英国国债收益率创1998年以来新高，斯塔默领导危机重创英镑

Each pillar is scored 0-100 where 50 = neutral. Fundamental blends ROE, P/TBV percentile, FY trajectory and balance-sheet risk (gilt MTM, surcharge tax exposure). Technical blends 20/50-day SMA structure, RSI(14), realised volatility regime and volume profile. Sentiment uses the IUX24 news scorer (-1 to +1) mapped to 0-100. Smart Money blends insider transaction direction, OBV trend and ADV-relative volume on directional sessions. Macro Overlay is the single most discriminating factor here — weighted at 30% — capturing UK-specific systemic risk that has flipped these names from idiosyncratic equity stories to macro factor proxies. HSBC's macro overlay scores 78 because Asia revenue insulates it; UK domestics score 8-18 because every transmission channel pushes the same way.

News articles are scored by an LLM mode that reads each article in full, identifies named tickers, and assigns: (a) directional polarity (bear/bull/neutral), (b) magnitude (-1 to +1), and (c) confidence weighting. Article-level scores are aggregated by ticker into a volume-weighted impact score over a 7-day window. A score below -0.50 indicates a regime where every credible-source headline is bear-biased; HSBC's neutral score is itself information — when the rest of the UK bank complex is pricing -0.65 to -0.72, the structural insulation hypothesis gains evidence weight.

Cumulative Abnormal Return (CAR) is computed by subtracting an event-window-matched benchmark return from each stock's realised return. UK domestic banks use the FTSE 350 Banks index as benchmark; HSBC uses the global MSCI World Banks index; FXB uses the DXY-equivalent G7 currency basket; EWU uses the MSCI EAFE. A 5-day CAR captures the immediate reaction; the 20-day CAR captures the persistence of the move. The May 11-12 window is still open, but the 5-day CARs for May 5-12 are already among the largest 20-day windows in the 6-month sample for LLOY (-4.4%), BARC (-3.9%) and NWG (-3.6%).

Trend slope is the ordinary least squares coefficient of log(price) regressed against trading-day index over the last 60 sessions, annualised. R² above 0.60 indicates a strong directional regime; readings between 0.45 and 0.60 indicate the trend is genuine but volatile. The 5/20-day targets project the slope forward — a useful short-horizon estimate when no obvious mean reversion is in play. UK domestic banks are all in steep negative-slope, high-R² downtrends; HSBC is in a positive-slope uptrend with the highest R² in the basket, validating the pair construction.

In a regime-shift environment, the momentum signal dominates the mean-reversion signal for the directional bias, while the mean-reversion signal indicates the squeeze risk on countertrend bounces. The trade construction is: enter on momentum (current levels), trail stops using the mean-reversion ceiling (R1 levels), and target the momentum extension (S3 levels). For HSBC the two signals align bullishly — both predict a move higher — which is part of why HSBC is the highest-conviction long in the book.

Alpha decay is modelled as a half-life function: π(t) = π_max · (1 - e^(-t/τ)) · e^(-t/decay_τ), where π_max is the modelled peak return (10-15% net), τ is the build-up half-life (10-15 days) and decay_τ is the post-resolution decay half-life (45-60 days). The estimated half-life of this trade is 2-4 weeks because the political pressure has 72+ Labour MPs already in revolt — the political resolution function is faster than gilt market normalisation. Kill conditions are explicit (see Module 6) to prevent giving back alpha after the half-life.

Trade construction uses (1) signal-score weighted entry — heavier sizing in highest-conviction names (BARC, NWG, LLOY, HSBC); (2) stop placement at the boundary of the prior 20-day range — a level whose violation invalidates the recent regime shift; (3) ATR-based position sizing — risk per name normalised to 25-40bp of portfolio; (4) explicit kill conditions tied to the macro mechanism, not just P&L; (5) monthly review of the alpha-decay path — half-life is 2-4 weeks, so positions are not held indefinitely. Friction estimates: 5-8bps spread on UK LSE bank lines, 1-2% borrow cost annualised on UK shorts, 30-50bps FX hedging if trading from USD base, total round-trip ~1.5-2.5% — manageable against a targeted 6-13% net move.

Jarque-Bera tests whether daily returns are normally distributed — p < 0.05 rejects normality and confirms fat tails, which is critical for sizing because Gaussian VaR understates true risk. Ljung-Box tests for serial correlation in returns; significant Q(20) indicates the momentum signal is real, not noise. Skewness measures asymmetry — all UK domestic banks have left-skew (large down moves more frequent than equally-large up moves), consistent with crisis dynamics. Excess kurtosis above 2 indicates fat-tailed distributions; combined with negative skewness, this creates the asymmetric risk that motivates the explicit kill conditions in Module 6. Historical-simulation VaR is preferred to parametric VaR given the fat-tailed regime.