This paper asks whether high-frequency technical trading rules generate statistically
robust and economically meaningful net returns in the Brazilian real/U.S. dollar
futures market, and whether such profits can be reliably exploited ex ante. Using
transaction-level data for U.S. dollar futures (DOL) traded on B3, we evaluate 31,986
candidate rules per horizon at five intraday aggregation levels under Stepwise Superior
Predictive Ability inference, with explicit transaction costs and strict day-trading
constraints. Significant rules peak at 774 at the 60-second horizon, with median
annualized excess returns reaching 49.74% and Sharpe ratios exceeding 4 among
selected rules. Rule significance is state-dependent, consistent with a model in which
episodic predictability arises from a latent directional component that dissipates at
longer horizons. Despite this in-sample evidence, fewer than 4% of significant rules
retain significance out of sample, supporting a conditional-viability interpretation:
profitable rules exist, but identifying and applying them ex ante remains elusive.
Comissão Organizadora
Comissão Científica