How to Read an Acubic Portfolio Proposal

When you finish the Acubic guided conversation, the product shows you a portfolio proposal before you connect a broker account or commit any capital. The proposal lists a set of holdings with their target weights, the construction method used, and the ranking context behind each selection. A lot of new users stall at this screen because the output looks technical when it is actually a straightforward answer to three questions you implicitly asked: which assets, how much of each, and why. This article explains what each part of the proposal means and how to decide whether to act on it, adjust it, or run a new one.

What the proposal contains

The proposal is the output of a three-stage construction process: screen, rank, and allocate. By the time you see it, the process has already narrowed a starting universe of more than 1,500 tickers down to those that pass your stated constraints, ranked the survivors by the method you chose, and assigned weights to the top holdings using an optimization algorithm. The proposal shows you that final output, not the intermediate steps.

Each line in the proposal carries a ticker symbol, its target weight as a percentage of the total capital you specified, and contextual data from the ranking step: the method applied, the lookback period used, and roughly where the asset ranked before the final cut. The weights sum to 100 percent of the capital figure you entered. No order has been placed. The proposal is a plan, not a trade.

For the full description of what data feeds each stage of this process, including the source of historical prices, how company classifications are applied, and what the risk-free rate is used for, see what data does Acubic use.

Why these tickers and not others

Each ticker in the proposal cleared two separate tests before appearing. First, it passed the screening step: your stated constraints, such as position size limits, sector exclusions, and geographic preferences, are applied as hard filters across the full starting universe before any ranking happens. An asset that does not fit your constraints does not appear in the proposal regardless of how well it scores on the ranking step. Constraints come first, and they are not softened during optimization.

Among the assets that passed the filter, the ones in your proposal ranked highest under the ranking method you chose. Market capitalization ranks by company size, favouring larger and more liquid companies. Trailing risk-adjusted return ranks by Sharpe ratio over the lookback period you selected. Trailing price momentum ranks by how much an asset's price moved over that same period. All three methods use historical data: none of them forecasts which asset will perform best over the period ahead.

The number of holdings in the proposal reflects the top of that ranked shortlist, cut at a size that keeps the portfolio diversifiable without spreading capital so thinly that each position becomes immaterial. A holding appears because it cleared your constraints, ranked competitively against the other survivors, and contributed a distinct risk profile to the weighting step that followed. If a particular ticker is absent and you expected it to appear, the most likely reason is that it did not pass the screening step: an excluded sector, a minimum market cap threshold, or a constraint on geographic exposure.

What the weights mean

Weights are assigned by the allocation step, which runs after the ranked shortlist is fixed. A higher-ranked asset does not automatically receive a higher weight, and a lower-ranked asset does not automatically receive a lower one. The allocation method, hierarchical risk parity by default, works from the historical correlation structure of the shortlisted holdings, not from their individual rankings.

The method groups holdings that tend to move together into clusters, then distributes capital across those clusters so that no single group of correlated assets dominates the total risk. Within each cluster, less volatile holdings receive a proportionally larger capital share and more volatile ones receive a smaller share. The result is a portfolio whose risk is deliberately spread across the natural structure that the data reveals, rather than concentrated in whichever assets had the best recent numbers.

A lower weight on a holding does not mean it is less important to the portfolio. In many cases the reverse applies: a holding with low correlation to the others adds genuine diversification precisely because it does not move in the same direction as the rest, and the algorithm holds it at a weight that reflects its contribution to the overall risk structure rather than its standalone size or ranking position. Why Acubic uses hierarchical risk parity explains the method in full, including when to choose mean-variance optimization or CVaR minimization instead.

What the proposal does not tell you

The proposal is not a prediction. The tickers listed are not forecast to outperform over the period ahead. The weights are not a view on which assets will rise relative to the others over the next month or quarter. The construction method uses historical returns, volatilities, and correlations to build a risk-adjusted allocation that fits your stated constraints. It does not project those same inputs forward into price targets.

This matters because it changes what the proposal is for. It is not a signal to act on before an opportunity closes. It is a structured, documented output that you can examine against your own judgment, compare to alternatives by re-running with different inputs, and adjust before committing to anything. The model limitations that apply, including the fact that historical correlations may not hold forward and that regime changes can affect how the method performs, are published in full at the methodology page.

Why Acubic does not predict the market explains this principle in detail: the construction method is designed to produce a robust, disciplined allocation from documented inputs, not to call which way prices are headed. Acubic does not promise returns and cannot predict short-term market moves. The proposal should be judged on whether the construction logic fits your stated goals, not on whether any specific holding is expected to rise.

How to judge whether the proposal fits your goals

Before connecting a broker, reviewing the proposal against a few straightforward checks is worth the time.

• Does the spread of holdings reflect the diversification you intended? The proposal is not simply the largest companies or the strongest recent performers. It reflects the correlation structure of the shortlist: if most of the holdings move together in the data, the optimizer will have distributed capital across distinct groups rather than loading up on correlated names.
• Are there holdings you want to exclude? The screening step respects the constraints you set, but it cannot anticipate every preference. If a holding appears that you do not want for any reason, you can re-run with that asset excluded or with an adjusted constraint.
• Is the number of holdings within a range you are comfortable monitoring? A smaller portfolio concentrates risk; a larger one distributes it more finely but also increases the number of active positions.
• Does the proposal match the risk profile you described in the quiz? The construction method cannot guarantee a specific level of risk in forward markets, but the weights should reflect the tolerance you stated at the input stage. If the overall composition looks more aggressive or more conservative than you expected, the risk constraint inputs are the place to start adjusting.
• Is any single holding carrying a weight that feels disproportionate? In most cases HRP produces well-distributed weights, but if the shortlist contains a cluster of very highly correlated assets, the optimizer may concentrate capital in the assets outside that cluster as the only available diversification. Understanding why a weight is high is more useful than simply reducing it without changing the underlying shortlist.

How to adjust the proposal before you act

Nothing is committed at the proposal stage. You can change the ranking method, the lookback period, the number of holdings, or any of the constraint inputs and run the process again to generate a new proposal. Each run produces a fresh output from the same pipeline using the updated inputs.

If you want to understand how a different construction method would weight the same shortlist, switching from hierarchical risk parity to mean-variance optimization or CVaR minimization changes the allocation step while leaving the screening and ranking steps unchanged. Comparing two proposals built from the same inputs under different optimization methods is often the clearest way to understand what each method is doing. The how does Acubic work guide covers each setting in context and explains the trade-offs between the three allocation methods.

If the shortlist itself is the issue, rather than the weights, the place to look is the screening and ranking inputs. A different ranking method, a different lookback period, or a tighter or looser constraint set will produce a different shortlist, and therefore a different allocation even under the same optimization method.

What happens when you act on the proposal

When you are satisfied with the proposal, the next step is to connect a broker and choose an automation mode. The proposal becomes the live portfolio target when the connection is active and the first rebalance runs. Connecting is a separate step from accepting the proposal: you can save the proposal, study the methodology further, and connect your broker when you are ready.

Once connected, Acubic keeps the portfolio aligned with its target over time. Prices move, and the actual weights drift away from the targets between rebalances. The scheduled rebalance reruns the full construction process from current data, derives new targets, and proposes the trades that close the gap. How that approval step works depends on the automation mode you choose: one-click vs full-auto rebalancing explains what monitoring, one-click, and full-auto each do, and which brokers support which options.

Getting started

The AI portfolio builder is the starting point if you have not yet generated a proposal. If you have one in front of you and want to go deeper on any part of the construction logic, the sources above cover each stage: what data does Acubic use for the full data pipeline, the methodology page for the authoritative method documentation including the stated limitations, and why Acubic uses hierarchical risk parity for the weighting logic in detail.

The proposal is the starting point for your own review, not a directive to act on. Acubic does not promise returns and does not predict short-term market moves. Build, review, adjust if needed, and connect when the proposal reflects what you actually want.