CIO Note
Event-Aware kNN
Execution Discipline
Executive summary
Earnings density rises materially this week with bellwethers across tech, consumer, and industrials (Tesla, Netflix, IBM, Intel, Coca-Cola, GM, American Airlines, GE Aerospace, among others), while cross-asset sentiment remains choppy and the VIX has pushed up to 6-month highs. Beat rates are healthy so far (86% of S&P reporters above EPS estimates), but surprise magnitudes are running below multi-year averages — a classic setup for idiosyncratic dispersion rather than clean index beta. That favors our stock-specific signal stack: kNN state matching + event proximity + residual momentum + crowding checks, with a thin news/sentiment layer used only for tie-breaks and context. Sources: FactSet Insight, MarketWatch, Investors.
Market regime snapshot (why we’re still running the playbook)
- Tape & trend. Index uptrend remains intact despite repeated bearish setups failing to confirm; we avoid macro heroics and focus on single-name edges. (MarketWatch)
- Catalyst density. One of the big earnings weeks of the quarter (60+ major S&P names), increasing dispersion and the value of our event-aware features. (Finimize)
- Vol & sentiment. VIX at 6-month highs argues for tight execution (VWAP windows, ATR guards) but does not preclude positive drift in correctly-scoped names. (Investors)
What we changed after last week (v1.3 → v1.4)
- Event-window neighbor sanitation. Exclude historical neighbors where surprise magnitude > |2σ| skewed post-earnings drift (both directions). Cuts false positives on high-IV names.
- Crowding penalty sharpened. Down-weight names with extreme HF ownership + elevated retail attention proxies (abnormal options volume).
- Sentiment cap tightened. News layer max weight reduced 20% → 12%; it may adjust only within a ±10% kNN score band. Prevents headlines from overruling the math.
Selection method (quick refresher)
- Primary: kNN state vector (5/10/20-day returns, curvature, EWMA(10/20) vol, vol-of-vol, quarticity, distance to earnings, IV vs. RV).
- Secondary: Residual momentum (sector/market-neutral), event proximity, dispersion, crowding, light valuation anchors (de-risk extremes only).
- News/sentiment: Direction/intensity of estimate revisions + headline tone; bounded to 12% weight.
- Risk: ≤3 names per sector; equal weights ($1,000 each); factor constraints (near-zero beta/momentum tilt ex-ante); stops at 2× 20-day ATR; time stop = Friday close.
The 10 for this week (equal-weight longs)
| Ticker | Company | kNN bias | Catalyst | Why now | Risks & guards |
|---|---|---|---|---|---|
| TSLA | Tesla | ↑ (modest) | Earnings this week (Investors) | Event proximity + historically positive post-earnings drift in compressed-vol → expanded-IV states. Residual momentum stabilizing; crowding penalty applied. | Gap risk on IV; margins/ASP mix. Tactic: If pre-mkt IV surges, half-size equity leg. |
| NFLX | Netflix | ↑ | Earnings this week (Investors) | Neighbors with improving 20/60 momentum into earnings + favorable revisions show short-horizon follow-through when IV is not extreme. | Sub growth/ARPU risk. Guard: ATR stop; no chase > +2.5× ATR gap. |
| IBM | International Business Machines | ↑ | Earnings this week (TipRanks) | Defensive tech factor + steady revisions; kNN favors low-vol positive drift clusters around print. | Consulting margins; FX. Standard guards. |
| INTC | Intel | ↑ (tempered) | Earnings this week (TipRanks) | Stabilizing vol + improving estimate momentum; residual momentum (sector-neutral) flipped positive. | Gross margin guide; foundry commentary. No add if pre-print drift > +2× ATR. |
| KO | Coca-Cola | ↑ | Earnings this week (Intellectia) | Defensive carry in volatile macro weeks; neighbors show lower left-tail risk around print. | FX translation; EM elasticities. Standard. |
| GM | General Motors | ↑ (balanced) | Earnings this week (Intellectia) | Event proximity + manageable estimate dispersion; improving residual momentum; crowding not severe. | Incentive spend; EV margin; labor. Tighten stop to 1.75× ATR if pre-print drift > +1.5× ATR. |
| AAL | American Airlines | ↑ (selective) | Earnings this week (Intellectia) | Neighbors prefer falling vol-of-vol + positive revisions into airline prints. | PRASM guide; fuel volatility. If WTI +>3% d/d, consider reduce. |
| GE | GE Aerospace | ↑ | Earnings this week (Investors) | Services + LEAP cycle; neighbors with steady uptrend + event proximity show favorable 5-day drift. | Supply chain; deliveries. Standard. |
| ISRG | Intuitive Surgical | ↑ | Earnings this week (EarningsWhispers) | High-quality growth; neighbors with pre-print consolidation + elevated (but not extreme) IV resolve higher 3–5 days. | Procedure growth; currency. Standard. |
| PG | Procter & Gamble | ↑ (defensive) | Staples cluster window; confirm timing (Nasdaq calendar) | Pairs with KO to stabilize portfolio beta; neighbors show modest positive drift in risk-off tapes. | FX, inputs, pricing power. Standard. |
Sector cap enforced: Tech (TSLA, IBM, INTC)=3; Comm Svcs (NFLX)=1; Staples (KO, PG)=2; Discretionary (GM)=1; Industrials (GE)=1; Airlines (AAL)=1; Healthcare (ISRG)=1.
Execution blueprint (mechanics matter this week)
- Entry: Monday auction → first 30–60 min VWAP window. Avoid chasing any name that opens > +1.5× 20-day ATR above Friday close.
- Sizing: Fixed $1,000 each; if same-week earnings and kNN bias only modest, half-size ($500) and reallocate residual to KO/PG.
- Risk controls:
- Stop-loss: 2× 20-day ATR from entry; trail to 1× ATR after +2.5× ATR move.
- Time stop: Exit Friday close regardless.
- Event gap rule: If a name gaps > +5% on results at the open, wait 15 minutes; if price > open and VWAP rising, permit entry at half size; otherwise skip and reassign.
What could blindside us (and how we’ll react)
- Macro shock: Inflation/policy rhetoric that shift terminal-rate odds; if VIX +>3 pts intraday, tighten stops to 1.5× ATR across the book. (Financial Times)
- Earnings landmines: If two or more names miss within the same sector, cut sector peers by 50% risk intra-week (dispersion clustering rule).
- Crowding unwind: If options-led crowding in TSLA/NFLX turns, do not add on weakness; follow mechanical exits only.
Why the math should dominate this week
Big weeks tempt narrative trading. Our kNN uses like-state analogues that already embed the market’s fear/greed balance around similar earnings clusters and vol regimes. The news layer is kept deliberately subordinate — it adjusts within the uncertainty band, not overrules the historical base rate. With healthy beat rates but below-average surprise magnitude, the edge is in name selection and risk discipline, not macro calls. See: FactSet Insight.
Housekeeping & checklist
- Re-validate exact report times on day-of for each name (pre/post-market). Aggregators: TipRanks, Intellectia, Nasdaq calendar.
- Keep sector cap ≤3; if a tech name slips off schedule, backfill with a Financials/Utilities candidate scoring next-best to maintain factor balance.
- Log fills vs. VWAP; update attribution by signal bucket next Saturday to roll weights for v1.5.
Do Your Own Research. This is a tactical, 5-day equal-weight basket guided by quantitative signals and public information. It is not investment advice and does not consider individual circumstances. Markets can gap; use the stop and time-exit rules as specified.