Investment Philosophy
Our Platform
Our scientific approach is powered by significant investments in three key areas: People, Technology, Data.
People
Two Sigma's ~1,700 employees include 250+ PhDs amongst more than 1,000+ data scientists, engineers and other technical professionals working collaboratively to create alpha through systematic identification of value opportunities.
Data
Two Sigma's diverse and expanding library of structured and unstructured information - from 10,000+ data sources - is housed in a proprietary data management system.
Technology
Computing power that would rank in the world's top 5 supercomputer sites, uses 380+ PB of storage capacity and infrastructure to run over 48,000 simulations daily.
Every day we apply a scientific approach to the investment management process.
- Data Sourcing & Preparation
- Our teams source, organize, and enrich economically relevant data to power research.
- Modeling
- We systematically capture persistent signals.
- Portfolio Construction
- We work to determine optimal portfolio.
- Execution
- We use efficient trading strategies to capture alpha.
Filter
-
Data Sourcing & Preparation: Our teams source, organize, and enrich economically relevant data to power research.
-
Modeling: We systematically capture persistent signals.
-
Portfolio Construction: We work to determine optimal portfolio.
-
Execution: We use efficient trading strategies to capture alpha.
Data Sourcing
We pursue the world's most valuable data--300 petabytes and counting--using our deep expertise to identify and source the most suitable datasets to test specific hypotheses.
Data Preparation
Ongoing investments in our proprietary technology ecosystem allow us to clean, label, and validate the data we’ve acquired (from more than 10,000 different sources) quickly and at scale.
Featurization
Transforming data into features (data + IP) enables ideation, testing, and implementation of models that incorporate nuanced hypotheses.
Modeling
We aim to capture a wide array of market drivers though a style-agnostic research process.
Modeling
We apply a toolkit of quantitative techniques--from simple ridge regressions to natural language processing--to extract signal from the noise in a wide spectrum of traditional and non-traditional data.
Modeling
Our research infrastructure runs over 100,000 simulations on market data daily.
Modeling
Through a long history of innovation, we've developed a set of core principles that guides our approach to AI.
Why Human Intuition Is Essential in Machine Learning
Claudia Perlich, Head of Strategic Data Science for Two Sigma Investment Management, explains why, despite a lot of hype, machine learning models can only be as good as the skills and intuition of the humans who create them.
Insights
by
Claudia Perlich
Updated on
June 24, 2024
Portfolio Construction
We analyze data in real time to generate independent forecasts and form a "consensus view" for each potential instrument in a portfolio. We systematically combine that view, plus trading costs and risks, to create a target allocation.
A Machine Learning Approach to Constructing an Inflation-Themed Equity Portfolio
Using a machine learning model, we show how to construct a thematic equity portfolio designed to outperform in high-inflation periods and underperform in low-inflation periods.
Insights
by
Simon Lau, Alex Botte
Updated on
January 26, 2022
Systematic and Discretionary Macro: Better Together
Systematic and discretionary macro funds have had similar risk-adjusted returns on average, and intuitively have different relative advantages. This analysis demonstrates that over the past ~15 years, a blend of the two delivered a higher Sharpe ratio than either approach alone.
Insights
by
Adam Akant
Updated on
August 2, 2023
Execution
Our world class-execution capabilities rely on a shared platform and proprietary technology to monitor risks in real time and adjust optimization parameter levels as necessary.
How We Approach AI
Apply thoughtfully
Use the right tool for the job; rigorous modeling process to minimize overfitting.
Constantly assess innovations
Monitor and adopt cutting-edge advances across disciplines, including generative AI.
Pursue collaboration
Promote cooperation among specialized modeling and engineering teams to build competitive techniques.
Seek academic partnerships
Enhanced institutional machine learning expertise through collaborations with leading academics.
Utilize extensive feature library
Leverage enriched, high-quality data to fuel machine learning insights.
Aim for automation and scale
Drive faster turnaround of research and model development
