cash method or add data. If you call the SetAccountCurrencyset_account_currency method more than once, only the first call takes effect.

// Set the account currency and your starting cash.
SetAccountCurrency("BTC"); // Set the account currency to Bitcoin and its quantity to 100,000 BTC.
SetAccountCurrency("INR"); // Set the account currency to Indian Rupees and its quantity to 100,000 INR.
SetAccountCurrency("BTC", 10);  // Set the account currency to Bitcoin and its quantity to 10 BTC.
# Set the account currency and your starting cash.
self.set_account_currency("BTC") # Set the account currency to Bitcoin and its quantity to 100,000 BTC.
self.set_account_currency("INR") # Set the account currency to Indian Rupees and its quantity to 100,000 INR.
self.set_account_currency("BTC", 10) # Set the account currency to Bitcoin and its quantity to 10 BTC.

Set Cash

To set your starting cash in backtests, call the SetCashset_cash method. By default, your starting cash is $100,000 USD. In live trading, LEAN ignores the SetCashset_cash method and uses the cash balances in your brokerage account instead.

// Set the initial cash balance for the account for supported currencies.
SetCash(100000);       // Set the quantity of the account currency to 100,000.
SetCash("BTC", 10);    // Set the Bitcoin quantity to 10.
SetCash("EUR", 10000); // Set the EUR quantity to 10,000.
# Set the initial cash balance for the account for supported currencies.
self.set_cash(100000)       # Set the quantity of the account currency to 100,000.
self.set_cash("BTC", 10)    # Set the Bitcoin quantity to 10.
self.set_cash("EUR", 10000) # Set the EUR quantity to 10,000.

Set Brokerage and Cash Model

We model your algorithm with margin modeling by default, but you can select a cash account type. Cash accounts don't allow leveraged trading, whereas Margin accounts can support leverage on your account value. To set your brokerage and account type, call the SetBrokerageModelset_brokerage_model method. For more information about each brokerage and the account types they support, see the brokerage integration documentation. For more information about the reality models that the brokerage models set, see Supported Models.

// Set the brokerage model and account type to model fees, supported order types, and available margin. 
SetBrokerageModel(BrokerageName.InteractiveBrokersBrokerage, AccountType.Margin);
# Set the brokerage model and account type to model fees, supported order types, and available margin.
self.set_brokerage_model(BrokerageName.INTERACTIVE_BROKERS_BROKERAGE, AccountType.MARGIN)

The AccountType enumeration has the following members:

Set Universe Settings

The universe settings of your algorithm configure some properties of the universe constituents. The following table describes the properties of the UniverseSettings object:

Property: Asynchronousasynchronous

Should the universe selection run asynchronously to boost speed?

Data Type: bool | Default Value: falseFalse

Property: ExtendedMarketHoursextended_market_hours

Should assets also feed extended market hours? You only receive extended market hours data if you create the subscription with an intraday resolution. If you create the subscription with daily resolution, the daily bars only reflect the regular trading hours.

Data Type: bool | Default Value: falseFalse

Property: FillForwardfill_forward

Should asset data fill forward?

Data Type: bool | Default Value: trueTrue

Property: MinimumTimeInUniverseminimum_time_in_universe

What's the minimum time assets should be in the universe?

Data Type: TimeSpantimedelta | Default Value: TimeSpan.FromDays(1)timedelta(1)

Property: Resolutionresolution

What resolution should assets use?

Data Type: Resolution | Default Value: Resolution.MinuteResolution.MINUTE

Method: Schedule.Onschedule.on

What days should universe selection run? If it's nullNone, it's the data resolution. For example, US Equity fundamentals universes are daily at midnight.

Argument Type: IDateRuleIDateRule | Default Value: nullNone

Property: ContractDepthOffsetcontract_depth_offset

What offset from the current front month should be used for continuous Future contracts? 0 uses the front month and 1 uses the back month contract. This setting is only available for Future assets.

Data Type: int | Default Value: 0

Property: DataMappingModedata_mapping_mode

How should continuous Future contracts be mapped? This setting is only available for Future assets.

Data Type: DataMappingMode | Default Value: DataMappingMode.OpenInterest

Property: DataNormalizationModedata_normalization_mode

How should historical prices be adjusted? This setting is only available for Equity and Futures assets.

Data Type: DataNormalizationMode | Default Value: DataNormalizationMode.AdjustedDataNormalizationMode.ADJUSTED

Property: Leverageleverage

What leverage should assets use in the universe? This setting is not available for derivative assets.

Data Type: decimalfloat | Default Value: Security.NullLeverageSecurity.NULL_LEVERAGE

To set the UniverseSettingsuniverse_settings, update the preceding properties in the Initializeinitialize method before you add the universe. These settings are globals, so they apply to all universes you create.

// Request second resolution data. This will be slow!
UniverseSettings.Resolution = Resolution.Second;
AddUniverse(Universe.DollarVolume.Top(50));
# Request second resolution data. This will be slow!
self.universe_settings.resolution = Resolution.SECOND
self.add_universe(self.universe.dollar_volume.top(50))

For more information about universe settings, see the related documentation for classic and framework algorithms.

Set Security Initializer

Instead of configuring global universe settings, you can individually configure the settings of each security in the universe with a security initializer. Security initializers let you apply any security-level reality model or special data requests on a per-security basis. To set the security initializer, in the Initializeinitialize method, call the SetSecurityInitializerset_security_initializer method and then define the security initializer.

// A custom security initializer can override default models such as 
// setting new fee and fill models for the security.
SetSecurityInitializer(CustomSecurityInitializer);

private void CustomSecurityInitializer(Security security)
{
    security.SetFeeModel(new ConstantFeeModel(0, "USD"));
}
# A custom security initializer can override default models such as
# setting new fee and fill models for the security.
self.set_security_initializer(self._custom_security_initializer)

def _custom_security_initializer(self, security: Security) -> None:
    security.set_fee_model(ConstantFeeModel(0, "USD"))

For simple requests, you can use the functional implementation of the security initializer. This style lets you configure the security object with one line of code.

// Disable the trading fees for each security by passing a functional 
// implementation for the SetSecurityInitializer argument.
SetSecurityInitializer(security => security.SetFeeModel(new ConstantFeeModel(0, "USD")));
# Disable the trading fees for each security by using lambda function 
# for the set_security_initializer argument.
self.set_security_initializer(lambda security: security.set_fee_model(ConstantFeeModel(0, "USD")))

In some cases, you may want to trade a security in the same time loop that you create the security subscription. To avoid errors, use a security initializer to set the market price of each security to the last known price. The GetLastKnownPricesget_last_known_prices method seeds the security price by gathering the security data over the last 3 days. If there is no data during this period, the security price remains at 0.

// Gather the last 3 days of security prices by using GetLastKnowPrice as the seed in Initialize.
var seeder = new FuncSecuritySeeder(GetLastKnownPrices);
SetSecurityInitializer(security => seeder.SeedSecurity(security));
# Gather the last 3 days of security prices by using get_last_known_prices as the seed in initialize.
seeder = FuncSecuritySeeder(self.get_last_known_prices)
self.set_security_initializer(lambda security: seeder.seed_security(security))

If you call the SetSecurityInitializerset_security_initializer method, it overwrites the default security initializer. The default security initializer uses the security-level reality models of the brokerage model to set the following reality models of each security:

The default security initializer also sets the leverage of each security and intializes each security with a seeder function. To extend upon the default security initializer instead of overwriting it, create a custom BrokerageModelSecurityInitializer.

public class BrokerageModelExampleAlgorithm : QCAlgorithm
{
    public override void Initialize()
    {
        // In the Initialize method, set the security initializer to seed initial the prices and models of assets.
        SetSecurityInitializer(new MySecurityInitializer(BrokerageModel, new FuncSecuritySeeder(GetLastKnownPrices)));
    }
}

public class MySecurityInitializer : BrokerageModelSecurityInitializer
{
    public MySecurityInitializer(IBrokerageModel brokerageModel, ISecuritySeeder securitySeeder)
        : base(brokerageModel, securitySeeder) {}    
    public override void Initialize(Security security)
    {
        // First, call the superclass definition.
        // This method sets the reality models of each security using the default reality models of the brokerage model.
        base.Initialize(security);

        // Next, overwrite some of the reality models
        security.SetFeeModel(new ConstantFeeModel(0, "USD"));    }
}
class BrokerageModelExampleAlgorithm(QCAlgorithm):
    def initialize(self) -> None:
        # In the Initialize method, set the security initializer to seed initial the prices and models of assets.
        self.set_security_initializer(MySecurityInitializer(self.brokerage_model, FuncSecuritySeeder(self.get_last_known_prices)))

# Outside of the algorithm class
class MySecurityInitializer(BrokerageModelSecurityInitializer):

    def __init__(self, brokerage_model: IBrokerageModel, security_seeder: ISecuritySeeder) -> None:
        super().__init__(brokerage_model, security_seeder)    
    def initialize(self, security: Security) -> None:
        # First, call the superclass definition.
        # This method sets the reality models of each security using the default reality models of the brokerage model.
        super().initialize(security)

        # Next, overwrite some of the reality models
        security.set_fee_model(ConstantFeeModel(0, "USD"))

To set a seeder function without overwriting the reality models of the brokerage, use the standard BrokerageModelSecurityInitializer.

var seeder = new FuncSecuritySeeder(GetLastKnownPrices);
SetSecurityInitializer(new BrokerageModelSecurityInitializer(BrokerageModel, seeder));
seeder = FuncSecuritySeeder(self.get_last_known_prices)
self.set_security_initializer(BrokerageModelSecurityInitializer(self.brokerage_model, seeder))

Add Data

You can subscribe to asset, fundamental, alternative, and custom data. The Dataset Market provides 400TB of data that you can easily import into your algorithms.

Asset Data

To subscribe to asset data, call one of the asset subscription methods like AddEquityadd_equity or AddForexadd_forex. Each asset class has its own method to create subscriptions. For more information about how to create subscriptions for each asset class, see Asset Classes.

AddEquity("AAPL"); // Add Apple 1 minute bars (minute by default)
AddForex("EURUSD", Resolution.Second); // Add EURUSD 1 second bars
self.add_equity("SPY")  # Add Apple 1 minute bars (minute by default)
self.add_forex("EURUSD", Resolution.SECOND) # Add EURUSD 1 second bars

In live trading, you define the securities you want, but LEAN also gets the securities in your live portfolio and sets their resolution to the lowest resolution of the subscriptions you made. For example, if you create subscriptions in your algorithm for securities with Second, Minute, and Hour resolutions, the assets in your live portfolio are given a resolution of Second.

Alternative Data

To add alternative datasets to your algorithms, call the AddDataadd_data method. For full examples, in the Datasets chapter, select a dataset and see the Requesting Data section.

Custom Data

To add custom data to your algorithms, call the AddDataadd_data method. For more information about custom data, see Importing Data.

Limitations

There is no official limit to how much data you can add to your algorithms, but there are practical resource limitations. Each security subscription requires about 5MB of RAM, so larger machines let you run algorithms with bigger universes. For more information about our cloud nodes, see Resources.

Set Indicators and Consolidators

You can create and warm-up indicators in the Initializeinitialize method.

private Symbol _symbol;
private SimpleMovingAverage _sma;

_symbol = AddEquity("SPY").Symbol;
_sma = SMA(_symbol, 20);
WarmUpIndicator(_symbol, _sma);
self._symbol = self.add_equity("SPY").symbol
self._sma = self.sma(self._symbol, 20)
self.warm_up_indicator(self._symbol, self._sma)

Set Algorithm Settings

The following table describes the AlgorithmSettings properties:

Property: AutomaticIndicatorWarmUpautomatic_indicator_warm_up

A flag that defines if the WarmUpIndicatorwarm_up_indicator method can warm up indicators.

Data Type: bool | Default Value: Falsefalse

Property: DailyPreciseEndTimedaily_precise_end_time

A flag that defines if daily bars should have an end_timeEndTime that matches the market close time (Truetrue) or the following midnight (Falsefalse).

Data Type: bool | Default Value: Truetrue

Property: FreePortfolioValuefree_portfolio_value

The buying power buffer value.

Data Type: decimalfloat | Default Value: 250m

Property: FreePortfolioValuePercentagefree_portfolio_value_percentage

The buying power buffer percentage value.

Data Type: decimalfloat | Default Value: 0.0025m

Property: LiquidateEnabledliquidate_enabled

A flag to enable and disable the Liquidateliquidate method.

Data Type: bool | Default Value: trueTrue

Property: MaxAbsolutePortfolioTargetPercentagemax_absolute_portfolio_target_percentage

The absolute maximum valid total portfolio value target percentage.

Data Type: decimalfloat | Default Value: 1000000000m

Property: MinAbsolutePortfolioTargetPercentagemin_absolute_portfolio_target_percentage

The absolute minimum valid total portfolio value target percentage.

Data Type: decimalfloat | Default Value: 0.0000000001m

Property: MinimumOrderMarginPortfolioPercentageminimum_order_margin_portfolio_percentage

The minimum order margin portfolio percentage to ignore bad orders and orders with small sizes.

Data Type: decimalfloat | Default Value: 0.001m

Property: RebalancePortfolioOnInsightChangesrebalance_portfolio_on_insight_changes

Rebalance the portfolio when you emit new insights or when insights expire.

Data Type: bool/NoneType | Default Value: trueTrue

Property: RebalancePortfolioOnSecurityChangesrebalance_portfolio_on_security_changes

Rebalance the portfolio when your universe changes.

Data Type: bool/NoneType | Default Value: trueTrue

Property: StalePriceTimeSpanstale_price_time_span

The minimum time span elapsed to consider a market fill price as stale

Data Type: TimeSpantimedelta | Default Value: TimeSpan.FromHours(1)timedelta(hours=1)

Property: TradingDaysPerYeartrading_days_per_year

Number of trading days per year for this algorithm's portfolio statistics.

Data Type: integerint | Default Value: 252

Property: WarmUpResolutionwarm_up_resolution

The resolution to use during the warm-up period

Data Type: Resolution/NoneType | Default Value: nullNone

To change the Settingssettings, update some of the preceding properties.

// Set the algorithm behavior using the Settings property on your algorithm.
Settings.RebalancePortfolioOnSecurityChanges = false;
Settings.TradingDaysPerYear = 365;
# Set the algorithm behavior using the settings property on your algorithm.
self.settings.rebalance_portfolio_on_security_changes = False
self.settings.trading_days_per_year = 365

To successfully update the FreePortfolioValuefree_portfolio_value, you must update it after the Initializeinitialize method.

Set Benchmark

The benchmark performance is input to calculate several statistics on your algorithm, including alpha and beta. To set a benchmark for your algorithm, call the SetBenchmarkset_benchmark method. You can set the benchmark to a security, a constant value, or a value from a custom data source. If you don't set a brokerage model, the default benchmark is SPY. If you set a brokerage model, the model defines the default benchmark.

// Set the benchmark to IBM
SetBenchmark("IBM");

// Set the benchmark to a constant value of 0
SetBenchmark(x => 0);

// Set the benchmark to a value from a custom data source
var symbol = AddData<CustomData>("CustomData", Resolution.Hour).Symbol;
SetBenchmark(symbol);
# Set the benchmark to IBM
self.set_benchmark("IBM")

# Set the benchmark to a constant value of 0
self.set_benchmark(lambda x: 0)

# Set the benchmark to a value from a custom data source
symbol = self.add_data(CustomData, "CustomData", Resolution.HOUR).symbol
self.set_benchmark(symbol)

If you pass a ticker to the SetBenchmarkset_benchmark method, LEAN checks if you have a subscription for it. If you have a subscription for it, LEAN uses the security subscription. If you don't have a subscription for it, LEAN creates a US Equity subscription with the ticker. Since the ticker you pass may not reference a US Equity, we recommend you subscribe to the benchmark security before you call the SetBenchmarkset_benchmark method.

Set Time Zone

LEAN supports international trading across multiple time zones and markets, so it needs a reference time zone for the algorithm to set the Timetime. The default time zone is Eastern Time (ET), which is UTC-4 in summer and UTC-5 in winter. To set a different time zone, call the SetTimeZoneset_time_zone method. This method accepts either a string following the IANA Time Zone database convention or a NodaTime.DateTimeZone object. If you pass a string, the method converts it to a NodaTime.DateTimeZone object. The TimeZones class provides the following helper attributes to create NodaTime.DateTimeZone objects:

SetTimeZone("Europe/London");
SetTimeZone(NodaTime.DateTimeZone.Utc);
SetTimeZone(TimeZones.Chicago);
self.set_time_zone("Europe/London")
self.set_time_zone(TimeZones.CHICAGO)

The algorithm time zone may be different from the data time zone. If the time zones are different, it might appear like there is a lag between the algorithm time and the first bar of a history request, but this is just the difference in time zone. All the data is internally synchronized in Coordinated Universal Time (UTC) and arrives in the same Slice object. A slice is a sliver of time with all the data available for this moment.

To keep trades easy to compare between asset classes, we mark all orders in UTC time.

Set Warm Up Period

You may need some historical data at the start of your algorithm to prime technical indicators or populate historical data arrays. The warm-up period pumps data into your algorithm from before the start date. To set a warm-up period, call the SetWarmUpset_warm_up method. The warm-up feature uses the subscriptions you add in Initializeinitialize to gather the historical data that warms up the algorithm. If you don't create security subscriptions in the Initializeinitialize method, the warm-up won't occur.

// Wind time back 7 days from the start date
SetWarmUp(TimeSpan.FromDays(7));

// Feed in 100 trading days worth of data before the start date
SetWarmUp(100, Resolution.Daily);

// If you don't provide a resolution argument, it uses the lowest resolution in your subscriptions
SetWarmUp(100);
# Wind time back 7 days from the start date
self.set_warm_up(timedelta(7))

# Feed in 100 trading days worth of data before the start date
self.set_warm_up(100, Resolution.DAILY)

# If you don't provide a resolution argument, it uses the lowest resolution in your subscriptions
self.set_warm_up(100)

Set Name and Tags

You can categorize your backtest with a name and tags. To set the algorithm Namename, call the SetNameset_name method.

// Set the name of your backtest to help categorize your results.
SetName("Backtest Name");
# Set the name of your backtest to help categorize your results.
self.set_name("Backtest Name")

The SetNameset_name method overwrites the names of backtests of an optimization, which contains the parameter values. To keep these values, create a name that includes them.

// Setting the backtest name to a mix of the parameters can help identify backtests in optimization results. 
var fastPeriod = GetParameter("fast_period", 100);
var midPeriod = GetParameter("mid_period", 200);
var slowPeriod = GetParameter("slow_period", 300);
SetName($"Backtest Name ({fastPeriod},{midPeriod},{slowPeriod})");
# Setting the backtest name to a mix of the parameters can help identify backtests in optimization results. 
fast_period = self.get_parameter("fast_period", 100);
mid_period = self.get_parameter("mid_period", 200);
slow_period = self.get_parameter("slow_period", 300);
self.set_name(f"Backtest Name ({fast_period},{mid_period},{slow_period})");

To add tags to you algorithm, call the AddTagadd_tag method.

// Add additional information to the algorithm with tags for better organization in the backtest results.
AddTag("Long Only");
AddTag("Momentum");
# Add additional information to the algorithm with tags for better organization in the backtest results.
self.add_tag("Long Only")
self.add_tag("Momentum")

Your algorithm can add 20 tags. The name and each tag can have up to 200 characters or else they are truncated.

Post Initialization

After the Initializeinitialize method, the PostInitializepost_initialize method performs post-initialization routines, so don't override it. To be notified when the algorithm is ready to begin trading, define an OnWarmupFinishedon_warmup_finished method. This method executes even if you don't set a warm-up period.

// Notify yourself when the algorithm is ready to begin trading.
public override void OnWarmupFinished()
{
    Log("Algorithm Ready");
}
# Notify yourself when the algorithm is ready to begin trading.
def on_warmup_finished(self) -> None:
    self.log("Algorithm Ready")

Examples

The following examples demonstrate some common practices for initializing algorithms.

Example 1: Set the Data Seed

Some data feeds might not have data on its first fetch, making the algorithm prone to error. The following algorithm seeds the initial price of each asset to its last known price to avoid issues.

// Seed with the last known price of assets to prevent $0 security price errors.
public override void Initialize()
{
    SetSecurityInitializer(new BrokerageModelSecurityInitializer(
        BrokerageModel, new FuncSecuritySeeder(GetLastKnownPrices)));
}
# Seed with the last known price of assets to prevent $0 security price errors.
def initialize(self):
    self.set_security_initializer(BrokerageModelSecurityInitializer(
        self.brokerage_model, FuncSecuritySeeder(self.get_last_known_prices)))

Example 2: Enable Extended Market Hours

Some traders use pre-market data for filtering assets. The following algorithm enables extended market hour data for assets in the universe.

// Add a universe of assets with extended market hours data.
public override void Initialize()
{
    UniverseSettings.ExtendedMarketHours = true;
}
# Add a universe of assets with extended market hours data.
def initialize(self):
    self.universe_settings.extended_market_hours = True

Example 3: Train Models When the Warm Up Finishes

The following algorithm uses the OnWarmupFinishedon_warmup_finished event handler to train a model with data from the warm-up period.

// Train machine learning models after the algorithm warm up is done, so your models 
// are ready as soon as the algorithm starts trading. 
public override void OnWarmupFinished()
{
    Train(MyTrainingMethod);
}
# Train machine learning models after the algorithm warm up is done, so your models 
# are ready as soon as the algorithm starts trading. 
def on_warmup_finished(self) -> None:
    self.train(self.my_training_method)

Other Examples

For more examples, see the following algorithms:

You can also see our Videos. You can also get in touch with us via Discord.

Did you find this page helpful?

Contribute to the documentation: