The First-Hour Fade: How Earnings Gaps Actually Trade After the Open

TL;DR — We measured every earnings reaction session for 40 of the most-traded US reporters across both reporting cycles of the season — 81 sessions, 66 of them opening on a gap of 2% or more. The retail script says ride the gap. The tape says otherwise: the first hour extended the gap in just 30 of 66 sessions — a coin flip — and on the eighteen monster gap-ups of 10%+, hour one continued only seven times. But the fade is not a reversal; it is a repricing of the entry. On the thirty ≥+5% gap-ups, buying the 09:30 print returned a median +0.45% by the close. Waiting for the first hour to dip and buying the 10:30 print instead returned a median +1.90%, positive 13 of 18 times — same names, same day, same news, four times the median of chasing the open. Chasing a first hour that was already up paid +0.25%. On the short side the pooled numbers are blunter: shorting the twenty-five ≥−5% gap-downs at the open and covering at the close lost a median 0.97% — the panic print charges you for panicking; the short-side money lives in the first-hour bounce (median 2.1% above the open) and in having the locate before the open. One more number to kill a retail habit: 39% of these gaps opened against the direction of the EPS surprise. The headline is not the trade. For a trader in Hong Kong or Singapore, the entire decision window — open, flush, 10:30 confirmation — runs from 21:30 to 23:30 local, after dinner, not in the middle of the night.

Earnings season is the one stretch of the calendar when the market hands day traders guaranteed volatility on a schedule. It is also when the most money is donated at 09:30. This study is a cousin of the gap-chase trap — which showed that ordinary strong opens fade at every gap size — and of the 09:30–09:45 auction work: the opening print is an auction artifact, not a considered price. Earnings mornings are that dynamic with the volume dial turned to maximum.

What we measured, and how

We took the earnings calendar for 40 of the most liquid reporters on the US tape — the mega-cap complex (AAPL, MSFT, META, GOOGL, AMZN, NFLX, TSLA, NVDA, AMD, AVGO, INTC, MU, QCOM, ORCL) plus the high-beta earnings movers a prop desk actually trades (COIN, MSTR, MARA, PLTR, HOOD, SOFI, AFRM, UPST, SHOP, SNOW, CRWD, NET, DDOG, MDB, RBLX, ROKU, DKNG, ABNB, UBER, CVNA, PYPL, XYZ, DIS, DELL, SMCI, MRVL) — across both reporting cycles in the window, December 18 2025 to June 25 2026. For each report we identified the reaction session: the same day for a pre-market release, the next trading day for a post-market release. That gives 81 reaction sessions, every one priced from consolidated-tape regular-session bars: the gap (09:30 open vs. prior close), the first hour (open to 10:30), and the rest of the day (10:30 to the 16:00 close).

The scatter is the finding. If earnings gaps trended out of the gate, the cyan dots would dominate. They don't: 30 of 66 gaps extended in the first hour, 36 went the other way. And size doesn't rescue it — of the eighteen gap-ups above 10%, only seven closed their first hour higher than they opened, and the quarter's three most spectacular gaps (SNOW +35.2%, DELL +31.4%, DDOG +30.9%) all dipped out of the gate. Six of the thirty ≥5% gap-ups printed their entire first-hour high within half a percent of the opening print — the open was the top. Whoever market-buys the first print of a hot earnings gap is filling the auction's imbalance at the day's worst price.

The gap-downs mirror it. RBLX opened −23.3% on May 1 and ripped +10.5% in the first hour; MDB opened −27.8% in March and bounced +3.1% — the two deepest gap-downs in the sample both went straight against the gap. Across the twenty-five ≥−5% gap-downs the median first-hour high printed 2.1% above the open. The panic print is where covering shorts and opportunistic buyers meet; it is not where new shorts get paid.

The honest nulls first

Consistent with how we published the Huddle Index, the things that don't work get printed too:

• The EPS headline is not a direction signal. In 26 of the 66 gapped sessions — 39% — the stock gapped against the direction of the EPS surprise. PYPL beat consensus in February and opened −17.9%. INTC printed ahead of estimates in January and opened −13.8%. ABNB missed in February and opened +9.1%. Guidance, revenue quality, and positioning set the gap; the number on the wire does not.
• Day-1 shorts at the open lose money by the close. Across all twenty-five ≥−5% gap-downs, short the 09:30 print and cover at 16:00: median −0.97%, only 10 of 25 profitable — before borrow cost, on the exact trade the panic invites you into. NET's −15.5% gap that kept collapsing all day is real, but it is the outlier the median already contains.
• Chasing first-hour strength is dead money. The twelve big gap-ups whose first hour was already green returned a median +0.25% from 10:30 to the close. Confirmation you paid full price for is not confirmation — it's the crowd already in.

What does work: let the auction lose, then buy the turn

The tradeable structure in this sample is specific: a large earnings gap-up, sold in the first hour, tends to spend the rest of the day going back up. Of the eighteen ≥+5% gap-ups whose first hour dipped (median dip −2.8%), thirteen closed above their 10:30 price, median +1.90% — four times the median of buying the open, with the worst trigger at −3.98%. MU's June quarter is the archetype: opened +17.7%, flushed −7.0% in the first hour, then climbed +5.75% off the 10:30 print into the close — the dip buyer got paid while the open chaser finished the day underwater. The mechanism is old and institutional: funds re-rating a name after a strong report cannot buy their size at the opening auction — they work orders all day. The open belongs to the imbalance; the afternoon belongs to the re-rating. Day traders and institutions are, for once, on the same side of the tape — the trader is just faster to the turn.

The rule, stated the way it runs on a desk: on a ≥5% earnings gap-up, do nothing at 09:30. If the first hour trades below the open, buy the 10:30 print with a stop under the first-hour low. If the first hour never dips, there is no trade. Thirty qualifying events, eighteen triggers, thirteen winners in the sample — we publish the n so you can weigh it honestly: this is a regime read on one earnings year, not a law of nature.

The case study is GOOGL's April 30 session. The report crosses the wires after the US close on April 29 — it is waiting in your feed at breakfast in Hong Kong. An Asia-based trader has the entire working day to read the release, mark the levels, and confirm the locate math on related names before the reaction session opens at 21:30 HKT. The open prints +6.9% at 374.16; the first hour sells it down to 365.82 — the dip every open-chaser is stopped into. At 10:30 ET (22:30 HKT) the tape turns; the buy at the 10:30 print rides an uninterrupted institutional bid to close at 385.12, +4.3%. Everything a trader had to decide happened between dinner and midnight in Asia; what remained was an exit bracket the desk's risk systems watch regardless of who is asleep.

How the desk uses it

• Treat the 09:30 print as data, not an entry. The opening auction on an earnings name is an imbalance-clearing price. The auction study makes the general case; earnings mornings are its extreme.
• The short side is won before the open. By the time a gap-down prints, borrow in that name is gone or repricing by the minute — the HTB mechanics piece walks the cost math. A desk with multiple locate vendors secures inventory pre-market and shorts the first-hour bounce — median 2.1% above the open — instead of donating to it at 09:30. And the day-1 close costs shorts a median 0.97%: the short thesis, if right, is a multi-day trade that needs borrow that stays.
• Execution is the difference at 10:30. The turn is a minutes-wide window on a fast tape. Hitting it takes hotkeys and direct routes — smart-route defaults are for the quiet middle of the day, not for an earnings tape — and a DMA stack that fills where you point it.
• Asia trades earnings season at the best desk hours in the world. A post-market US report lands in Asia's morning; the reaction session opens after dinner; the whole entry decision closes by 23:30 local. The Singapore–Hong Kong session plan is the daily routine; earnings season is that routine with the payout multiplied.

The single sentence to take: on earnings day the crowd pays the open to feel involved; the desk pays the 10:30 print to get paid.

Related reads

The Gap-Chase Trap · The 09:30–09:45 Auction · Hard-to-Borrow Mechanics · Four HTB Vendors · Singapore & Hong Kong Session Plan.

Joining the desk

If your earnings-season P&L is a string of great reads filled at terrible prices, the problem usually isn't the read — it's entering where the auction wants you to, on infrastructure that can't do anything else. Vortex backs experienced US-equity day traders with DMA via Sterling Trader Pro, multi-vendor HTB locates secured before the open, and a desk that treats the 10:30 turn as an executable event, not a chart annotation. The trader application takes about ten minutes.

---

Methodology: 81 earnings reaction sessions, 2025-12-18 to 2026-06-25, for 40 liquid US reporters (AAPL, MSFT, META, GOOGL, AMZN, NFLX, TSLA, NVDA, AMD, AVGO, INTC, MU, QCOM, ORCL, COIN, MSTR, MARA, PLTR, HOOD, SOFI, AFRM, UPST, SHOP, SNOW, CRWD, NET, DDOG, MDB, RBLX, ROKU, DKNG, ABNB, UBER, CVNA, PYPL, XYZ, DIS, DELL, SMCI, MRVL), covering both reporting cycles of the season. Reaction session = report date for pre-market releases, next trading day for post-market releases; report dates, timing, and EPS surprise from published earnings records. Prices are consolidated-tape (SIP) regular-session bars — hourly bars for one sub-sample, 30-minute SIP bars for the rest, with identical field definitions (gap = 09:30 open vs. prior 16:00 close; first hour = open → 10:30; rest of day = 10:30 → 16:00 close) and cross-validated on overlapping sessions to the penny; the GOOGL case study uses 15-minute bars, spot-verified bar-by-bar against a second source. Reaction days were additionally sanity-checked for an earnings-day volume signature; low-volume, near-zero-gap reactions (e.g., MARA, MSTR) are kept as data. Buckets by |gap|: ≥2% (66 events), ≥5% (55), ≥10% (32). No commissions, borrow cost, or slippage modeled; entries and exits are bar prints. All results are one earnings year's regime on liquid names — published as measured, nulls included. Compiled from public market data — VCG Research.

#EarningsSeason #GapTrading #MarketStructure #OpeningAuction #DayTrading #PropTrading #ShortSelling #AsiaTrading #ExecutionIntelligence #VortexCapitalGroup