Research deep dive: why this basket, why now, and what we excluded
Framework. We approach the weekly “buy once, set stops, review Friday” sleeve as a cross‑sectional dispersion trade powered by event density and attention‑amplified liquidity. In plain English: when many companies report in a short window, surprises cluster; the crowd focuses on a narrow set of narratives; and flows become lumpy. That mix reliably increases the spread between winners and losers. Our edge is to hold a compact, equal‑weight portfolio that tilts toward names with (i) a dated, observable catalyst within the week, (ii) high probability of elevated relative volume (RVOL), and (iii) clean technical posture heading into the print. We then enforce downside symmetry with −8% GTC stops to maintain convexity at the portfolio level.
Expected value (EV) arithmetic. Equal weight at $1,000 per name implies a $10,000 sleeve. Let each name’s weekly return be a discrete variable with three states: beat/trend (+12% to +20%), in‑line (−3% to +5%), and miss/fade (−8% via stop). In earnings‑dense weeks, our internal hit‑rate on beat/trend for high‑attention, liquid names typically ranges 25–35% with another 30–40% in the in‑line bucket. Using a conservative mix (pwin=0.3, pflat=0.4, ploss=0.3) with payoffs (+15%, +2%, −8%), the per‑name EV is 0.3×0.15 + 0.4×0.02 − 0.3×0.08 = +2.5%. Compounded over 10 names with modest correlation (ρ≈0.25 due to shared macro), the portfolio EV approximates +2–3% for the week before outlier effects. The reason we target +10% is not the mean but the right‑tail: one or two outliers at +25–40% (common around AI, platforms, or crypto‑beta) can lift the entire basket, while stops bound the left tail.
Kelly‑lite risk budgeting. A full Kelly fraction on the sleeve would be EV/Var; with weekly vol per name in the 6–12% band and low cross‑name correlation, full Kelly would over‑allocate for a discretionary account. We prefer “Kelly‑lite”—a fixed equal‑weight and a hard stop per name—to ensure that a sequence of misses cannot overwhelm the basket. The portfolio stop suggestion (consider flattening if ≥3 names stop by Wednesday close) is a circuit‑breaker that caps realized drawdown when the macro tape overwhelms micro catalysts.
Selection mechanics. The basket is intentionally crowded into narratives that sponsor strong post‑print trends: AI infrastructure and inference (PLTR, AMD, QCOM), scaled consumer platforms with operating leverage (UBER, SHOP, SPOT, ABNB), retail‑flow/option‑elastic names (HOOD), and a macro‑beta kicker (COIN) that can produce convex outcomes if crypto rallies mid‑week. We avoid very low‑float microcaps despite their raw torque because the “buy once, walk away” constraint punishes names where spread/halts/liquidity are variables we would otherwise manage intraday.
Why each pick fits the brief. PLTR carries AI/government data optionality and prints Monday after the close, giving us two clear sessions of positioning and then a binary that often trends for 48–72 hours post‑call. AMD sits at the intersection of DC GPU demand and roadmap credibility; guidance color on MI deployments can produce double‑digit moves. QCOM has handset recovery and on‑device AI tailwinds plus a history of decisive gaps on prints. UBER and SHOP are classic operating‑leverage vehicles; even modest guide changes can produce multi‑sigma reactions because incremental margins and capital return narratives are now core to their multiple. HOOD is deliberately included for its options‑flow elasticity and retail participation—when positioning is off‑side, moves compound. ABNB and DBX contribute idiosyncratic dispersion with well‑understood drivers (take‑rate/seasonality for ABNB; FCF/efficiency for DBX). Finally, SPOT is a momentum “follow‑through” candidate when margin trajectory and user growth re‑align, and COIN gives us convex beta to any crypto headline impulse after last week’s numbers.
Exclusion logic (what we didn’t buy). We screened and passed on: (1) Binary biotech prints with single‑asset risk (great if monitored intraday, poor fit for a set‑and‑forget sleeve); (2) Ultra‑low float small caps where halts and 5–10% spreads can convert a textbook −8% stop into substantially worse slippage; (3) High‑multiple software without a near‑term event—without a calendar catalyst, the tape’s factor drift dominates outcome and reduces the probability of a large, idiosyncratic move within five sessions; (4) Legacy high dividend cyclicals that may be fundamentally sound but lack near‑term attention—without flows, dispersion tails shrink; and (5) Over‑crowded mega‑caps reporting next week—the setup would force us to hold through a non‑event week, violating the weekly objective.
Liquidity and slippage. Every selected name typically trades deep, tight markets at the open. The “wait up to five minutes” clause exists purely to avoid paying peak opening‑auction spreads on gap days; a quick VWAP‑reclaim trigger improves average fill quality without introducing day‑trading complexity. Stops are stop‑market (versus stop‑limit) to guarantee exit—our experience shows that avoided gap‑through risks are worth occasional adverse prints.
Correlation management. While several positions sit in tech/platforms, their drivers are sufficiently orthogonal (AI infrastructure vs. handset cycles vs. take‑rate/margin vs. crypto beta) to keep realized correlation acceptable over one week. If the macro regime deteriorates sharply at Monday’s open (e.g., VIX spike, ES −0.6%), the plan allows one discretionary swap—rotating a single slot into an energy or staples name with an imminent print—to preserve the event‑density characteristic while diversifying factor exposure.
Path to +10% in practice. The distribution that most often gets us there is (a) one outlier at +25–40% (e.g., surprise AI guide, crypto impulse) plus (b) two to three followers at +6–12%, with the rest mixed and a couple stopped. Because we size equally and cut losers quickly, the winners dominate the P&L map. This is exactly the behaviour we want from a weekly dispersion sleeve: small, bounded left tails and fat, portfolio‑relevant right tails.
Governance and process. The weekend scan (news/filings/analyst actions, short interest, scheduled earnings) feeds the basket; the Monday pre‑market RVOL screen (3× vs 30‑day) informs fill priority; and the Friday scorecard enforces accountability to the +10% target. We deliberately avoid intraday overrides to prevent “story drift”—most of the edge is captured by being correctly exposed and disciplined, not by hyperactive tinkering.
Bottom line. This is a deliberately simple, rules‑based weekly vehicle for capturing earnings‑week dispersion without sitting at the screens. The chosen names maximize the probability of right‑tail outcomes while bounding losses via uniform, mechanical stops. In a market where narrative and flows increasingly co‑determine returns, that combination—event density, attention, RVOL, and risk symmetry—offers a clean, repeatable edge.