Markets

Overnight Bets on AMD Point to a Lower Open

When shares followed a big move in a perp, a contract that tracks the stock overnight, a median 55% of the move came by midnight. The study measured each move through 8 a.m. Eastern Time (ET).

Key points

  • By then, perps had made a median 24% of their eventual morning move, measured from each contract’s price at the regular stock market close.
  • On Tuesday, Aug. 11, the CoreWeave (CRWV) perp had gained 15.7% by the start of overnight share trading.
  • The study covers 22 stocks with hourly perp prices available from exchange price data.

The AMD$454.37-15.74% perp, a contract that tracks AMD shares through the night, fell 6.5% ahead of Friday’s open. That points to a lower opening price for shareholders, a prospect already visible Thursday evening. Much of the overnight move typically came early in the 614 cases OptimistFi studied: the perp moved 2% or more by 8 a.m. ET, and shares opened at least 0.5% from their prior close in the same direction.

On those nights, a median 55% of the perp’s move from the regular stock market close to 8 a.m. ET had already happened by midnight. The study ran from the night of April 27 to the night of Nov. 18, 2026.

Tonight's stock perps since Thursday, Nov. 19's close
StockPerp by 8 p.m.By midnightBy 4 a.m.By 8 a.m.
AMD-3.4%-3.5%-3.4%-6.5%
CoreWeave (CRWV)-2.3%-3.4%-2.6%-5.5%
Micron (MU)-1.8%-2.9%-2.3%-5.0%
Intel (INTC)-1.9%-2.6%-3.0%-4.6%
Rocket Lab (RKLB)-0.9%-2.0%-2.5%-3.9%
Palantir (PLTR)-1.5%-2.1%-2.0%-3.9%
Meta (META)-1.3%-1.6%-1.5%-3.5%
Nvidia (NVDA)-0.9%-1.7%-1.3%-3.3%
Oracle (ORCL)-1.0%-1.8%-1.9%-3.2%
Broadcom (AVGO)-0.9%-1.4%-2.1%-3.1%
Hims & Hers (HIMS)-1.0%-1.1%-1.2%-2.9%
Robinhood (HOOD)-0.9%-2.4%-2.0%-2.7%
Circle (CRCL)-1.0%-2.5%-1.7%-2.6%
Strategy (MSTR)-0.4%-1.7%-0.7%-2.3%
Coinbase (COIN)-1.3%-2.3%-1.4%-2.3%
Amazon (AMZN)-0.3%-0.8%-0.7%-2.3%
Tesla (TSLA)-1.1%-1.0%-1.4%-2.2%
Microsoft (MSFT)-0.7%-0.8%-1.1%-2.2%
Alphabet (GOOGL)+0.3%-0.2%+0.1%-0.6%
Netflix (NFLX)-0.1%0.0%+0.2%+0.3%
Alibaba (BABA)-0.5%+2.5%+1.9%+0.2%
Apple (AAPL)+0.7%+0.2%+0.4%+0.1%
Each stock's perp on a 24/7 exchange: its move since the 4 p.m. ET close at each hour, to 8 a.m. ET Friday, Nov. 20.

The AMD perp had fallen 3.4% by 8 p.m. on Thursday, Nov. 19. It stayed close to that level through the start of premarket trading, then fell further.

For an AMD shareholder, the evening price showed the direction of the overnight move well before its full size became clear. When overnight share trading began, some perps were already close to their morning price while others had barely moved.

Median share of the move made by midnight
55%
On big moves the shares later followed
Median share made by 6 a.m.
92%
Most of the morning move was in place
Overnight decline in the AMD perp
6.5%
Loss ahead of Friday’s regular session
Indicated AMD opening price
$457.58
Implied by the overnight contract price

From evening trading to the hours before the bell

Overnight trading in some U.S. shares begins at 8 p.m. ET. By then, perps had made a median 24% of their eventual morning move, measured from each contract’s price at the regular stock market close.

For Intel (INTC$52.80-7.34%), the evening brought almost the entire decline on Thursday, Oct. 29. The Intel perp was already down 22.9% when overnight share trading began.

Circle (CRCL$50.03-9.47%) gave shareholders little indication of a big move at the same hour on Thursday, July 9. The Circle perp was up just 0.2%.

By midnight, the balance across the study had shifted. On 351 of 614 qualifying nights, the perp had covered at least half the distance from its afternoon price to where it stood the next morning.

The gap between the Intel and Circle examples persisted. At midnight the Intel perp was down 23.0% from the regular close. The Circle perp had gained 1.5% and would give back some of that gain before morning.

Premarket share trading begins at many brokers at 4 a.m. ET. The median perp had made 74% of its move by then, though the Circle perp was still up just 0.7% on its July night.

By 6 a.m. ET, the median reached 92%. For a trader holding a perp position overnight, most of the change in its value through the morning had typically come before that hour.

The Intel perp ended the test window down 25.2%. Intel shares opened 25.8% lower on Friday, Oct. 30, close to the loss already visible the evening before.

The Circle perp reached the end of its window up 12.8%. Circle shares opened 12.1% higher on Friday, July 10. Almost none of that rise had been visible when premarket trading began.

AMD trade alerts · free

An email when Advanced Micro Devices perps make a big move

When a large account opens, closes or flips a AMD position, funding changes sides, or the price moves 10% in a day.

At most one email a day. Unsubscribe in one click.

Even the same stock kept different hours

An early move was no fixed trait of a company. On Tuesday, Aug. 11, the CoreWeaveCRWV$42.09-15.88% (CRWV) perp had gained 15.7% by the start of overnight share trading.

The gain narrowed a little by midnight before widening to 18.2% by morning. CoreWeave shares opened 20.3% higher on Wednesday, Aug. 12. The evening price had already captured much of the rise facing shareholders.

CoreWeave had given a different picture on Wednesday, July 29. The CoreWeave perp rose 2.3% by evening, then surrendered most of its gain. It was up just 0.4% when premarket trading began.

By morning the CoreWeave perp had gained 7.2%. The shares opened 12.3% higher on Thursday, July 30. Even the late rise in the contract understated the change in the stock.

Both kinds of night recurred across the study. On 86 occasions, three quarters of the move was already in place by evening. On 73 occasions, less than a quarter was in place when premarket trading began.

Which nights the clock measures

The study covers 22 stocks with hourly perp prices available from exchange price data. OptimistFi compared each contract with its own 4 p.m. ET price and used Yahoo Finance for the shares’ official opens and closes.

The sample spans 115 weeknights and leaves out weekends and nights next to a U.S. market holiday. Across 2,504 observations, each stock counted separately on each night.

Shares opened far enough from their prior close to score on 1,901 occasions. The timing test narrows that group to big perp moves the shares followed. It gives no success rate for a call made the evening before.

The morning price also matters to what counts as an early move. If a perp rises sharply in the evening and gives some back, the early gain can exceed the entire move measured by morning.

Where overnight trading overlaps

In the U.S., Kalshi lists perps on individual stocks. Across the contracts in the study, a median $4.4 million traded per stock per night.

A shareholder comparing the contract with the stock faces a timing limit. By the final morning check, the shares themselves are trading at many brokers. The comparison cannot establish that the perp moved first.

What Friday’s open will settle

AMD’s decline was part of a broad overnight move ahead of trading on Friday, Nov. 20. Only 3 of 22 perps were higher by morning. In all, 18 of 22 met the study’s threshold for a big move.

The CoreWeave perp pared its decline between midnight and the start of premarket trading before falling further. By morning it was down 5.5%, pointing to an opening price of $42.46.

The MicronMU$654.23-11.14% (MU) perp followed a similar path. Its 2.9% decline at midnight narrowed before deepening to 5.0% by morning.

For AMD, the overnight decline points to $457.58. The official share price at 9:30 a.m. ET will settle whether the stock opened far enough below Thursday’s close to qualify, even if it opens above that estimate.

Make us preferred
ScoutJUST THIS STORY

LET’S MAKE THIS USEFUL · 1 OF 5

How would you like me to explain this?

Tell me what you want from this read. Start with a quick note, then adjust the details. Your other stories stay as they are.

We’ll remember your view of this story. The published article stays unchanged.

THIS STORY’S READING VIEWStandard · The whole picture

Clear explanations with context. Both sides of the evidence.