How Acubic Selects Assets for Your Portfolio

A portfolio proposal from Acubic lists specific tickers, not a vague theme. That raises an obvious question before you look at any weight: why these names, and not the thousands of others you could have held instead? The answer is not a black box and it is not a forecast. It is a two-stage process, screen and select, that runs before any weight is decided, and every stage is documented rather than hidden behind the word "AI". This article walks through both stages in order. For the allocation step that comes after selection, see why Acubic uses hierarchical risk parity, and for the full pipeline end to end, see how Acubic works.

The universe you start from

Selection has to start somewhere, and Acubic starts from a broad universe of more than 1,500 tickers rather than a curated shortlist. That size matters: a narrow starting list quietly limits what a portfolio can ever contain, no matter how good the ranking or allocation steps downstream turn out to be. A wide universe pushed through a disciplined screen is a different thing from a short list assembled by hand, and it is the difference this section is about.

The full mechanics of where that universe's underlying data comes from, including the market-data provider and the specific fields used, are covered separately in what data does Acubic use. This article picks up from there and focuses on what happens to that universe on the way to a shortlist.

Screening: a hard filter, not a soft preference

Before any ranking happens, the starting universe is narrowed to the tickers that fit the constraints you set during the guided conversation: your risk tolerance, and any exclusions or limits you specified. This step is a hard filter. A ticker either satisfies your stated constraints or it does not, regardless of how attractive its numbers look in isolation. A name you excluded, or one that falls outside a limit you set, is removed before ranking begins. It is not carried forward at a reduced weight and hoped away.

The order here is deliberate. Constraints are applied first, on the full universe, so every step that follows only ever sees eligible candidates. That is a different design from building an unconstrained portfolio first and adjusting it toward your preferences afterward, which tends to produce a compromise rather than a portfolio that actually reflects what you asked for. Assets are also dropped at this stage if they lack sufficient price history over the evaluation window under consideration, since a ranking computed on incomplete history is not comparable to one computed on a full window.

Selection: ranking what survives the screen

Once the eligible universe is fixed, Acubic ranks the candidates that remain using one of several methods. No single method is treated as universally correct; the right choice depends on what you are optimizing for, which is why the choice is exposed to you rather than made silently on your behalf.

Market capitalization

Ranking by market capitalization favors larger, more established companies. It is the simplest of the available methods and tends to produce a shortlist weighted toward names with longer trading histories and deeper liquidity.

Sharpe ratio

Ranking by Sharpe ratio orders candidates by trailing risk-adjusted return: return earned per unit of historical volatility, measured over your chosen lookback window, using treasury yields as the risk-free rate. This method rewards consistency over the measurement window rather than raw magnitude of return.

Price momentum

Ranking by price momentum orders candidates by how strongly their price has moved over a trailing window. It is worth being precise about what this is and is not. It orders candidates by a historical statistic, not a forecast of where a price is headed next. Strong trailing momentum is not a prediction that a name will keep rising; it is one input, drawn entirely from the past, used to decide which tickers are eligible for the allocation step that follows.

Why the lookback window matters

The two returns-based ranking methods, Sharpe ratio and momentum, look back over a period you can adjust, from three months up to three years, with twelve months as the default. That choice changes what the ranking rewards, and it is worth understanding rather than skipping past as a technical detail.

A shorter window, such as three or six months, reacts faster to a company's recent numbers but is more exposed to a single unusual quarter. A longer window, up to three years, smooths out short-term noise at the cost of reacting more slowly when a company's underlying picture genuinely changes. Twelve months is the default because it is a reasonable middle point for most goals, not because it is presented as objectively correct. An advanced selector exposes the choice for anyone who wants to change it; most people never touch it and get the default automatically.

What this shortlist is, and what it is not

None of the three ranking methods involves predicting which asset will outperform next. Each one turns a historical statistic, whether size, risk-adjusted return, or price trend, into an ordering that the allocation step then works from. That distinction matters more than it might seem: a ranking is a description of the past, applied consistently, not a claim about the future. Acubic does not attempt to time markets or forecast short-term price moves at any stage of this process. The reasoning behind that stance, and why it shapes the whole product rather than just this step, is covered in why Acubic does not predict the market.

From a ranked shortlist to a portfolio

Selection produces an ordered, eligible shortlist. It does not, by itself, decide how much of each name to hold; that is the job of the allocation step, which takes the shortlist and assigns weights using a portfolio optimizer subject to your constraints. Acubic's default optimizer is hierarchical risk parity, which groups holdings by how their returns move together and spreads risk across those groups rather than concentrating it in a handful of correlated names. The reasoning behind favoring that method over a plain risk/return trade-off is covered in why Acubic uses hierarchical risk parity.

Keeping screening, selection and allocation as three separate, explicit stages rather than one opaque scoring step is a deliberate design choice. It means a final allocation can be traced back to the universe it came from and the ranking that produced it, and that changing one stage, such as switching from Sharpe ratio to momentum, does not silently change how the others behave.

Why the order matters before you connect a broker

By the time you see a portfolio proposal, screening and selection have already happened, and the proposal itself is the place to see how selection and allocation combine into specific holdings and weights; that reading is covered separately in how to read an Acubic portfolio proposal. Nothing here requires connecting a broker or moving any money: screening and selection run on the universe and your stated constraints alone, before you decide whether to act on the result at all.

Understanding this sequence is also the honest answer to "why these assets and not others". It is not a hidden signal and not a prediction. It is a starting universe, narrowed by constraints you set, ordered by a method you chose, and handed to an allocator that turns the order into weights. Every one of those steps is documented so that a reader can check it rather than take it on faith.

To see the full construction process from the initial conversation through to a proposal, read how Acubic works, or start building a portfolio directly at the Acubic portfolio builder. The complete methodology and data-source documentation, including model limitations, is published at Acubic's methodology and data sources page, and the broader set of how-to and explainer articles lives in the Acubic guides section.