​1. A Structured, Systematic Approach to Equity Portfolio Construction

Cherries was designed to support disciplined equity portfolio construction in environments where noise, behavioral bias, and inconsistent judgment can materially affect outcomes. Rather than relying on discretionary stock selection or short-term market prediction, the platform applies a structured, repeatable framework to portfolio design, with the objective of improving long-term risk-adjusted performance.

​

The system treats portfolio construction as a mathematical allocation problem under uncertainty. Its purpose is not to forecast individual securities, but to enforce consistency, diversification, and explicit risk control across changing market conditions. Optimization is used as a tool to compare feasible portfolio alternatives within a defined structure, not as a promise of optimal outcomes in all environments.

 

2. Equity Universe and Data Inputs

Cherries evaluates a broad universe of U.S.-listed equities drawn from the major exchanges, including NYSE, NASDAQ, and AMEX. The platform incorporates long-term historical data, with approximately fifteen years of market and fundamental history used to inform analysis and risk modeling.

Data inputs are updated daily and include, among others:

​

• End-of-day prices and return series
   

• Corporate actions such as dividends and splits
   

• Liquidity and trading-volume measures
   

• Volatility and downside behavior metrics
   

• Fundamental indicators derived from reported financial statements
   

All data undergo automated validation and normalization before entering the modeling process. The objective at this stage is breadth and robustness rather than selectivity, ensuring that downstream analysis is applied to a consistent and well-defined investment universe.

 

3. Daily Mathematical Ranking Framework

Each trading day, Cherries applies a mathematically driven ranking process to the equity universe. This ranking step precedes portfolio optimization and serves to reduce noise by identifying securities that exhibit more stable and robust behavior across multiple dimensions.

The ranking framework evaluates characteristics such as:

​

• Consistency of historical returns
   

• Volatility patterns and downside behavior
   

• Liquidity and tradability
   

• Measures of financial stability and fundamental quality
   

The purpose of this ranking is not to predict short-term performance or identify individual “winners,” but to define a ranked universe of candidates suitable for disciplined portfolio construction. By filtering and ordering securities before optimization, the system reduces sensitivity to outliers and unstable inputs, improving the reliability of subsequent risk modeling and allocation decisions.

 

 

4. Covariance Modeling and Risk Structure

Effective diversification depends not only on the individual risk of securities, but on how those risks interact. To capture this structure, Cherries constructs a daily covariance matrix representing the historical relationships between securities within the ranked universe.

​

The covariance matrix provides a quantitative map of co-movement across assets, enabling the system to assess how combinations of stocks contribute to overall portfolio volatility and drawdowns. This approach allows diversification to be treated as a measurable property of the portfolio rather than an assumed outcome of holding multiple securities.

​

By explicitly modeling cross-asset relationships, the system distinguishes between portfolios that appear diversified at the security level and those that are diversified at the risk level. This risk structure forms the foundation for all subsequent portfolio construction and optimization steps.

 

5. Portfolio Optimization Across the Risk Spectrum

Using the covariance structure and the ranked universe, Cherries generates a broad set of feasible portfolio candidates under defined constraints. Portfolio construction is framed as a constrained optimization problem, with the objective of identifying portfolios that efficiently convert risk into return rather than maximizing returns in isolation.

​

The optimization process applies principles from modern portfolio theory to construct an efficient frontier representing optimal risk–return combinations within the feasible solution space. Only portfolios that satisfy explicit constraints—such as stock counts, exposure limits, liquidity requirements, or mandate-specific rules—are considered.

From this frontier, portfolios are selected based on predefined risk targets or benchmark-relative risk levels.

 

The resulting portfolios exhibit the expected monotonic relationship between risk and return and are designed to maintain internal consistency across different risk profiles, rather than relying on ad hoc adjustments or discretionary overrides.

 

6. Role of AI Models in Portfolio Construction

AI models within Cherries are used to support the portfolio construction process, not to replace the underlying mathematical framework. The system does not employ AI to rank the equity universe or to predict short-term price movements.

Instead, AI models operate on the ranked universe and the optimization outputs to assist with:

​

• Translating ranked securities into investable portfolios under multiple interacting constraints
   

• Managing trade-offs between competing objectives, such as diversification, turnover, and exposure limits
   

• Supporting allocation logic where deterministic rules alone may be insufficient
   

This separation of roles ensures that portfolio behavior remains grounded in transparent mathematical structure, while AI is applied selectively to improve robustness and operational flexibility. The result is a controlled integration of AI that enhances portfolio construction without introducing opaque decision-making or reliance on predictive signals.

​

7. Mandates, Constraints, and Customization

Cherries is designed to operate within the practical constraints that govern professional investment mandates. Portfolio construction is performed subject to explicit, user-defined rules that reflect regulatory requirements, investment policy statements, and strategy-specific objectives.

Supported constraints include, among others:

​

• Sector and industry exposure limits
   

• Minimum and maximum position sizes
   

• Liquidity and tradability thresholds
   

• Stock count requirements
   

• Benchmark-relative risk targets
   

• Exclusion rules (e.g., ESG, geographic, or issuer-based)
   

Constraints are applied directly within the optimization process rather than through post-construction adjustments. This ensures that portfolios remain internally consistent and that risk characteristics reflect the full set of imposed rules, avoiding distortions that can arise when constraints are enforced manually or sequentially.

This framework allows investment teams to implement systematic strategies while retaining full control over mandate design and risk expression.

 

8. Portfolio Output, Transparency, and Auditability

The output of the Cherries process is a fully specified equity portfolio with explicit weights, accompanied by a complete set of supporting analytics. Each portfolio is generated as a single, coherent allocation rather than as a collection of independent signals or recommendations.

​

For each constructed portfolio, the platform provides:

• Full position weights
   

• Portfolio-level risk metrics
   

• Historical performance simulations
   

• Benchmark comparisons
   

• Allocation and exposure breakdowns
   

All calculations are reproducible, and portfolio histories can be exported for independent review. Cherries does not execute trades, hold assets, or interface with custody, ensuring a clear separation between portfolio construction and implementation.

​

This emphasis on transparency and reproducibility supports internal review, audit processes, and external reporting requirements common to institutional investment environments.

 

9. Scope and Limitations

Cherries is a systematic portfolio construction platform, not a predictive trading system. The methodology is designed to improve the structure and consistency of equity portfolios over time, but it cannot eliminate risk or prevent periods of underperformance.

Key limitations include:

​

• Reliance on historical data, which may not fully reflect future market conditions
   

• Exposure to broad equity market risk
   

• Sensitivity to structural changes in correlations and market regimes
   

• Implementation effects such as transaction costs, liquidity constraints, and timing differences
   

Performance results are provided for illustration and evaluation purposes only. Past performance does not guarantee future results, and actual outcomes will vary depending on market conditions and how portfolios are implemented.

Cherries does not attempt to forecast short-term market movements. As with all equity-based strategies, losses are possible, including during periods of market stress.