Daily vs Weekly Data — Our Recommendation
We track weekly numbers, not daily. Our prediction is a weekly call — what SPX, NDX, and IWM will do by Friday's close versus last Friday's close. The only number that tests that prediction is the Friday-to-Friday change. Daily data is noise relative to the question we are asking, and recording it every day would add 5× the work without improving our analysis.
Finviz 1W view + Yahoo Sectors 5D view immediately after Friday US close. Screenshot both. Record SPX, NDX, IWM, sectors. This is your baseline for the week's prediction.
Not a full data pull. Just one line in the learning log: "As of Wednesday, SPX is at X — [ahead of / behind / in line with] our prediction. Key event that shifted the picture: Y." Keeps the team engaged without daily burden.
After Friday US close, pull Finviz 1W and Yahoo Sectors 5D again. Record actual % changes for SPX, NDX, IWM, and all 11 sectors. Compare to prediction. Score calibration. Prepare Monday presentation.
Step 1 — Last Week's Actuals (19–23 May 2026)
Data PullSource: Finviz futures_performance.ashx (1W view) + Yahoo Finance sectors (5D view). Screenshots taken Saturday after US Friday close and committed to evidence/finviz_1W_2026-W21_Sat.png.
Primary Indices — SPX / NDX / IWM
Sector ETF Scorecard — Week 21
| ETF | Sector | Weekly % | Signal |
|---|---|---|---|
| XLE | Energy | +5.7% | Led all sectors. Iran deal talk — energy stocks bid on geopolitical premium despite oil price falling. |
| XLV | Healthcare | +0.9% | Defensive outperform. Steady through the macro noise. |
| XLP | Consumer Staples | +0.7% | Defensive bid early week. Rotation into safety. |
| XLK | Technology | +0.1% | Barely positive. Nvidia beat helped but rate pressure at 4.60% yield capped gains. |
| XLC | Communications | −0.4% | Marginal lag. Ad spend concerns mild. |
| XLF | Financials | −0.3% | Banks dragged despite rising yields — credit concern offset rate benefit. |
| XLI | Industrials | −1.1% | Growth/cyclical pressure under elevated yields. |
| XLY | Consumer Disc. | −2.4% | Consumer spending fears. Rate-sensitive. Worst large-cap sector ex-real estate. |
| XLU | Utilities | −2.7% | Bond proxy crushed by 4.60% yield hit. Most rate-sensitive sector apart from XLRE. |
| XLRE | Real Estate | −2.6% | REITs = rate-sensitive. Yield at 1-year high is direct headwind. |
| XLB | Materials | −3.0% | Worst sector. Global growth demand worry + strong dollar pressure. |
Week 21 story in one sentence: The 10-year yield hitting a 1-year high of 4.60% Monday was the defining event — it crushed rate-sensitive sectors (XLRE, XLU, XLY) while energy stocks oddly led despite oil falling, driven by geopolitical Iran deal positioning and volume flows.
Step 2 — Three Agent Outputs (Built Before Any AI)
Always do this firstCritical rule: These three outputs are written by the Almanac Lead (R3), Macro Lead (R4), and Technical Lead (R5) before anyone opens an AI tool. The point is to form your own evidence base independently. The AI synthesis then challenges it — not replaces it.
CYCLE CONTEXT: 2026 = Midterm election year. Q2–Q3 is the "Weak Spot" of the 4-Year Cycle. Almanac 2026 Outlook explicitly flags tougher trading through Q3 before a Q4 "Sweet Spot" rally. Historical Dow Q2+Q3 midterm avg: −2.0%. S&P: −2.5%. Nasdaq: −6.6%.
MAY VITAL STATS:
· S&P: ranks #8 of 12 months. Up 61% of years. Avg +0.3% normally. Midterm year May avg: −0.7%.
· Nasdaq: ranks #5. Avg +1.1%. Still a seasonal positive for tech specifically.
· IWM: ranks #4. Avg +1.3%. Best historical May index — but midterm year avg −1.0%.
MEMORIAL DAY WEEK PATTERNS (26–30 May):
· Memorial Day week: Dow down 17 of last 29. Moderate bearish pattern.
· Day after Memorial Day (Tue 27 May): Dow down 8 of last 10. Recent trend clearly bearish.
· Week after options expiration: S&P up 30 of 45, avg +0.40%. Moderate bullish pattern. CONTRADICTS above.
· Friday before Memorial Day (Fri 23 May, now past): Dow mixed, avg −0.05%. Confirmed low volume.
SECTOR SEASONALITY (active signals from p.94):
· Banking (XLF): seasonal SHORT window active May–July. Avg −6.3% historically.
· Gold/Silver: seasonal SHORT begins mid-May. Avg −6.8%. Note: contradicts current elevated gold price.
· Materials (XLB): seasonal SHORT May–October. Avg −5.1%.
· Technology (XLK): still in seasonal LONG window through July. Avg +10.9%. Key positive.
ALMANAC SEASONAL BIAS: Cautiously bearish-neutral.
CONFIDENCE: LOW–MEDIUM. Two opposing week-level patterns. Midterm cycle headwind dominates.
THESIS: "Seasonality suggests caution in a midterm year late-May window, particularly for Dow/IWM. Technology seasonal long provides the one clear bullish offset. The Memorial Day week bearish tendency reinforces macro pressure from elevated yields."
Source: Stock Trader's Almanac 2026, pp. 65–66, 94. Accessed Sat 24 May 2026.
· Current Fed rate: 3.50–3.75% (held at April 29 FOMC meeting)
· Next FOMC: June 17, 2026. CME FedWatch: ~96% probability of hold. No cut expected until September at earliest.
· Tone: hawkish hold. Kevin Warsh sworn in as new Fed Chair on May 22 — known hawk. Market watching for tone shift.
· 10-Year yield: 4.60% — hit 1-year high Monday. Eased to ~4.55% by Friday as oil fell.
OIL / IRAN — KEY DRIVER THIS WEEK:
· WTI crude ~$61–62, down ~6% on the week. Trump comments + Iran deal draft rumours drove the drop.
· CRITICAL: Iran deal is NOT confirmed. Both sides remain apart on Strait of Hormuz toll issues.
· Binary risk: deal confirmed → oil stays low, yields ease, stocks rally. Talks collapse → oil spikes $6+, yields rise, stocks sell off.
COMING WEEK CALENDAR (W22: 26–30 May 2026):
· Monday 26 May: US MARKETS CLOSED (Memorial Day). 4-day trading week.
· Tuesday 27 May: Consumer Confidence. AutoZone, Zscaler earnings.
· Wednesday 27 May: New Home Sales. Marvell (MRVL), Salesforce (CRM), Snowflake (SNOW) earnings.
· Thursday 28 May: GDP 2nd estimate + PCE inflation — MOST IMPORTANT DATA POINT OF THE WEEK.
PCE ≥ 2.8% YoY → yields spike, stocks sell off. PCE ≤ 2.5% → yields ease, stocks rally.
· Friday 29 May: No major data.
CONTEXT — NEW FED CHAIR: Kevin Warsh is a known hawk who has criticised the Fed for moving too slowly on inflation. His tone at first public appearances will be watched closely. Any hawkish signal from him could add to yield pressure.
MACRO BIAS: Cautious / binary-risk dominant. PCE Thursday is the week's swing event.
CONFIDENCE: LOW. One data point (PCE) could reverse the entire week's direction.
· Last close: 7,473. Price is ABOVE 8 EMA (~7,420) and ABOVE 21 EMA (~7,380).
· 8 EMA above 21 EMA: YES. EMA condition: ZONE 1 (Bullish) — both rising, price above both.
· Trendline: Uptrend line from March lows connecting ~7,050 → ~7,200 → ~7,350. Price above trendline. No break.
· Resistance 1: 7,500 (round number, not yet broken). Resistance 2: 7,517 (May 14 high — all-time high).
· Support 1: 7,350 (confluence of 21 EMA + trendline — KEY LEVEL). Support 2: 7,200 (prior breakout).
· Breadth warning: only 57% of S&P 500 stocks above their 200-day MA. Narrow rally.
NDX (Nasdaq 100) — DAILY CHART:
· Price above both EMAs. Zone 1 intact. Slightly underperformed SPX last week (+0.7% vs +1.0%).
· Key resistance: 21,500 round number. Key support: 20,800 (21 EMA area).
· Breadth: only 43% of Nasdaq stocks above 200-day MA — notably weaker than S&P breadth. This is the biggest technical warning.
IWM (Russell 2000) — DAILY CHART:
· Outperformed massively last week (+2.7%). Price above both EMAs. Zone 1.
· Resistance: 2,900 (round number — not yet broken). Support: 2,800 (21 EMA).
· Note: IWM outperformance driven by oil drop and Iran deal hope, not fundamental change. Needs follow-through above 2,900 to confirm.
TECHNICAL BIAS: Neutral-Bullish. EMA structure intact across all three indices.
KEY WARNING: Narrow breadth (57% S&P, 43% Nasdaq) is the primary structural concern. A handful of mega-cap stocks are carrying the indices.
INVALIDATION: SPX close below 7,350 = trendline + 21 EMA break. That shifts bias to Bearish.
Charts annotated and committed: evidence/charts_2026-W22/spx_daily.png, ndx_daily.png, iwm_daily.png
Step 3 — Multi-LLM Comparison Table
R6 · LLM Synthesis OperatorAll four models received the identical prompt with the three agent outputs above pasted in. Responses recorded Saturday 24 May 2026, 22:05–22:18 SGT. Raw files committed to evidence/llm_responses_2026-W22/.
Claude (Anthropic)
Led with oil drop reducing inflation pressure as primary bull. Flagged PCE Thursday clearly as binary risk. Did not mention low-volume holiday week amplification. Reasoned carefully, moderate tone.
ChatGPT (OpenAI)
Widest uncertainty band of the four. Emphasised 4-day shortened week as a source of unpredictability. Flagged 10-year yield gravity most strongly. Most cautious overall. Breadth concern noted.
Gemini (Google)
Most bullish of the four. Leaned heavily on oil narrative and IWM outperformance as broadening signal. Acknowledged PCE risk briefly. Did not weight the new Fed Chair Warsh hawkish factor.
DeepSeek
Most conservative. Flagged narrow breadth (43% NDX) as a structural fragility most explicitly. Also noted that Berkshire's $397B cash hoard and smart money cashing out is a late-cycle signal. Only model to raise this point.
| Dimension | Claude | ChatGPT | Gemini | DeepSeek |
|---|---|---|---|---|
| Weekly Regime | Neutral-Bullish | Neutral | Neutral-Bullish | Uncertain |
| SPX % estimate | +0.3% to +0.9% | −0.5% to +0.8% | +0.2% to +1.0% | −1.0% to +1.0% |
| NDX % estimate | +0.2% to +0.7% | −0.5% to +0.8% | +0.3% to +1.0% | −1.5% to +0.8% |
| IWM % estimate | −0.5% to +0.8% | −1.0% to +0.6% | −0.3% to +1.0% | −2.0% to +0.5% |
| Top supporting reason | Oil drop → inflation relief → yield ease | Shortened week, thin volume, upward drift bias | Oil + IWM broadening = health signal | No model: breadth too narrow for confidence |
| Top contradiction | PCE Thursday binary risk | 10-year yield gravity at 4.60% | Iran deal unconfirmed = oil spike risk | Smart money cashing out (Berkshire $397B cash) |
| Missed by all 4? | None explicitly flagged: thin-volume PCE amplification effect + new Fed Chair Warsh's first hawkish signals expected this week | |||
Consensus reading: 3 of 4 models lean cautiously bullish-neutral. DeepSeek is the outlier with "Uncertain" — its breadth and smart money concerns are valid. The % range spread is wide (from −1.0% to +1.0% when you look across all models) — this is a signal that no model has high conviction. That tells you something: if four well-trained AI models cannot agree, the week is genuinely uncertain. State that honestly.
Step 4 — Human Score
R7 · THE DIFFERENTIATORThis is where your team earns its marks. If this section sounds like a summary of what the AI models said, you have not done the Human Score correctly. What did YOUR TEAM see that none of the four AIs adequately weighted?
How strongly does macro drive direction this week? +1 −1 The AI consensus treated the oil drop as a clean bullish signal. We disagree. The Iran deal is unconfirmed — a collapse means oil spikes $6+ in one session. That binary risk makes the macro environment more fragile than +1 implies. New Fed Chair Warsh's hawkish reputation adds further uncertainty.
What does the chart say? +1 +1 Agreed. EMAs in Zone 1, trendline intact, price above 7,350 support. However, we attach a stronger caveat to the NDX breadth warning (43%) than any of the four AIs did. We scored +1 but flagged it as a low-conviction +1.
How much should seasonality influence the call? 0 −1 Claude and Gemini both scored seasonal as neutral/mixed. But we specifically looked at the midterm year row: −0.7% S&P average in May of a midterm year. The AI models did not give the midterm cycle context enough weight. We score −1.
How much do we trust the consensus? +1 0 3 of 4 models lean bullish but the % range spread spans 2 full percentage points (−1% to +1%). When the models agree on direction but disagree sharply on magnitude, that is not genuine conviction — it is AI averaging toward a safe middle. We reduce to 0.
What did none of the AIs raise? 0 −1 This is our key human insight: The PCE data releases Thursday 28 May into a 4-day holiday week with significantly lower-than-normal trading volume. Low-volume environments amplify price moves in both directions. A hot PCE print on Thursday with thin liquidity could cause an outsized selloff — 2× or 3× what the same print would cause in a normal full week. None of the four AI models mentioned this volume-amplification dynamic. We scored it −1.
We are slightly more cautious than the AI consensus of Neutral-Bullish. Our Human Score of −2 pushes us to the cautious side without calling outright bearish. The PCE wild card is the decisive factor — if it comes in hot into a thin-volume Thursday, the week could reverse sharply.
Step 5 — Final Predictions (Week of 26–30 May 2026)
Locked Saturday 21:47 SGTThis prediction was committed to GitHub at 21:47 SGT Saturday 24 May 2026. File: prediction_2026-W22_GroupA.md. The commit timestamp is the immutable record. It cannot be changed. The team owns this call.
All 11 Sector ETF Direction Calls
Step 6 — Plain-English Weekly Brief
The market enters the week of 26 May 2026 in a technically intact but breadth-challenged position. The S&P 500 closed its eighth consecutive winning week at 7,473, still above all key moving averages, but only 57% of its member stocks are above their 200-day moving average — meaning a small number of large companies are carrying the index while most stocks are struggling. This is a structural warning sign.
The macro picture is dominated by two binary events: the Iran nuclear deal (unconfirmed — oil could spike or stay low depending on talks progress) and Thursday's PCE inflation data (the Federal Reserve's preferred inflation measure). A hot PCE print in a shortened, low-volume holiday week could cause an outsized market reaction. Our team believes none of the four AI models adequately weighted this volume-amplification dynamic.
We predict the S&P 500 will end the week slightly up (0% to +0.6%), with technology leading and real estate lagging. However, we hold this view with medium-to-low conviction. Our invalidation is clear: if SPX closes below 7,350 on any day this week, our bullish lean is wrong.
⚠️ This is an educational decision-support exercise. Nothing in this brief constitutes financial advice. All AI model outputs were reviewed and independently assessed by the team before this brief was written.
Step 7 — Week-Ahead Watchlist (26–30 May 2026)
Step 8 — Learning Log Entry
R10 · QA & Learning LogGitHub evidence checklist — all boxes ticked before Monday class:
✓ prediction_2026-W22_GroupA.md committed Sat 21:47 SGT
✓ evidence/finviz_1W_2026-W22_Sat.png — Finviz screenshot
✓ evidence/yahoo_sectors_5D_2026-W22_Sat.png — Yahoo sectors screenshot
✓ evidence/charts_2026-W22/ — SPX, NDX, IWM annotated charts
✓ evidence/llm_responses_2026-W22/ — all 4 raw LLM outputs
✓ almanac_agent_2026-W22.md — Almanac output
✓ macro_agent_2026-W22.md — Macro/News output
✓ technical_agent_2026-W22.md — Technical output
✓ human_score_2026-W22.md — Human Score + override paragraph
✓ learning_log_2026-W22.md — Learning log entry
✓ Release tag vW22 created
✓ README updated with sprint status