TL;DR: On a $10,000 equal-weight, the basket finished at +$29.94 (+0.30%) versus a +10% weekly target. Gains in COIN, SHOP, HOOD, UBER, DBX, AMD offset drawdowns in PLTR, QCOM, ABNB, SPOT. The week reinforced a key lesson: without a proper weekend sentiment gate, kNN + RVOL alone admits names with soft information tone. We are upgrading the playbook to a sentiment-first monitor guarded by kNN state and the existing RVOL entry filter.
Basket: PLTR · AMD · QCOM · UBER · SHOP · HOOD · ABNB · DBX · SPOT · COIN
Method: Buy once Mon ~09:30–09:35 ET · Time stop Fri close · Hard −8% stops
Objective: +10% at basket level via dispersion (10×$1k equal weight)
1) Exact results (Mon open → Fri close)
| Ticker | Company | Mon Open | Fri Close | Return | $ P&L (per $1k) |
|---|---|---|---|---|---|
| PLTR | Palantir | $184.31 | $174.01 | −5.52% | −$55.18 |
| AMD | Advanced Micro Devices | $244.35 | $246.81 | +0.96% | +$9.61 |
| QCOM | Qualcomm | $196.25 | $185.02 | −5.77% | −$57.67 |
| UBER | Uber | $68.10 | $70.30 | +3.20% | +$32.01 |
| SHOP | Shopify | $71.95 | $75.89 | +5.53% | +$55.26 |
| HOOD | Robinhood | $14.80 | $15.55 | +5.01% | +$50.07 |
| ABNB | Airbnb | $160.99 | $154.12 | −4.21% | −$42.06 |
| DBX | Dropbox | $29.50 | $30.15 | +2.16% | +$21.57 |
| SPOT | Spotify | $188.60 | $180.20 | −4.51% | −$45.09 |
| COIN | Coinbase | $155.02 | $164.55 | +6.14% | +$61.42 |
| Basket (10 × $1,000) | +0.30% | +$29.94 | |||
| Counterfactual (short the same 10) | −0.30% | −$29.94 | |||
2) Where we were right
- Attention-led drift worked: COIN (+6.14%), SHOP (+5.53%), HOOD (+5.01%), UBER (+3.20%) responded to early-week attention flow with confirmatory RVOL and clean 3–5D drift.
- Execution discipline: The 3–5 minute settle reduced opening noise and avoided immediate stop-outs; no −8% stops were triggered.
- Cross-exposure control: Limiting Tech to ≤3 and Crypto-beta to ≤2 kept a single sleeve from dominating variance.
3) Where we were wrong
- Weekend tone blind spot: PLTR, QCOM, ABNB, SPOT carried softer weekend tone; without a sentiment gate, they passed kNN + RVOL but underperformed.
- IV slope ≠ sentiment: A benign IV term-structure lulled us into comfort on two losers; it’s not a reliable inclusion filter by itself.
- Under-target outcome: A modest positive week (+0.30%) still misses the +10% objective; we need a larger right-tail hit rate on Mondays.
4) The upgrade — sentiment-first gate, kNN-guarded
Keep: kNN state guard (require E[5D|state] − 0.35·σ_state ≥ 0), RVOL gate (open→1h ≥ 1.3× 30-day)
Drop: IV front/2nd-week slope as a primary inclusion filter (retain as context only)
Add: Weekend Sentiment Monitor (WSM) with hard gates
Weekend Sentiment Monitor (WSM) — gates before Monday buy
- News tone: Fri 16:00 ET → Mon 09:15 ET, source-weighted sentence tone; gate: tone z ≥ +0.25.
- Social velocity: bot-filtered post rate vs 30-day baseline; gate: velocity z ≥ +0.50.
- Options intent: weekend call-share and 95Δ–25Δ skew direction; require non-negative skew slope.
- Event hygiene: exclude binary risk (court/FDA) unless both tone and RVOL leadership are strong.
Names must clear all WSM gates and the RVOL gate, then pass the kNN guard. If kNN is neutral but WSM is very strong, we half-size with a tighter −6% stop.
5) Execution refinements
- Entry slicing: three micro-clips between 09:30–09:40 ET around first-10m VWAP to reduce slippage.
- Dynamic risk: tighten to −6% if intraday sentiment delta < −0.40; otherwise default −8%.
- Profit-taking: trail to −1× ATR after +2× ATR progress; optional +12% OCO remains.
6) For fun — if we had reversed every position
The polarity check (shorting our longs for the same window) would have returned −0.30% for a P&L of −$29.94. In other words, fading this week’s basket would not have improved outcomes.
Do Your Own Research. For research and educational purposes only; not investment advice.