C — Citigroup Inc.

Report date: 2026-05-26 | Sector: Financial Services | Last price: $125.09 | Horizon: 30d | Generated: 2026-05-26T07:29:43.821466Z

Forecast summary

Ensemble point +6.01%

80% CI [-2.47%, +15.43%]

95% CI [-6.54%, +20.75%]

Method dispersion 0.0602

Beats RW baseline YES

MC drift (annual) 98.33%

MC sigma (annual) 35.90%

MC paths 1000

Bull target (90th pct) +30.31% -> $163.00

Base target (50th) +11.64% -> $139.66

Bear target (10th pct) -4.84% -> $119.04

Macro regime risk_on_growth

10y yield 4.56%

3m yield 3.59%

Yield-curve slope +0.97%

VIX level 16.59

VIX z-score (252d) -0.48

Sector ETF XLF

Sector relative (90d) +15.98%

Forecast plot (interactive)

Realised volatility

Yang-Zhang annualized (60d) 35.90%

Close-to-close annualized (60d) 30.94%

Per-method comparison

Method	Weight	Point	80% lo	80% hi	95% lo	95% hi
linear	25.0%	+12.42%	+8.74%	+16.08%	+7.16%	+18.09%
monte_carlo	25.0%	+11.64%	-4.84%	+30.31%	-12.82%	+40.76%
ar1	25.0%	-0.01%	-2.23%	+2.27%	-3.38%	+3.49%
random_walk	25.0%	+0.00%	-11.56%	+13.07%	-17.12%	+20.66%

Factor contributions (interactive waterfall)

Factor contribution table

Factor	Loading	Ticker value	Contribution
value_score	+0.0495	+0.4358	+0.0216
quality_score	+0.0192	-0.7351	-0.0141
momentum_score	+0.0286	+2.1413	+0.0612
lowvol_score	+0.0173	-0.0468	-0.0008
revisions_score	+0.0236	+2.3050	+0.0545
news_activity_score	+0.0402	+0.0583	+0.0023
(intercept)	-	-	-0.0005

Factor decay over horizon

Factor IC over time

Per-factor IC backtest summary

Factor	Mean IC	ICIR	% positive	Cumulative attribution	n periods
value_score	+0.2563	+2.690	100.0%	+0.5383	252
quality_score	+0.1105	+1.096	86.5%	+0.2401	252
momentum_score	+0.1428	+1.354	92.5%	+0.2460	252
lowvol_score	+0.0878	+0.887	82.1%	+0.1870	252
revisions_score	+0.1375	+1.319	88.9%	+0.2738	252
news_activity_score	+0.3540	+3.832	100.0%	+1.0893	252

Snapshot — fundamentals + technical

Market cap $213.35B

P/E (trailing) 15.5

P/B 1.1

Forward P/E 10.0

PEG 0.70

Dividend yield 192.00%

Beta 1.12

52w high / low $135.29 / $73.49

Distance from 52w high -7.54%

Cross-sectional rank 13 / 92 above median

Panel source (trained model: 2026-05-26 (0d old), universe_size=92, lookback_days=252)

RSI(14) 51.6

SMA50 / SMA200 $121.32 / $108.78

ATR(14) $3.17

Avg volume (20d) 10.91M

Profit margin 20.36%

ROE 7.65%

Revenue growth (YoY) +15.90%

Earnings growth (YoY) +56.10%

Debt/Equity -

Current ratio -

Short ratio 0.03

Forecast accuracy — walk-forward backtest (60d lookback)

Horizon	n predictions	Directional accuracy	MAE (return)	RMSE	80% CI hit rate	Pearson(pred, real)
1d	691	50.7%	1.36%	1.90%	79.2%	-0.044
30d	662	57.3%	10.83%	13.61%	66.3%	-0.287
60d	632	57.4%	19.66%	24.53%	56.2%	-0.501

Volume momentum (Granville / CGW / Quong-Soudack)

Composite z -0.492

Active signals 7 of 7

MFI(14) 51.96

CMF(20) -0.1689

OBV z (252d) +1.166

VPT z (252d) +1.066

VW-momentum z -0.707

Volume z (60d) -1.352

Relative volume (vs 20d ADV) 0.58x

CMF z (252d) -1.703

Sector rotation & AI-spillover (v6)

Sector ETF XLF

Rel-return 5d / 20d / 60d -0.29% / -2.82% / +8.41%

Sector mom-z 1m / 3m / 6m -1.17 / -0.72 / -1.08

Rotation phase cyclical

AI-factor beta (60d) -0.068

AI spillover score -0.0062

Risk-off corr regime (60d) 0.33 (>0.6 = risk-off; <0.35 = stock-picking)

Sector breadth (% > SMA50) -

Sector dispersion 20d -

Peer comparison (sector-relative valuation & momentum)

Peer signal

NO PEERS

score +0.00 (0 peers)

P/E percentile

20d momentum percentile

Market-cap percentile

only 0 rows in Financial Services; need >=4

Decision input: cheap_leader → nudge bias one bucket toward LONG; expensive_laggard → nudge toward SHORT; cheap_laggard = value-trap (neutralize); expensive_leader = crowding risk (neutralize). Folded into Claude critique and one-line PM summary.

Multi-timeframe technical analysis (1wk / 1d / 1h)

Per-timeframe verdict (TradingView-style aggregate of 17 indicators: RSI, MACD, BBands, ADX, Stoch, %R, OBV, MFI, CMF, ATR, SMA50/200, EMA9/21, Donchian, price-vs-SMA200). Counts = how many indicators fired BUY / SELL / NEUTRAL; net strength is the weight-sum delta. Crosses (golden/death, MACD signal cross, Donchian breakout) and the SMA200 trend filter carry 1.5x weight.

1wk verdict

NEUTRAL

3 buy / 1 sell / 13 neutral · net +4.0

1d verdict

NEUTRAL

2 buy / 2 sell / 13 neutral · net +1.0

1h verdict

NEUTRAL

2 buy / 3 sell / 12 neutral · net -2.5

1wk NEUTRAL 3/1/13

RSI(14) 59.9

MACD hist +0.2989

Bollinger %b 0.72

ADX(14) 27.7 RISING

ATR % 6.00%

MFI(14) / CMF(20) 48.7 / -0.004

Donchian pos / break 0.69

vs SMA200 ABOVE

n_bars521

1d NEUTRAL 2/2/13

RSI(14) 51.1

MACD hist -0.5553

Bollinger %b 0.41 SQUEEZE

ADX(14) 13.2

ATR % 2.54%

MFI(14) / CMF(20) 52.0 / -0.169

Donchian pos / break 0.44

vs SMA200 ABOVE

n_bars1253

1h NEUTRAL 2/3/12

RSI(14) 55.3

MACD hist +0.0484

Bollinger %b 0.59 SQUEEZE

ADX(14) 26.3 RISING

ATR % 0.67%

MFI(14) / CMF(20) 41.8 / -0.109

Donchian pos / break 0.57

vs SMA200 BELOW

n_bars1729

Cross-timeframe confluence

Confluence LONG (0-1) 0.20

Confluence SHORT (0-1) 0.00

Patterns are intentionally selective (Elder triple-screen, daily/hourly divergence, BB-squeeze+breakout, SMA-alignment, MACD/OBV). Most tickers fire 0-2 patterns on any given day. Below: each pattern with its current state (✓ firing / ✗ not firing).

Long patterns

✗ elder_triple

✗ daily_oversold_hourly_bull

✗ bb_squeeze_break_up

✓ sma_align_long

✗ macd_cross_obv

Short patterns

✗ elder_triple

✗ daily_overbought_hourly_div

✗ bb_squeeze_break_dn

✗ sma_align_short

✗ macd_cross_obv

Intraday nowcast — 1h-bar short-horizon (v6)

Next 24h nowcast

+0.26%

over next 24h on 1h bars

80% band

[-3.61%, +4.28%]

n_bars = 1729 · sigma_1h = 0.0063

Direction vs 30d ensemble

AGREES

1h direction: LONG · 30d direction: LONG

Source: intraday_forecaster_v10_trained_lgbm.
(trained model: 2026-05-26 (0d old), mean_directional_accuracy=0.523, n_features=17) Decision input only -- the 30d ensemble + factor regression remain the primary call. Conflicts surfaced into the Claude critique as horizon_conflict flag when |1h move| > 1%.

Meta-label gate — AFML Ch.3 secondary classifier (v6)

Verdict

meta-label: NEUTRAL

p_take = 0.60 · threshold 0.60

Bias adjustment

meta-label NEUTRAL (p_take=0.60 between 0.40 and 0.60) -- no adjustment.

Method: secondary LightGBM trained on (features + primary direction) -> P(primary is right). Source: meta_labeler_v10_trained_lgbm.
(trained model: 2026-05-26 (0d old), roc_auc=0.641, accuracy=0.608, n_train_rows=8,284) When ABSTAIN with p_take < 1-threshold AND bias non-neutral, the bias is demoted one bucket and a meta_abstain flag is added to the critique. See AFML Ch.3.6 (Lopez de Prado 2018).

Implied vol vs realized vol — options-trade gate (v8 W2)

Regime

NEUTRAL

no decisive vol mispricing — delta-1 setup over options

IV / RV ratio

1.23

z-score: n/a (insufficient history)

30d IV vs 30d RV

IV 33.3% (Polygon)

RV 27.1% (yang-zhang)

Provenance: source polygon_options:contracts_call; expiry 2026-06-26 (31d) · strike $125.00 · 1 strike(s) used.
Caveat: 15-min delayed end-of-day quotes; free-tier Polygon (no Greeks, no real-time). Regime is qualitative input only — NOT a numeric factor on the regression.

FMP fundamentals & analyst consensus (v7)

Quality / leverage

ROE 7.6%

ROIC -

FCF yield -

Debt / EBITDA -

Current ratio -

Margins / valuation

Gross margin 0.0%

Operating margin 34.1%

Net margin 20.4%

P/E (TTM) 15.5

EV / EBITDA -

Analyst consensus

Target mean -

Target high / low - / -

Upside vs last close -

Revisions score +0.000

Trade-bias signal — foundation for trade-guidance layer

Bias strong_buy

Composite z-score +3.117

Conviction 1.0

Recommended playbook A_strong

Suggested position size 4.99% of equity

kelly_0.25 x meta_p_take 0.60 x singleton-proxy -> raw 4.988%, capped at 5.00%; direction=+1; singleton-proxy: news_activity_z=+0.06 (nonzero=True), wf30d_dir_acc=0.5725075528700906 (>0.55=True) — suggestion only, not an order

Strategies on bias side

•Aggressive long-call (40-60 DTE, delta 0.55-0.70, size up)

•Bull call-debit spread (long ATM + short 1 SD OTM, 40-60 DTE)

•Long-dated LEAPS for thesis with multi-month conviction

•Cash-secured short put as covered-position entry

Why this bias

composite_z = +3.12 | bias = strong_buy | conviction = 1.00 | macro_regime = risk_on_growth | ensemble = +6.01%

Sensitivity (OFAT tornado ±1 sd) and stress scenarios

Factor	Loading w_i	Ticker z-value	+1 sd impact	-1 sd impact
value_score	+0.0495	+0.436	+0.0495	-0.0495
news_activity_score	+0.0402	+0.058	+0.0402	-0.0402
momentum_score	+0.0286	+2.141	+0.0286	-0.0286
revisions_score	+0.0236	+2.305	+0.0236	-0.0236
quality_score	+0.0192	-0.735	+0.0192	-0.0192
lowvol_score	+0.0173	-0.047	+0.0173	-0.0173

Stress scenarios (forecast shift from base point)

Scenario	Description	Stressed point	Delta vs base
rates_+100bps	Yield curve +100bps (lowvol -1sd, value -0.5sd)	+8.22%	-4.20%
recession_risk_off	Recession (momentum -1.5sd, lowvol +1sd, news -1sd)	+5.83%	-6.58%
quality_flight	Quality flight (quality +1sd, revisions +1sd, momentum -0.5sd)	+15.28%	+2.86%

Multi-source live sentiment (free real-time)

Composite z-score -0.531

Active signals 3 of 3

Sources stocktwits, apewisdom, iv_skew

StockTwits bull 6

StockTwits bear 1

StockTwits net +0.71

Message volume 30

ApeWisdom mentions 1

24h change +0

Reddit rank 290

25-delta IV skew +2.3076

IV-skew read bearish

Analyst critique

Agreement with model agree_with_caveats

Confidence 0.5

PM one-liner C Financial Services: quant +6.01%/30d, disp +0.0602, beats RW, macro neutral, 0 catalyst(s).

Sensitivity concern If value_score (the highest-loading factor) is mis-measured by +/-1 stdev, the bias likely flips.

What the model may have missed

fewer than 3 same-sector peers in mcap band 0.2x-5x (only 0 rows in Financial Services; need >=4) -- peer cross-check unavailable for this ticker
no scheduled catalysts within 30d -- model is a pure factor bet, no event-pricing component
IV-RV neutral (33.3% IV vs 27.1% RV, ratio 1.23x) -- no decisive vol-mispricing edge; delta-1 setup over options.

Critique flags

{'type': 'model_uncertainty', 'severity': 'medium', 'evidence': 'method dispersion +0.0602 exceeds 5% (80% band width +17.90%)', 'suggested_check': 'inspect per-method points -- if outlier method drags ensemble, consider trimmed mean or median'}
{'type': 'placebo_audit_failed', 'severity': 'high', 'evidence': 'GT-score 0.52 > 0.5: real-acc 50.7% vs max placebo-acc 52.6% (AMG -0.019)', 'suggested_check': 'setup may be spurious -- inspect feature-importance for noise-like features; consider increased abstention threshold'}

Factor contribution methodology — where the numbers come from

Point return formula (linear factor model): r_hat = alpha + sum_i (w_i × f_i) where w_i is the cross-sectional loading for factor i (estimated from the cross-section of US large-cap names) and f_i is the z-scored factor value for this ticker (Grinold-Kahn winsorization at ±3).

Per-factor contribution shown in the waterfall above is exactly w_i × f_i (decimal-return units), so the bars sum (with intercept) to the linear forecast point. The ensemble point combines linear / Monte Carlo / AR(1) / random-walk per per-method-comparison table.

Bias mapping (composite z-score → LONG/NEUTRAL/SHORT): weighted sum of value (1.0), quality (0.7), momentum (1.0), low-vol (0.5), revisions (1.0), news-activity (0.6); thresholds ±1.0 with macro override (risk-off + LONG → reduce to NEUTRAL; risk-on-growth + SHORT → reduce to NEUTRAL). LONG → Playbook A (long-call / call-debit-spread / covered-call); SHORT → Playbook B (long-put / put-debit-spread / bear-call-spread); NEUTRAL → cash.

Sensitivity (one-factor-at-a-time tornado above): impact = w_i × 1sd per factor (MSCI / Two-Sigma Venn convention). Stress scenarios are canned multi-factor shocks (rates +100bps; recession risk-off; quality flight).

Factor decay uses exponential half-life per factor type: catalysts 7d, news 14d, revisions 30d, momentum 45d, value 90d, quality 90d (Tetlock 2007; Cohen-Malloy-Pomorski 2012). The heatmap above shows contribution × exp(-ln(2) × t / half_life) on a 5-day grid.

Factor IC backtest: Spearman rank correlation between each factor's cross-sectional ranking on day t and the realized h-day-forward return, aggregated across the panel (Grinold-Kahn 2000 ch.4 fundamental law of active management). Source for this report: synthetic_panel_252d_100tickers_seed42 (parallel runner).

Placebo audit (v6, GT-Score): GT-score = 0.52 SPURIOUS — setup likely lacks real signal . Computed from 3 seed(s) across placebo kinds: shuffled, gaussian_iid, garch. Method: real directional accuracy = 50.7%; max placebo accuracy = 52.6%; AMG = -0.019. Source: placebo_audit_v6_ar1_runner. Provenance: arXiv 2604.15531 "Spurious Predictability in Financial ML" (Paper #3). Risk-flag spurious_predictability_audit_failed added to narrative.

Catalysts in window (30d)

No catalysts within the forecast window.

Analyst narrative

Bull case

C ensemble forecast is +6.01% with 80% band [-2.47%, +15.43%] over 30 days. Positive drivers: momentum_score, revisions_score. 0 catalyst(s) in the window provide event-driven upside potential.

Bear case

Inter-method dispersion (+0.0602) and beats random-walk argue caution. Negative drivers: quality_score. Macro regime risk_on_growth historically caps single-name conviction.

Synthesis

Weight the ensemble's +6.01% center against +0.0602 method dispersion. Risk On Growth regime + 0 catalyst(s) define the setup. Treat the 80% band as the working trade-sizing envelope.

Risk flags

wide method dispersion -- model uncertainty elevated
no catalysts in window -- pure factor bet
spurious_predictability_audit_failed (GT-score=0.52 > 0.5; AMG=-0.019)