book
Checkout our new book! Hands on AI Trading with Python, QuantConnect, and AWS Learn More arrow

Alpha

Supported Models

Introduction

This page describes the pre-built Alpha models in LEAN. The number of models grows over time. To add a model to LEAN, make a pull request to the GitHub repository. If none of these models perform exactly how you want, create a custom Alpha model.

Null Model

The NullAlphaModel doesn't emit any insights. It's the default Alpha model.

Select Language:
# NullAlphaModel doesn't emit insights.
self.add_alpha(NullAlphaModel())

To view the implementation of this model, see the LEAN GitHub repository.

Constant Model

The ConstantAlphaModel always returns the same insight for each security.

Select Language:
# Emit up insights that expire in 30 days for all assets.
self.add_alpha(ConstantAlphaModel(InsightType.PRICE, InsightDirection.UP, timedelta(30)))

The following table describes the arguments the model accepts:

ArgumentData TypeDescriptionDefault Value
typeInsightTypeThe type of insight
directionInsightDirectionThe direction of the insight
periodtimedeltaThe period over which the insight will come to fruition
magnitudefloat/NoneTypeThe predicted change in magnitude as a +/- percentageNone
confidencefloat/NoneTypeThe confidence in the insightNone
weightfloat/NoneTypeThe portfolio weight of the insightsNone

To view the implementation of this model, see the LEAN GitHub repository.

Historical Returns Model

The HistoricalReturnsAlphaModel buys securities that have a positive trailing return and sells securities that have a negative trailing return. It sets the magnitude of the Insight objects to the trailing rate of change.

Select Language:
# Add HistoricalReturnsAlphaModel to leverage historical return data for generating alpha signals, identifying trends based on past performance.
self.add_alpha(HistoricalReturnsAlphaModel())

The following table describes the arguments the model accepts:

ArgumentData TypeDescriptionDefault Value
lookbackintHistorical return lookback period1
resolutionResolutionThe resolution of historical dataResolution.DAILY

This model cancels all the active insights it has emit for a security when the security has a 0% historical return.

To view the implementation of this model, see the LEAN GitHub repository.

EMA Cross Model

The EmaCrossAlphaModel uses an exponential moving average (EMA) cross to create insights. When the fast EMA crosses above the slow EMA, it emits up insights. When the fast EMA crosses below the slow EMA, it emits down insights. It sets the duration of Insight objects to be the product of the resolution and fast_period arguments.

Select Language:
# Add EmaCrossAlphaModel to generate trading insights based on the crossing of fast and slow exponential moving averages identifying trends and potential reversals. The duration of insights is set by the resolution and fast_period to ensure timely and relevant trading signals.
self.add_alpha(EmaCrossAlphaModel())

The following table describes the arguments the model accepts:

ArgumentData TypeDescriptionDefault Value
fast_periodintThe fast EMA period12
slow_periodintThe slow EMA period26
resolutionResolutionThe resolution of data sent into the EMA indicatorsResolution.DAILY

To view the implementation of this model, see the LEAN GitHub repository.

MACD Model

The MacdAlphaModel emits insights based on moving average convergence divergence (MACD) crossovers. If the MACD signal line is 1% above the security price, the model emits an up insight. If the MACD signal line is 1% below the security price, the model emits a down insight. If the MACD signal line is within 1% of the security price, the model cancels all the active insights it has emitted for the security.

Select Language:
# Use MacdAlphaModel to generate insights based on MACD crossovers, emitting up signals when the MACD signal line is 1% above the security price and down signals when it is 1% below, while canceling insights if within 1% capturing significant trend changes while avoiding noise from minor fluctuations.
self.add_alpha(MacdAlphaModel())

The following table describes the arguments the model accepts:

ArgumentData TypeDescriptionDefault Value
fast_periodintThe MACD fast period12
slow_periodintThe MACD slow period26
signal_periodintThe smoothing period for the MACD signal9
moving_average_typeMovingAverageTypeThe type of moving average to use in the MACDMovingAverageType.EXPONENTIAL
resolutionResolutionThe resolution of data sent into the MACD indicatorResolution.DAILY

To view the implementation of this model, see the LEAN GitHub repository.

RSI Model

The RsiAlphaModel generates insights based on the relative strength index (RSI) indicator values. When the RSI value passes above 70, the model emits a down insight. When the RSI value passes below 30, the model emits an up insight. The model uses the Wilder moving average type and sets the duration of Insight objects to be the product of the resolution and period arguments.

Select Language:
# The RsiAlphaModel generates insights based on the RSI indicator: emits a down insight when RSI exceeds 70 (overbought) and an up insight when RSI falls below 30 (oversold). This helps identify potential price reversals based on momentum.
self.add_alpha(RsiAlphaModel())

The following table describes the arguments the model accepts:

ArgumentData TypeDescriptionDefault Value
periodintThe RSI indicator period14
resolutionResolutionThe resolution of data sent into the RSI indicatorResolution.DAILY

To view the implementation of this model, see the LEAN GitHub repository.

Base Pairs Trading Model

The BasePairsTradingAlphaModel analyzes every possible pair combination from securities that the Universe Selection model selects. This model calculates a ratio between the two securities by dividing their historical prices over a lookback window. It then calculates the mean of this ratio by taking the 500-period EMA of the quotient. When the ratio diverges far enough from the mean ratio, this model emits generates alternating long ratio/short ratio insights emitted as a group to capture the reversion of the ratio.

Select Language:
# Use BasePairsTradingAlphaModel to analyze security pairs from the universe, generating long/short insights based on deviations from a 500-period EMA of their price ratio to capture mean-reversion opportunities.
self.add_alpha(BasePairsTradingAlphaModel())

The following table describes the arguments the model accepts:

ArgumentData TypeDescriptionDefault Value
lookbackintLookback period of the analysis1
resolutionResolutionAnalysis resolutionResolution.DAILY
thresholdfloatThe percent [0, 100] deviation of the ratio from the mean before emitting an insight1

To view the implementation of this model, see the LEAN GitHub repository.

Pearson Correlation Pairs Trading Model

The PearsonCorrelationPairsTradingAlphaModel ranks every pair combination by its Pearson correlation coefficient and trades the pair with the highest correlation. This model follows the same insight logic as the BasePairsTradingModel.

Select Language:
# Add PearsonCorrelationPairsTradingAlphaModel to rank and trade the most correlated security pairs based on Pearson correlation coefficients, using a similar insight logic as BasePairsTradingModel to capture mean-reversion opportunities.
self.add_alpha(PearsonCorrelationPairsTradingAlphaModel())

The following table describes the arguments the model accepts:

ArgumentData TypeDescriptionDefault Value
lookbackintLookback period of the analysis15
resolutionResolutionAnalysis resolutionResolution.MINUTE
thresholdfloatThe percent [0, 100] deviation of the ratio from the mean before emitting an insight1
minimum_correlationfloatThe minimum correlation to consider a tradable pair0.5

To view the implementation of this model, see the LEAN GitHub repository.

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: