Tonight, within roughly six hours, the market gets the Federal Reserve's interest-rate decision (2am Thursday SGT), the chair's press conference, and then earnings from Microsoft and Meta — two of the biggest companies on earth. Tuesday's session set the stage in the strangest way possible: the Nasdaq 100 fell to a new low for this decline, while nearly everything underneath it improved.
The Nasdaq 100 (NDX) dropped 0.98% to 27,763, and during the day it fell as low as 27,453 — a new low for this whole pullback, briefly breaking below a level our framework has watched for weeks (27,753, a key retracement line). It clawed back above that line by the close, but finished below its lower "Bollinger Band" — think of that band as the market's normal trading lane. Closing outside the lane is rare and usually means one of two things: an exhausted move about to snap back, or the start of a faster slide. We'll know which within hours.
Meanwhile the S&P 500 (SPX) actually rose 0.21% to 7,429, bouncing off a floor it has now tested four times — the last two bounces came within a fifth of a point of each other. The broad NYSE index hit another all-time high. The weakness is concentrated almost entirely in big tech.
◼ Price: NDX new cycle low, below its normal trading band — the worst structural signal of this decline.
◼ Market internals: the share of Nasdaq stocks in uptrends jumped +28.6% in one day (second big jump in a row), more stocks rose than fell for a third straight day, and junk-bond prices — a stress gauge — actually improved. The average stock is quietly repairing while a handful of giants drag the index down.
These two signals cannot coexist for long. Usually a market this stretched would take weeks to resolve the tension. This one gets its verdict on a schedule — tonight.
Markets currently price roughly a 62–64% chance the Fed holds rates steady and a 36–38% chance it hikes by a quarter point. Our framework leans similarly: 65% hold / 35% hike. Why not lower hike odds, given oil just crashed? Because the new Fed chair has been unusually tough on inflation, and oil is still up ~20% this month. Why not higher? Because short-term bond yields — the market's most honest Fed forecaster — are pricing hikes for later this year, not today.
A hold with softer language likely triggers a quick pop. The checkpoints, in order: 27,809 (back inside the trading lane — the first quality test), 27,890, then the gap left behind on Tuesday at 27,927–28,039. A rally that stalls before 28,043 is a weak bounce. Clearing it opens 28,231–28,460. One tell to watch: a friendly Fed that still can't lift the index back above 27,809 would be the single most bearish outcome of the night.
A surprise hike would hit fast, because most investors aren't positioned for it. The downside checkpoints: 27,664 (a major moving average — futures are sitting exactly on it this morning), 27,453 (Tuesday's low), then 27,000 and 26,818. The market's fear gauge (VIX), unusually calm at 18.2, would likely jump toward 20–22.
Both report tonight after the US close. Here's the twist: whether they beat expectations barely matters — prediction markets already price ~92–95% odds of beats. What moves the stocks is their AI spending guidance. Last quarter, Meta beat on everything and still fell 6–7% purely because it raised spending plans. Wall Street now wants proof the hundreds of billions going into AI data centres will pay off. Microsoft is down ~21% this year, Meta ~10% — so some worry is priced in, but the spending question is the live wire.
Combining the Fed (65/35) with the earnings reaction (we estimate a 55% chance the spending story disappoints again, 45% chance it relieves), here's where NDX futures most likely sit by 11am Singapore time Thursday — after both events and the earnings calls have been digested:
Relief rally squeezes through Tuesday's gap. Strongest bounce scenario.
Fed relief at 2pm, reversed after 4pm on capex worries. A round trip.
The two shocks offset. Violent chop, little net movement.
Both land badly. Tuesday's floor breaks; the deep support zone comes into play.
Weighing it all: the probability-weighted landing spot is ≈27,600–27,750 — roughly where futures already trade. Translation: the market's "average expectation" is a wash, but the tails are fat in both directions. Expect a 1.5–2.5% move; direction is the coin the next six hours will flip.
| NDX support | Why it matters | NDX resistance | Why it matters |
|---|---|---|---|
| 27,753 | Key retracement line — held at the close | 27,809 | Trading-lane floor, now overhead |
| 27,664 | 100-day average — futures sit on it now | 27,927–28,039 | Tuesday's gap — first rally magnet |
| 27,453 | Tuesday's low — the break level | 28,231–28,274 | The broken floor from last week |
| 26,818 | Halfway-back level of the whole rally | 28,460 | Monday's high |
| 26,296–26,461 | The deep-value zone (long-term averages) | 28,564–28,814 | The upper shelf |
| SPX support | Why it matters | SPX resistance | Why it matters |
|---|---|---|---|
| 7,410 | Reclaimed average — first cushion | 7,452 | Tuesday's high |
| 7,383–7,337 | The four-times-tested floor | 7,462–7,492 | A dense cluster of averages overhead |
| 7,013 | 200-day average | 7,521 | Upper retracement line |
The "rotation" scenario gained the most this week (+5) on the improving internals; the "shock" scenario was trimmed (−3) as oil calmed and credit markets strengthened. Structural rollover remains the base case: this Friday's monthly close is set to lock in the third momentum peak the framework has tracked since 2018 — a pattern that, in past cycles (2018, 2021, and 1999 as the lone older precedent), preceded multi-month declines.
Illustrations of how the framework maps tonight's scenarios to structure — not recommendations. Levels refer to NDX/SPX/VIX futures.
| # | Idea | Horizon | Shape |
|---|---|---|---|
| 1 | Two-sided breakout bracket around the 2pm decision (buy-stop 28,090 / sell-stop 27,420, whichever fires cancels the other; flatten before earnings) | 1 day | Captures either branch without predicting it |
| 2 | Fade a relief squeeze into 28,231 / 28,460 / 28,700 — only after earnings resolve; stop 28,990; targets 27,809 / 27,453 | 1 day–1 wk | The primary structural idea, staged in thirds |
| 3 | If a hike prints: momentum entry below 27,650, adds on failed retests; stop 27,960; targets 27,453 / 27,000 / 26,818 | 1 day | The 36% tail, structured not chased |
| 4 | Staged buying interest at the deep-value zone 26,820 / 26,540 / 26,320; stop 25,900 | 1 wk | Prepared capital for the worst-case night — historically the framework's most reliable zone |
| 5 | Volatility two ways (futures, not spot): sell vol ~19.5–20.2 on a calm hold; sell a panic spike ~22.5–24.5 only after earnings | 1 wk | Vol is unusually cheap into a double catalyst |
| 6 | The split-tape pair: long SPX at its four-tested floor 7,395/7,383 against short NDX 28,040/28,231, equal size | 1 wk | Owns the rotation itself; Fed-direction-neutral |
| 7 | Structural fade of any August rally into 28,564 / 28,810 / 29,030; stop 29,260; targets 27,000 / 26,461 / 25,882 | 1 mo | The core medium-term thesis |
| 8 | If Friday's monthly close confirms the momentum peak: fade the first early-August bounce; invalidated only if monthly momentum exceeds its 2021 peak (77.88 RSI) | 1 mo | Trading the monthly signal itself — with a hard falsification line |
Next week: five more catalysts follow tonight (Amazon and Apple on Thursday, GDP and inflation data, the Bank of Japan, and Friday's monthly close). Base case: a settling range of 27,300–28,300 with the monthly momentum peak locked in.
August: the framework expects rallies to keep failing under the broken averages (28,600–29,100) and a test of 26,818, then the deep-value zone near 26,300–26,500 — about 5% below current prices.
Six months / year-end: if the 2018/2021 pattern repeats, a staircase lower toward 24,200–26,500 with the S&P holding up better than the Nasdaq throughout. Year-end central range: NDX 25,000–27,000. The falsification line for this entire thesis is public and precise: if monthly momentum (RSI) closes above 77.88, the framework's bearish structure is invalidated — full stop.