CPI Mỹ chạm mức 3,8% khi chi phí năng lượng trói buộc Fed — Việc cắt giảm lãi suất bị hoãn đến năm 2027

Each pillar is scored 0-100. Fundamental uses EV/Sales, EV/EBITDA, ROIC, FCF yield, and net debt percentiles versus historical 5-year ranges. Technical combines RSI(14), distance to SMA50/SMA200, 30-day vol, and price location within 52-week range. Sentiment integrates news flow polarity from the topic-news set (250+ scored articles) plus analyst price-target gap. Smart Money weighs insider net flow, institutional position changes, and price-target dispersion. For shorts, higher score = stronger bearish edge; for longs, higher score = stronger bullish edge. TLT is scored only on technical and sentiment pillars because it is a passive ETF with no fundamental cash flows.

Each ticker's sentiment score aggregates topic-news polarities tagged in the 506-article news scan from May 9-13, 2026. Each article's analyzed_assets field provides a -1 to +1 sentiment score per affected symbol with confidence and impact (direct vs indirect). We aggregate via direct-impact-weighted average, then truncate to articles within the 7-day window. Scores below -0.50 are strong-bearish; above +0.50 are strong-bullish.

Cumulative Abnormal Return measures stock performance relative to the SPY benchmark over 5 and 20 trading-day windows following the event. We compute geometric daily returns minus benchmark-beta-adjusted SPY returns, then accumulate. CARs above +10% over 20 days are statistically significant given typical large-cap dispersion of 3-5%. The +85% 20-day CAR on INTC is a tail observation that historically reverses within 60 days.

Momentum forecast uses log-linear regression of last 30 trading days, projecting via constant daily drift. Higher R² indicates a more reliable extrapolation, but R² > 0.90 with double-digit weekly slopes (INTC, HUM) is itself a sign of an unsustainable parabolic regime — these are explicit mean-reversion candidates. The Z-score column compares current price to its 90-day mean, normalized by standard deviation; values beyond +2σ have a 95% historical probability of reverting within 60 trading days.

Realized volatility above 80% annualized in any large-cap name almost always signals positioning extremes rather than fundamental risk. INTC's 100.6% reading is comparable to the 2020 SPACpark dispersion peak and the 2021 meme-stock cohort — historical priors show 60-day forward vol decay of 40-60% in such regimes, typically accompanied by price normalization toward the 90-day mean.

The half-life estimate is derived from prior Fed hawkish-pivot regimes (2010 Warsh QE2 dissent, 2013 taper tantrum, 2018 Powell pivot, 2022 Q1 hawkish turn). In each case, the style-rotation pair-trade alpha peaked at 8-12% net over 40-60 trading days, then decayed as positioning normalized. Our central estimate is 10.5% peak alpha at day 60, decaying to 10.5% terminal by day 120. Kill conditions force exit at -5% drawdown to preserve capital for regime-confirmation re-entry.

Insider net flow aggregates Form 4 filings since Feb 1, 2026. We classify M-Exempt option exercises followed by sales as effective selling. 10b5-1 scheduled plans (Ruemmler at GS) are flagged separately but not signal-weighted. The OBV proxy uses sign-of-day return weighted by volume; volume anomaly compares 1-day volume to 30-day moving average. Smart Money score combines these three components with insider net flow weighted 50%, OBV direction 30%, volume anomaly 20%.

Entry, stop, and target levels are derived from technical levels (SMA20/50/200, prior congestion zones, analyst consensus PT) cross-validated by ATR-based volatility scaling. Position sizing uses the 0.5% risk-per-trade rule: shares = (0.5% × portfolio) / (entry − stop). This auto-scales positions inversely with volatility, giving HIMS and BAC larger share counts than INTC and HUM. The expected Sharpe range of 1.2-1.8 forward is based on prior Fed-pivot regime backtests (2018, 2022 Q1) where similar pair-trade structures delivered 1.4-2.0 Sharpe over 60 trading days.

Jarque-Bera tests for normality of daily returns using skewness and excess kurtosis — small p-values reject normality. Ljung-Box Q-statistic tests for autocorrelation in returns through 10 lags; small p-values indicate non-random structure (momentum or mean reversion). VaR is historical (non-parametric) at the 95th and 99th percentile of the 90-day return distribution. CVaR (Expected Shortfall) is the conditional mean of returns below the VaR threshold — a more conservative measure of tail risk.