What is the momentum factor and why it works on B3

In 1993, two academics from the University of Illinois — Narasimhan Jegadeesh and Sheridan Titman — published a paper in the Journal of Finance that changed how the market thinks about stock selection. The title was dry: Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency. The discovery, however, was not: stocks that rose the most over the past 3 to 12 months tend to continue rising over the next 3 to 12 months. And the effect was large enough to deliver statistically significant alpha even after discounting transaction costs.

Thirty-three years later, the momentum factor continues to survive in global data — including on B3. This post explains what exactly momentum is, why it works despite being known for decades, and how VORTEX QSP uses the factor within a multifactorial architecture.

What it is, mechanically

The canonical formulation is simple. At each rebalancing moment (monthly, in the case of VORTEX QSP), you calculate for each stock in the universe the cumulative return over the past L months, excluding the most recent month (to avoid the short-term reversal effect, which operates over 1-4 week periods). This L is called the lookback period. Typical values range between 6 and 12 months.

Next, you rank the universe from highest to lowest cumulative return. Stocks at the top of the ranking enter a long portfolio. In the original paper, Jegadeesh-Titman also built the short leg with the worst performers; in practical Brazilian commercial use, most investors ignore the short due to operational constraints and still capture most of the premium.

Why it works

There are three mainstream explanations — two behavioral, one based on risk.

1. Under-reaction to news

Investors take time to fully incorporate good news. When a company surprises with earnings, the price rises on the day, but typically continues rising in the months following as analysts revise estimates, institutional funds rebuild positions, and the market processes the new reality. The same applies in the opposite direction — stocks that disappoint continue to fall. Bernard and Thomas (1989) documented this post-earnings announcement drift in detail and it explains much of momentum.

2. Over-reaction over longer horizons

Investors who entered late extrapolate the recent past as the future. "Rose 40% in 6 months, will keep going." Demand feeds the price. The effect eventually reverses — momentum reverts over 3-5 year horizons — but in the 3-12 month interval the trend persists.

3. Risk that nobody measured correctly

Defenders of the Efficient Market Hypothesis argue that momentum is not alpha but rather compensation for some type of risk that CAPM doesn't capture — perhaps crash risk (momentum breaks violently in sudden reversals, like April 2009 globally or November 2020 in Brazil). Daniel and Moskowitz (2016) showed that the factor has extreme negative returns during market reversals, so part of the premium may indeed be compensation for tail risk.

The three explanations are not mutually exclusive. Momentum probably exists because of a mix of all three — and that's why it's so robust.

Momentum on B3: the empirical evidence

Brazilian studies confirm the factor in our market. Mussa, Rogers and Securato (2009) analyzed data from 1995 to 2008 and found statistically significant momentum in 6 and 12-month windows. Heineberg and Procianoy (2003) had already observed the same pattern in earlier data. More recently, studies using the IBrX-100 universe show average annual returns of 4 to 8 percentage points above the benchmark for the top momentum decile portfolio (before costs).

The good news: the factor is strong in Brazil. The bad news: it's also more volatile here than in developed markets — B3 has fewer liquid stocks, and momentum crashes tend to be sharper. That's why momentum alone is risky. Combined with other factors (such as low volatility and quality), it delivers risk-adjusted returns far superior.

Common mistakes when applying momentum

Lookback too short

Looking at only the last 1-3 months captures noise, not trend. Worse: it captures exactly the short-term reversal — stocks that rose sharply in the last month often fall the following month. The empirical sweet spot is between 6 and 12 months, and excluding the most recent month is part of the method.

Forgetting about costs

Momentum requires frequent rebalancing. Each rebalance generates turnover — and turnover generates brokerage costs, bid-ask spreads and market impact. In academic backtests these costs are underestimated (or ignored). In practice, a poorly calibrated momentum strategy can have excellent gross returns and negative net returns. VORTEX QSP uses a hysteresis band — a stock in the top-15 only leaves the portfolio when it falls outside the top-25 — precisely to control this.

Pure momentum without regime hedging

In sudden reversals, momentum bleeds. Anyone running pure momentum in March 2020 saw sharp drawdown. The solution is not to abandon the factor, but to combine it with others that have low or negative correlation during crises — typically low volatility and quality, which defend during risk-off moments.

How we apply it in VORTEX QSP

Momentum is one of five pillars of the composite score. The calculation is the standard Jegadeesh-Titman one (12-1 lookback), normalized cross-section by z-score, aggregated with equal weight to other factors. The combination reduces the volatility of the isolated factor and improves the Sharpe ratio by around 30-50% relative to pure momentum, in B3 backtests.

The complete architecture explanation is in Technology. The empirical results from 7.3 years walk-forward — including month-by-month decomposition — are in Performance.

To wrap up

Momentum is not the only factor. It's not even the best in isolation (low volatility has higher Sharpe). But it may be the most well-documented factor in modern financial literature, and it captures something real about how information diffuses in the market. Decades of attempts to arbitrage it away haven't made it disappear — it just got more sophisticated to exploit.

For a retail investor in Brazil, trying to implement momentum manually is difficult: it requires clean data, monthly calculations, discipline to rebalance even when the ranking says counterintuitive things, and controlled costs. That's exactly the gap VORTEX QSP solves — systematization of the method with discipline and honest disclosure.