The 5 pillars of systematic stock picking

#music #musicdistribution #brazil #indie

ARCHITECTURE · MULTIFACTORIAL

The 5 pillars of systematic stock picking

~12 min read · published on May 30, 2026 · complete architecture

Building a quantitative portfolio is not "ranking stocks by one factor and buying the top 10". Anyone who tried that naive approach quickly discovered that: (a) the portfolio becomes concentrated in a few sectors, (b) turnover eats returns via costs, (c) drawdowns are worse than necessary, (d) in some regimes the factor simply fails for months on end.

Serious systematic stock picking requires five pillars that complement each other. This post explains each one, and how VORTEX QSP connects all five.

Pillar 1: Multifactorial composite score

Instead of betting on a single factor, you combine several orthogonal factors (with low correlation to each other). Each one captures a different "regularity" in the market:

Momentum — market under-reaction to news. Dedicated post here.
Low volatility — leverage aversion and lottery preference. Dedicated post here.
Quality — high ROE, controlled debt, stable margins. Captures "good companies stay good".
Value — Price/Book and P/E low relative to sector. Captures "mean reversion" of risk premium.
Low beta — sensitivity to IBOV below 1.0. Close relative of low volatility, but with nuance: captures systematic beta, not idiosyncratic volatility.

The choice to combine 5 factors (and not 3 or 8) is empirical. More factors generate dilution — you end up buying "all stocks". Fewer factors increase regime-adversity risk. VORTEX QSP fixed 5 before backtesting began and didn't change it.

Equal-weight among pillars

Why equal weight, and not weight by "which factor is best"? Because any optimized weighting is overfitting on the past. Equal-weight is the choice that has the lowest variance of performance across different regimes — you don't bet on which factor will lead over the next 12 months, you gain a bit from each one.

PRINCIPLE

"Equal-weight is the best estimate of optimal weights when you don't trust your individual Sharpe estimate." — adapted from DeMiguel, Garlappi and Uppal (2009), who showed that 1/N portfolios beat mean-variance optimization in most out-of-sample windows.

Pillar 2: Hysteresis band

Here's the trick that separates operational strategy from academic backtest. Without hysteresis, a stock that misses the top-15 by a little in one month enters again the next. Transaction costs become what defines the result — usually for the worse.

Solution: a stock enters the portfolio when it reaches the top-15. But it only exits when it falls outside the top-25. The 10 slots between 15 and 25 form the hysteresis band — stocks in this zone "stay" in the portfolio if they were already there, and "don't enter" if they weren't yet.

The effect: monthly turnover drops from ~40% to ~12-15%. Costs drop proportionally. And surprisingly, gross return stays virtually the same — the band introduces minimal noise in selection but eliminates most unnecessary costs.

Pillar 3: Inverse-variance weighting

After choosing the 15 stocks, what weight should each one get? Three common options:

Equal-weight (1/15 each). Simple, transparent, but concentrates risk in the most volatile ones.
Cap-weighted (proportional to size). Becomes basically concentrated IBOV — you lose alpha from stock picking.
Inverse-variance (proportional to 1/volatility²). More volatile stocks get lower weight, more stable stocks get higher weight. Balances risk contribution among positions.

VORTEX QSP uses inverse-variance. The practical result: no individual stock contributes more than ~12% of total portfolio risk, even if it has the highest composite score. Drawdowns are more controlled, and the Sharpe ratio increases.

Pillar 4: Anti-concentration restriction

Without sectoral restriction, a multifactorial strategy typically concentrates in 2-3 sectors that dominate the ranking at that moment. In 2020 it was tech/digital retail. In 2022 it was commodities. Sectoral concentration = macro bet disguised as stock picking.

VORTEX QSP restriction: no sector can represent more than 30% of the portfolio. If pure selection would violate this, the algorithm replaces the marginal stocks of the saturated sector with the next ones in the ranking from another sector.

Cost: slight loss of returns in periods where one sector dominates (and it would be optimal to concentrate). Benefit: much less dependent on a specific macro thesis, more predictable drawdowns, and the product makes sense for an investor who doesn't want extreme sectoral exposure.

Pillar 5: Disciplined rebalancing

The most underestimated and emotionally most difficult pillar. All the engineering above only works if you actually rebalance on scheduled dates, even when:

The market fell and your positions are in the red — exiting now "locks in losses".
A top-ranked stock seems "too obvious" — your intuition says to wait for correction.
A stock you love falls out of the ranking — selling it means "abandoning ship".

Discipline beats intuition in the absolute majority of cases. The average investor's intuition is good at detecting when "something is wrong" — and those are exactly the moments when the backtest empirically would have followed the rules while the human investor quits.

VORTEX QSP generates signal systematically before each trading session opens. Execution is the subscriber's decision. But the signal exists without the day's bias.

How the 5 pillars connect

The pillars are not independent — they operate in sequence:

Composite score (pillar 1) ranks the universe by combined z-score of the 5 factors.
Hysteresis (pillar 2) applies the enter-15/exit-25 rule on the ranking, generating the tentative portfolio for the month.
Inverse-variance (pillar 3) calculates individual weights based on recent volatility of each stock.
Sectoral restriction (pillar 4) checks if any sector exceeded 30%. If so, replaces marginal positions.
Rebalancing (pillar 5) executes before opening, with costs explicitly debited.

What this means for the investor

Well-done systematic stock picking is less about "finding the next Petrobras" and more about replicable process that captures documented premiums in the literature, with controlled risk. The return won't be glamorous — there's no narrative of an exotic pick that multiplied 10x. But historically it beats IBOV with much superior Sharpe, and without depending on perfect investor timing.

VORTEX QSP is exactly this operationalized for retail: the 5 pillars running every month, before opening, in direct interface. No manager opinion dependency, no timing, no narrative of the moment. Process. And honest disclosure of what returned, what drew down, and what failed.

To close

The 5 pillars seem complicated in text. In practice they're fixed rules that run automatically. The complicated part is the discipline of following the signal — especially when the market is hurting. If that's hard for you (it is for most), systematization is exactly the value VORTEX QSP delivers: the hard part is already written in code.