What If You Bought at the Open? Overnight vs Intraday Returns and Seasonal Timing, Tested on 293 Stocks and ETFs

We tested seven buy-and-sell rules and three seasonal-signal rules on every stock, ETF and index with reliable opening prices. Most gains came overnight, periodic rules came close to buy and hold, and the seasonal signal picked slightly better days but rarely beat holding after costs.

Does it matter when you buy and sell? We tested seven simple rules on every stock, ETF and index on SeasonalityCharts.com: buying at the open and selling at the close, buying at the close and selling at the next open, holding open to open and close to close, and buying at the start of each week, month or year and selling at its end. Then we asked whether the site’s three seasonality methods could pick better days. Here is what $10,000 would have become under each rule, using only historical prices up to November 2026.

  • Most gains came overnight, not during trading hours. Across 293 symbols with at least ten years of reliable opening prices, the median overnight return (close to next open) was +6.1% a year and the median intraday return (open to close) only +2.4% a year. For 106 symbols, trading hours alone lost money.
  • For broad funds the gap is striking: $10,000 in SPY since 1994 grew to about $162,000 held, to about $127,000 owned only overnight, and to only about $13,000 owned only during trading hours. In QQQ since 1999, trading hours alone lost three quarters of the money while the overnight hours turned $10,000 into about $550,000.
  • Buying at the start of each year, month or week and selling at its end came close to buy and hold, with the yearly version having the smallest worst-case drawdown.
  • The seasonal signal picks better days, but not by enough. On days when all three methods pointed up, judged only from earlier years, the average day was better than average for 196 of 293 symbols. But trading only those days left money in cash most of the time, and sitting out the down-signal days beat holding for only 80 of 293 symbols once a 0.05% trading cost was included.
  • The signal worked best for broad funds, industrials and financials (Microsoft, Caterpillar, the Russell 2000), and worst for fast-growing momentum stocks and commodity-driven funds (Nvidia, Tesla, gold, energy).

The seven rules, in plain English

  • Open to close (trading hours only): buy at each day’s opening price, sell at that day’s closing price. You hold nothing overnight.
  • Close to next open (overnight only): buy at each close, sell at the next morning’s open. You hold nothing during trading hours.
  • Open to next open and close to close: buy and sell at the same point each day, which is the same as simply holding.
  • Each week, month or year, open to close: buy at the first open of the period, sell at its last close, and sit in cash until the next period starts (for weeks, that means over every weekend).

Each rule starts with $10,000 and reinvests everything. Prices are split-adjusted daily open and close prices, without dividends. We only use years in which the opening price is reliable: older split-adjusted prices are rounded to the cent, which makes a $0.41 share’s open and close look identical, so each symbol starts in the first year from which at least 95% of its sessions have a genuine opening price (1994 for SPY, 2010 for Apple). Full details are on the methodology page.

Results across 293 stocks, ETFs and indexes

Every rule across 293 stocks, ETFs and indexes with 10+ years of reliable open prices (medians; each symbol from its own start date to today).
RulePer year (median)$10,000 became (median)After 0.05% per tradeWorst drawdown (median)Beat buy and hold
Close to close (buy and hold)Buy at the close, sell at the next close: the same as holding+8.7%$51,308$4,347−65%–
Open to next openBuy at the open, sell at the next session's open+8.6%$50,814$4,306−65%120 of 293
Close to next open, every dayBuy at each close, sell at the next open (overnight only)+6.1%$29,296$2,712−51%107 of 293
Open to close, every dayBuy at each open, sell at the same day's close (trading hours only)+2.4%$15,406$1,624−62%50 of 293
Each week, open to closeBuy the week's first open, sell its last close (out over weekends)+8.0%$44,815$28,054−61%145 of 293
Each month, open to closeBuy the month's first open, sell its last close+7.5%$40,333$35,852−59%32 of 293
Each year, open to closeBuy the year's first open, sell its last close+8.3%$45,506$44,918−45%69 of 293
Up-signal days, open to closeOn each day when all three seasonal methods point up: buy at that day's open, sell at its close. Otherwise in cash.+1.5%$13,459$8,034−34%29 of 293
Up-signal days, close before to open afterBuy at the close the day before an up-signal day, sell at the open the day after it. Back-to-back signal days are held as one trade.+3.5%$19,283$13,211−41%50 of 293
Hold, but sit out down-signal daysHold all the time, except: sell at the close before a day when all three methods point down, stay in cash through that day (and any down days right after it), buy back at the close of the last one.+8.6%$49,020$37,286−62%148 of 293

Two things stand out. First, the daily rules collapse once trading costs are included: at 0.05% per round trip, buying every open and selling every close turned the median $10,000 into about $1,600. Second, the periodic rules (weekly, monthly, yearly) trade rarely enough that costs barely matter, and they come close to buy and hold.

Overnight returns vs intraday returns

The chart shows SPY, the S&P 500 ETF, since 1994. Buy and hold is the thick line; the overnight line follows it closely, while the trading-hours line barely moves.

This “overnight effect” is well documented in academic research on U.S. stocks. Explanations include retail orders placed outside market hours that execute at the opening price, a reward for holding the risk of overnight news, and how news and attention arrive. It is not universal: in our data the overnight hours beat trading hours for 84% of funds and indexes and 94% of energy stocks, but for only about a third of consumer-staples and real-estate stocks.

By sector (symbols with 10+ years of reliable open prices). Median yearly return of each daily rule; share of symbols where overnight beat trading hours, where up-signal days beat the average day, and where skipping down-signal days beat buy and hold.
SectorSymbolsOvernight beat trading hoursTrading hours, per year (median)Overnight, per year (median)Up-signal days beat averageSkipping down days beat holding
Communication Services757%−0.6%+3.6%57%71%
Consumer Discretionary3566%+3.3%+10.9%74%37%
Consumer Staples2035%+4.3%+0.8%65%55%
Energy1894%−6.2%+10.2%44%50%
Financials3155%+2.8%+5.1%65%45%
Funds and indexes5784%−1.7%+7.2%79%58%
Healthcare2843%+5.8%+4.0%61%29%
Industrials2871%+0.8%+8.4%75%86%
Materials944%+4.2%+2.5%56%56%
Real Estate729%+3.8%+1.8%57%71%
Technology4648%+7.9%+8.3%65%35%
Utilities743%+1.0%+1.2%43%71%

Part of the difference is dividends. A stock’s price drops by roughly the dividend at the open on its ex-dividend date, so in price-only data the dividend shows up as an overnight loss. High-yield sectors such as consumer staples, utilities and real estate therefore look weaker overnight than they really were for a shareholder who collects the dividend.

What $10,000 became under each rule, from the first year with reliable open prices to date (price only, before costs). Green: beat buy and hold.
SymbolClose to close (buy and hold)Close to next open, every dayOpen to close, every dayEach week, open to closeEach month, open to closeEach year, open to close
SPY since 1994$162,335$126,941$12,755$105,964$104,609$130,899
QQQ since 1999$133,416$549,811$2,427$78,329$74,921$102,168
IWM since 2000$58,540$208,080$2,848$26,073$34,348$45,506
DIA since 1998$66,381$38,169$17,392$55,829$39,854$55,373
AAPL since 2010$466,322$85,107$54,864$416,725$267,460$353,637
MSFT since 1999$151,105$106,052$14,248$271,300$95,187$126,044
NVDA since 2020$271,417$146,349$22,352$459,989$239,572$300,092
KO since 2008$32,488$22,840$14,141$43,893$29,768$23,083
XOM since 2008$18,243$6,336$28,583$23,062$14,713$12,096
GLD since 2004$80,092$90,817$8,809$53,027$61,867$78,918
TLT since 2002$9,621$4,917$19,528$17,628$26,393$10,794

Can the seasonal methods pick better days?

Every symbol page measures seasonality three ways: by calendar date, by week and weekday, and by trading day of the month. For each day we asked whether all three had historically been up (or all three down) on that day, using only the years before it, so the test never uses information that was not yet available. Then we compared those days with all other days, and tested three rules:

  • Up-signal days, open to close: on each day when all three methods point up, buy at the open and sell at the close; otherwise stay in cash.
  • Up-signal days, close before to open after: buy at the close the day before an up-signal day and sell at the open the day after it, so the position also captures the two overnight moves around the day; back-to-back signal days are held as one trade.
  • Hold, but sit out down-signal days: hold all the time, except sell at the close before a day when all three methods point down, stay in cash through that day and any down-signal days right after it, and buy back at the close of the last one.

The signal does carry information. For SPY, up-signal days averaged +0.035% from open to close against +0.008% on all days and −0.013% on down-signal days. Across all 293 symbols, up-signal days beat the average day for 196, and beat down-signal days for 207. But the differences are tiny (a median +0.018% per day), and acting on them costs money: trading only the up-signal days misses most of the market’s gains, which come overnight and on ordinary days, and sitting out down-signal days needs hundreds of round trips (a median of 541).

Where it worked

  • Microsoft (MSFT), since 1999: sitting out down-signal days turned $10,000 into about $252,000 before costs and $175,000 after, against $151,000 for holding.
  • Caterpillar (CAT), since 2008: about $146,000 before costs and $107,000 after, against $85,000 held. On its down-signal days Caterpillar averaged −0.134% from open to close.
  • The Russell 2000 (IWM), since 2000: up-signal days averaged +0.046%, down-signal days −0.080%; sitting them out beat holding before costs ($83,000 vs $58,500) and roughly matched it after ($60,500).
  • QQQ: up-signal days averaged +0.064% against −0.011% on all days; sitting out down days beat holding before costs ($152,000 vs $133,000), though not after.
Seasonal-signal rules vs buy and hold, $10,000 before costs. The per-symbol pages also show results after a 0.05% cost per trade.
SymbolClose to close (buy and hold)Hold, but sit out down-signal daysUp-signal days, open to closeUp-signal days, close before to open after
MSFT since 1999$151,105$251,954$14,751$27,465
CAT since 2008$84,978$145,759$20,162$73,257
IWM since 2000$58,540$83,214$14,932$26,649
QQQ since 1999$133,416$151,865$18,295$45,844
SPY since 1994$162,335$160,266$15,774$29,945

Where it did not

  • Nvidia (NVDA), since 2020: down-signal days were actually better than up-signal days (+0.273% vs +0.177%). Sitting them out left $89,000 against $271,000 for holding.
  • Tesla (TSLA), since 2014: $171,000 sitting out down days versus $390,000 held.
  • Gold (GLD) and energy (XLE): up-signal days were slightly worse than an average day, so the calendar told us nothing useful about the next session.
  • Coca-Cola (KO): sitting out down days left $23,000 against $32,500 held, and with costs only $17,000.
Where the seasonal signal did not help: $10,000 before costs.
SymbolClose to close (buy and hold)Hold, but sit out down-signal daysUp-signal days, open to closeUp-signal days, close before to open after
NVDA since 2020$271,417$89,405$19,392$49,139
TSLA since 2014$389,830$170,616$31,811$122,964
GLD since 2004$80,092$108,389$6,108$9,870
XLE since 2004$47,175$49,020$7,225$6,892
KO since 2008$32,488$23,052$22,836$25,670

What made the difference

  • Calendar-driven assets vs news-driven ones. Broad funds, industrials and financials respond to recurring flows (month-end rebalancing, payroll and pension money at the turn of the month, tax and holiday seasons), which is exactly what the three methods measure. Fast-growing stocks such as Nvidia and Tesla move on earnings, product news and momentum; missing even a few of their big days costs far more than the signal saves.
  • Commodities follow their own drivers. Gold and oil prices respond to interest rates, the dollar, supply and geopolitics, not the stock-market calendar, and the signal showed no edge.
  • Costs decide it. Sitting out down-signal days beat holding for 148 of 293 symbols before costs but only 80 after a 0.05% cost per trade, and real-world spreads, slippage and taxes on short-term gains would cut further.
  • Dividends move gains to the overnight hours, so high-yield stocks look better intraday than overnight in price-only data.

How to use this

These are historical, hypothetical results, not a trading system. The clearest lessons are about when returns have arrived (overnight and on ordinary days, not in the trading session) and how quickly frequent trading eats into them. Seasonal signals are best read as context: a reason to be patient, or to time an entry you already planned, rather than a reason to trade every day. Every symbol page now has a What if section with these rules for that symbol, including results after costs. For longer rhythms, see our study of stock market cycles, and for the calendar pattern of the whole market, which stocks follow the S&P 500’s seasonality.

Frequently asked questions

What is the overnight effect in stocks?

The overnight effect is the tendency of U.S. stocks and funds to earn most of their long-run gains between the close and the next open rather than during trading hours. In our data, $10,000 in SPY since 1994 became about $127,000 owned only overnight and about $13,000 owned only from open to close.

Is it better to buy at the open or at the close?

Historically, buying at the close and selling at the next open captured more of the gains than buying at the open and selling at the close for 61% of the 293 symbols we tested. Both daily rules lose most of their gains once trading costs are included, and buy and hold beat both.

Does buying at the start of the month and selling at the end work?

It came close to buy and hold for most symbols (median +7.5% a year vs +8.7%) with few trades, but it beat holding for only 32 of 293 symbols because it misses the moves between the last close of one month and the first open of the next.

Can seasonality predict which days will be up?

Only slightly. Days when all three seasonal methods pointed up were better than the average day for about two thirds of symbols, but by a median of only 0.018% per day, too little to beat holding once trading costs are included for most symbols.

Why do dividend stocks look worse overnight?

Because a stock opens lower by about the dividend on its ex-dividend date. In price-only data that drop appears as an overnight loss, even though a shareholder receives the dividend in cash.

Are these results investment advice?

No. They are hypothetical backtests on historical prices without dividends, taxes or realistic trading costs beyond a 0.05% example. Past results do not predict future results.

Figures as of mid-November 2026 from daily open and close prices; the tables above update automatically. Hypothetical results, price only, before taxes; not investment advice.