Methodology & risk note: Backtested event study of SPY daily price data, 1993-2026 (8,397 big range days), frictionless. Results are hypothetical and not investment advice, past patterns don't guarantee future results. Full method and disclaimer below.
TL;DR
- "Big" isn't one number. A big range day can be measured three ways, in raw index points, as a percent of price, or against the market's own recent volatility (ATR). The three rulers disagree about which days even count as big, so picking the right one is the whole game.
- Raw points lie. A fixed-dollar definition of "big" just flags the recent past. Of the 635 days with a range of at least 6 points, 549 happened in the 2020s, because SPY climbed from $44.34 to $741.75. A points ruler mistakes a higher price for a wilder market.
- The honest signal is what comes next, and it's volatility, not direction. The bigger the day relative to its own recent range, the wilder the next day. Next-day range climbs steadily from 0.715x the recent average on the calmest days to 1.928x on the most extreme, a 2.7x spread.
- And it lingers. After the biggest days, the next day runs about 1.3x the normal range, and it's still elevated at 1.13x twenty days later. Wild begets wild for weeks, not hours.
- The direction tells are smaller, but real. A big down day bounces harder than a big up day continues (+0.305% vs +0.104% over five days), and where the day closes is a read: closing near the low leads (+0.117% next day), closing near the high leads to nothing (−0.006%).
A massive day just printed. Now what?
You know the feeling. The S&P 500 has one of those days, a wide, violent range, the kind that fills the financial news and lights up every group chat. And the instinct that follows is always about direction. If it ripped higher, people ask whether the rally has legs. If it cratered, they ask whether to buy the dip or run for the exit. The day's size is treated as a clue about which way the market goes from here.
That instinct is aimed at the wrong target. We measured every trading day in the S&P 500 since 1993, 8,397 sessions in SPY's daily price history, and asked what actually follows a big range day. The cleanest answer has almost nothing to do with up or down. A big day's most reliable message is about volatility: after days of at least 1.5x ATR, average subsequent range stayed elevated through twenty sessions. Big days come in clusters.
But before any of that, there's a trap to step around. The very first thing you have to decide is what "big" even means, and the obvious way of measuring it will fool you. We'll take it one question at a time.
What counts as a "big range day"?
The range of a day is just the distance between its high and its low, how much ground price covered between the session's extremes. A big range day is one of the wide ones. Simple enough. The hard part is the word "wide," because wide compared to what?
Here's the dataset, stated plainly so you know exactly what's behind every number below:
- Instrument: SPY, the S&P 500 ETF, daily bars.
- Window: February 1993 through June 2026, 33 years.
- Events: 8,397 big range days detected and measured.
- Costs: frictionless. No commission or slippage. The magnitudes here are relative, not what you'd net after trading costs.
Every day gets sized three different ways and tagged with what came next, the average return over the following 1 and 5 days, how wide the next day's range was, and where price closed inside the day's own range. This is measurement, not a trading system: it tells you the base rates, and you decide what to do with them.
Measure a big day three ways
"Big" depends entirely on the ruler. A 6-point range was a screaming move when SPY traded at $44.34 in February 1993 and a quiet afternoon at $741.75 in June 2026. So we sized every day three ways and compared what each one revealed:
- In raw index points, the high-minus-low distance in dollars. The way a lot of people instinctively think about it.
- As a percent of price, the same range scaled by where the market is trading. A 6-point day is 13.5% of a $44.34 SPY and 0.8% of a $741.75 SPY.
- Relative to the market's own recent range (ATR), sizing the day against how much SPY has been moving lately. ATR, or average true range, is a standard gauge of recent daily movement. A day "worth 2x ATR" is twice as wide as a normal recent session.
Running all three side by side is the honest way to do this. If a pattern only shows up under one definition, it's probably an artifact. If it holds across all three, it's real. And the comparison turns up something genuinely useful right away: one of these rulers is broken.
Why raw points lie
Watch what happens when you define "big" in raw index points. A points ruler doesn't flag the wildest days in history, it mostly flags the most recent ones.
| Decade | Big days by Points (≥6 pts) | Big days by Price% (≥2.5%) | Big days by ATR (≥2x) |
|---|---|---|---|
| 1990s | 7 | 101 | 61 |
| 2000s | 30 | 351 | 50 |
| 2010s | 49 | 79 | 85 |
| 2020s | 549 | 119 | 57 |
| Total | 635 | 650 | 253 |
Look at the points column. Of the 635 days with a range of at least 6 points across the entire 33 years, 549 of them, 87%, happened in the 2020s. That is not because the market suddenly got wilder. It's because SPY climbed from $44.34 to $741.75. At 16.7 times the price, a "6-point move" is a sixteenth as meaningful, so the points ruler quietly redefines half its "big" days as whatever happened most recently. A genuinely violent day from 1998 doesn't even make the list.
The percent-of-price ruler tells the opposite story, its big days cluster in the 2000s (351 of 650), the era of the dot-com bust and the financial crisis, when the market really was tearing itself apart in percentage terms. And the ATR ruler is the most evenly spread of all (61 / 50 / 85 / 57 across the four decades), because it re-grades every day against its own moment in time. A wild day in 1998 and a wild day in 2025 both register as wild, regardless of the price level or the era.
That even spread is the tell. The ATR ruler is the only one that isn't secretly measuring when a day happened instead of how wild it was. So that's the ruler we trust for the real question, what comes next.
The headline: a big day predicts tomorrow's volatility
This is the finding the whole study turns on, and it has nothing to do with direction. Size the day against its own recent volatility (ATR), then look at how wide the next day's range turns out to be, measured the same way, against the recent average. The relationship is steep and it's clean.
| Day's size (vs ATR) | Events | Next-day range (vs recent avg) |
|---|---|---|
| Under 0.5x | 1,000 | 0.715x |
| 0.5-1x | 4,513 | 0.886x |
| 1-1.5x | 2,067 | 1.016x |
| 1.5-2x | 544 | 1.177x |
| 2-2.5x | 164 | 1.422x |
| 2.5-3x | 49 | 1.703x |
| Over 3x | 40 | 1.928x |
Read that last column top to bottom. After the calmest days, those under half an ATR, the next day runs only 0.715x the recent average range, noticeably quieter than usual. After the most extreme days, over 3x ATR, the next day runs 1.928x, nearly double the normal range. That's a 2.7x spread, and it climbs in clean order with no zig-zag. A value above 1.0 means tomorrow was wider than the recent average; below 1.0 means calmer.
This is volatility clustering, measured here on daily SPY bars from 1993 to 2026: big moves are followed by big moves, calm by calm. Robert Engle's 1982 paper on autoregressive conditional heteroskedasticity in Econometrica, and the GARCH literature that followed it, is where the general claim comes from. This study measures it on one ETF. The market doesn't reset overnight. A violent session leaves the next day's range elevated, because whatever was driving the violence, whether news or positioning or fear, doesn't evaporate at the closing bell. Notice what this is not: it isn't telling you the market will go up or down. It's telling you to expect another wide day. That's a statement about size, not sign, and it's the most reliable thing a big day tells you.
It decays slowly
A fair objection: maybe the cluster is just a one-day echo, tomorrow is wild, then everything snaps back to normal. The data says no. The elevated volatility fades, but it fades slowly, and it's still measurable weeks later.
Here's the next-day-style range ratio tracked across horizons, comparing the biggest days (at least 1.5x ATR) against the calmest (under 1.0x ATR):
| Days later | After a big day (≥1.5x ATR) | After a calm day (<1.0x ATR) |
|---|---|---|
| 1 | 1.297x | 0.855x |
| 2 | 1.26x | 0.86x |
| 3 | 1.234x | 0.87x |
| 5 | 1.202x | 0.887x |
| 10 | 1.2x | 0.908x |
| 20 | 1.128x | 0.934x |
The day right after a big one runs 1.297x the normal range. Five days out it's still 1.202x, and even twenty trading days later, a full month of sessions, it's running 1.128x, still meaningfully above normal. The calm days do the mirror-image thing: they start quiet at 0.855x and slowly drift back up toward normal as the unusual calm wears off.
So the cluster isn't a one-day flicker. After days of at least 1.5x ATR, the range ratio stayed above 1.0 through twenty sessions. The measured follow-through is weeks of wider days, not a one-session spike.
Does size predict returns too? Only at the extremes
Size predicts tomorrow's range beautifully. Does it predict tomorrow's return? Mostly no, and the exception is worth stating precisely, because it lives entirely in the tails.
Walk up the ATR ladder and the average next-day return does not climb with size: −0.019% for the calmest days, then +0.050%, +0.037%, +0.032%, +0.043% through the middle buckets. The two most extreme buckets sit on different horizons and sit below the 50-event floor: 2.5 to 3x ATR averaged +0.517% the next day (49 events), and days over 3x ATR averaged +1.071% over the next five (40 events). The percent lens says the same thing: every bucket below 2.5% shows five-day drift of 0.204% or less, while days of 4% or more, 150 events in 33 years, averaged +0.456% the next day and +0.980% over five.
Ordinary big days, even days most traders would call dramatic, carry essentially no return edge. The rare tail prints a real upward drift, at those counts and those two horizons. That lines up with the down-day bounce in the next section, where Down events carry the stronger drift. Nothing here classifies those extreme days as sell-offs or whipsaws, so panic and rebound is a reading of the result and not a tested mechanism.
The shape, flat middle, jump at the tail, is the finding. The exact decimals of any one extreme cell will wobble on fresh data.
The direction tells are smaller, but they're real
Everything so far has been about size, because size is where the strong signal lives. But there are two genuine direction tells hiding underneath, and they're worth knowing precisely because they cut against the popular instinct.
The first: a big down day bounces harder than a big up day continues.
| Direction | Events | Avg next-day return | Avg 5-day return | Hit +1% within 20 days |
|---|---|---|---|---|
| Down | 3,836 | +0.069% | +0.305% | 88.5% |
| Up | 4,561 | +0.016% | +0.104% | 82.9% |
A big down day is followed by +0.305% over the next five days. A big up day by only +0.104%, roughly three times less. And a big down day trades at least 1% above its close at some point within the next 20 sessions 88.5% of the time, versus 82.9% for an up day. Both numbers point the same way, up. Down events exceeded Up events, and both were positive in this sample. But the down days bounce harder, which is the opposite of the "a big drop means weakness, get out" reflex.
Where it closes is the read
The second direction tell is sharper than the first, and it's about where inside the day's range price finished. Did the market close near the low (sellers in control into the bell), in the middle, or near the high (buyers in control)? Split every big day into those three thirds and the follow-through separates cleanly.
| Closed in the day's... | Events | Avg next-day return | Avg 5-day return | Next-day range (vs avg) |
|---|---|---|---|---|
| Low third | 2,485 | +0.117% | +0.351% | 1.037x |
| Mid third | 2,385 | +0.029% | +0.212% | 0.945x |
| High third | 3,527 | −0.006% | +0.074% | 0.860x |
Closing near the low leads on every dimension. It produces the strongest next-day drift (+0.117%), the strongest five-day drift (+0.351%), and the widest next day (1.037x the recent average). Closing near the high does the opposite, essentially flat the next day (−0.006%), the weakest five-day drift, and the calmest follow-through (0.860x).
This lines up with the down-day finding and sharpens it. A big day that closes near its low is a day the sellers won into the close, and that's exactly the setup that tends to snap back hardest, with a wider, more active next day. A day that closes near its high has already spent its energy. The next day tends to be quieter, with average next-day drift near zero. So if you only remember one direction read from this study, make it this one: the close location is the tell. Near the low leads. Near the high lags.
What about actually trading it?
Everything above is measurement of what big days do. The natural follow-up is what happens when you try to trade them, long versus short. To check, we ran a backtest sweep of 2,736 variants across the three size lenses, both directions, and a range of holding periods, then looked at risk-adjusted return (return relative to worst drawdown). One result is blunt: the long side's median beats the short side's in all three lenses. Neither clears the random control's spread, so this is a ranking of the two sides, not proof that either has beaten luck.
Across the reliable variants (those with at least 50 trades), median CAR/MaxDD on the long side is +0.010 in the ATR and percent-of-price lenses and +0.020 in points, and −0.020 on the short side in all three. Down events averaged +0.305% over five days against up events' +0.104%, and the short-side median CAR/MaxDD is negative in all three lenses. The long side's medians beat the short side's in all three segments, but all three sit below the random control's 0.025, inside its seed-to-seed spread of 0.032. That control enters SPY on random days, as often and for as long as the study's median variant. On return against drawdown, trading big range days long did not beat it. Turning any of this into a real strategy is a separate, careful piece of work, not something this measurement study hands you ready-made.
The honest caveats
None of this is a finished trading system, and it would be dishonest to present it as one. Three things to keep front of mind:
- Frictionless. Every number here is computed with no commission and no slippage. The magnitudes are relative, useful for comparing one bucket against another, not what you'd actually net after costs. On days under 2.5% of price, the five-day drift runs from +0.105% to +0.204%, and those are gross figures. Set your own round-trip cost against them before you read any of them as a return you'd keep.
- An event study, not a system. This measures base rates. The events overlap (we're cataloguing what big days do, not trading them sequentially), and the backtest sweep is a flat-only measurement tool, not a tuned strategy. The job here is to replace gut-feel myths with measured odds, turning those odds into a robust strategy is separate, careful work.
- Thin cells are flagged, not trusted. The most extreme buckets are small, the over-3x ATR bucket holds just 40 events, and the 2.5-3x bucket only 49. We flag these rather than drop them, because they carry the cleanest signal, but their exact numbers can wobble on fresh data. Lean on the shape of the climb, calm-to-wild, low-close-leads, not the third decimal place of any one extreme cell.
A companion study on what the direction of an S&P 500 opening gap really tells you publishes the following week. It walks through gap size and fill rates the same way this one walks through big days.
The findings at a glance
| Finding | The number | What it means for a trader |
|---|---|---|
| Three rulers | 549 of 635 "big" point-days are in the 2020s | Don't define "big" in raw points, it just flags recent years. Use ATR. |
| The headline | next-day range climbs 0.715x → 1.928x by ATR bucket | A big day predicts tomorrow's volatility, not its direction. |
| Persistence | still 1.128x elevated 20 days after the biggest days | The cluster lasts weeks, not hours. |
| Direction | down days +0.305% / up days +0.104% over 5 days | A big down day bounces harder than a big up day continues. |
| Close location | low close +0.117% / high close −0.006% next day | Where it closes is the read, near the low leads, near the high lags. |
Key takeaways
- "Big" depends on the ruler. Raw points lie. 549 of 635 big point-days fall in the 2020s, as SPY's price rose 16.7 times. ATR is the honest measure because it grades each day against its own volatility. Percent-of-price sits in between.
- A big day predicts volatility, not direction. The next day's range climbs cleanly from 0.715x the recent average on the calmest days to 1.928x on the most extreme. Big days cluster.
- And the cluster lasts. After the biggest days the next session runs ~1.3x normal, and it's still elevated at 1.128x a full month later. Plan for weeks, not a spike.
- The direction tells are real but secondary. A big down day bounces harder than a big up day continues (+0.305% vs +0.104% over five days), and the close location sharpens it, near the low leads (+0.117% next day), near the high lags (−0.006%).
- If you trade it, stay long. Median CAR/MaxDD is +0.010, +0.020 and +0.010 on the long side across the ATR, points and percent lenses, and −0.020 on the short side in all three. Shorting big days fights SPY's structural drift. Even the long medians sit below a random-entry control's 0.025, so long is the better of two sides, not an edge over luck.
This is one study in an ongoing series. If you want the next one, the same kind of large-scale, myth-busting, frictionless-but-honest backtest, join The Overfit newsletter at StatOasis.com/Overfit. It's where we publish the data behind the trading ideas everyone argues about.
Methodology: how this data was generated
This is a backtested event study, not live trading results. These numbers come from our own event-study research engine, which we built in-house and run over the full SPY price history, not from third-party summaries or reproduced figures. Here's exactly how it was built, in plain terms.
- Data source: Daily OHLCV (open, high, low, close, volume) price data for SPY, the S&P 500 ETF.
- Date range: February 1993 through June 2026, 33 years of daily bars.
- What an "event" is: A big range day, a session whose high-to-low range is large. All 8,397 days were detected and sized three ways: raw index points, percent of price, and relative to ATR (average true range, a standard volatility gauge). Each was also tagged by direction (up / down) and by where price closed inside its own range (low / mid / high third).
- Forward-outcome measurement: For each day, we recorded the average return over the following 1 and 5 trading days, and how wide subsequent days' ranges were relative to the recent average (the "range ratio" used throughout). This is measurement of base rates, not a trading system, events are allowed to overlap.
- The trading sweep: Separately, we ran a flat-only backtest of 2,736 variants (the three size lenses x direction x holding periods), starting from $35,000 in capital, ranked by risk-adjusted return (return relative to maximum drawdown), to check which side actually pays.
- Frictionless assumption: Results are computed without commissions or slippage. Real-world trading costs would reduce any edge shown here. Treat these as indicative base rates, not net-of-cost returns.
- Reliability: Findings rest on large samples drawn from 33 years of data. Any segment with fewer than 50 events is treated as low-reliability and flagged rather than trusted; the ATRx buckets behind the headline volatility-clustering finding hold between 40 and 4,513 events, and the two most extreme, at 49 and 40 events, sit below that floor and are flagged.
For more on why even a clean backtest is not the same as a live edge, and how costs, slippage, and overfitting eat into hypothetical results, watch Why Most Profitable Backtests Fail in Live Trading on the StatOasis YouTube channel.
Disclaimer
All results above are derived from historical backtesting using daily SPY price data and do not represent actual trading results. Backtested performance is hypothetical. Past performance of any pattern does not guarantee future results. This page is for educational and informational purposes only and does not constitute investment advice. StatOasis is not a registered investment advisor. Nothing here is a recommendation to buy or sell any security. Please consult a licensed financial professional before making any investment decision.
Transparency: StatOasis sells trading-education products, some of which are linked above. Our research is produced independently and is not altered to favor a sale.
Freshness: The data is current through 2026. We re-review these studies when the underlying dataset is extended.
