An overview of how
our algorithm works
How our algorithm works
All of NightSky’s strategies are powered by our proprietary algorithm that we developed in-house. Built in Python, the algorithm analyzes historical market and company data to form forward-looking expectations, then uses mathematical optimization to construct diversified portfolios that seek to balance expected return and risk.
Our algorithmic process is consistent across our strategies, while the investment universe and constraints are tailored to each objective—allowing us to apply a consistent, disciplined process across different mandates, such as U.S. Large Cap and U.S. Small Cap.
The animations below illustrate a simplified version of how our algorithm functions:
① 5+ million data points
We maintain a growing database of over five million data points, combining historical stock returns with detailed information on nearly every U.S.-listed company—such as its sector, industry, size, and trading activity.
② One framework, two
proprietary strategies
Our investment process is driven by one proprietary algorithmic system. We use that same system to run both our U.S. Large-Cap and U.S. Small-Cap strategies, with the distinction coming from the universe of stocks each strategy targets.
③ Proprietary forecasting
Next, the algorithm seeks to detect where capital is flowing within the market and identify stocks that appear more attractive by using historical data to measure momentum. Using this framework, the algorithm makes a prediction about how each stock may perform over the next three months.
④ Mathematical
optimization
Once forecasts are complete, the algorithm applies mathematical optimization to identify the mix of stocks and position sizes that best balances expected return and risk by using the Sharpe Ratio, a standard investment measure that assesses return per unit of risk.
⑤ Final output: A high-conviction, systematic portfolio
The result is a curated portfolio of 20–40 stocks — each selected and sized to complement the others, creating a collection of holdings that are often uncorrelated and designed to be collectively optimized for the quarter ahead.
⑥ A new portfolio
every 3 months
We re-run this entire process every three months, at the start of each calendar quarter. This cadence seeks to keep our strategies nimble and adaptive to evolving markets without being overly reactive to short-term fluctations and noise.