What Is a Pullback in Stocks? 67 S&P 500 Dips Since 1950, and How Deep They Went

Michael Whitman · Trading since 2017, full-time since 2020 · YouTube creator, 137K subscribers · Published 2026-09-05 · Updated 2026-11-02 · Reviewed 2026-11-01

Covered on this page: TradingView.

A pullback is a stretch where price moves against its trend for a while and then the trend carries on. In a rising stock it is a dip; in a falling one it is a bounce. The catch is in the last four words: whether the trend carries on is only known later.

Most explanations stop at the definition and a cartoon. This page counts the real thing.

The usual rule of thumb says a stock-market pullback is a 5% to 10% fall, so the data section takes every time the S&P 500 closed 5% below a record since 1950 and shows how many stopped there, how long they took, and how two recent ones looked identical at the start.

How it forms

A trend pauses for ordinary reasons. Buyers from lower down sell some of their gains. Newer buyers decline to pay the top price and wait. Neither group thinks the trend is over, yet together they are enough to push price back for a few days or a few weeks.

What separates it from a turn is what stays intact. In an uptrend, the low of the dip should hold above the previous low, so the staircase of higher lows that market structure describes is still there when buying resumes.

Traders often expect the dip to slow near something they could see beforehand: a prior high, a support level, or a moving average.

Activity often thins out on the way down. Shrinking volume during the dip is read as sellers being absent rather than buyers being scared. That is a reading, not a rule, and plenty of real turns begin quietly too.

Two meanings of one word

On a chart, a pullback has no size. A day trader calls a 0.3% dip on a five-minute chart a pullback; a swing trader uses the same word for a three-week slide on a daily chart. The test is structural: did the move leave the trend’s last low standing?

In market news, the word has a size. Reporters use pullback for a fall of roughly 5% to 10% from a high, correction for 10% or more, and bear market for 20% or more.

None of those thresholds does anything. They are round numbers that make headlines comparable, and every deeper fall passed through each one on the way down.

This page uses both. The worked example and the data use the news version, because it is the one people search for and the one that can be counted on 76 years of index closes.

A worked example: the S&P 500 in early 2026

The record came first. The S&P 500 closed at 6,978.60 on 27 January 2026. That close is the reference point: every percentage below is measured from it, using daily closes only.

The slide was slow. It took 35 trading sessions for the index to close 5% below the record, which it did on 18 March at 6,624.70. By the news convention, the 2026 dip had become a pullback.

Then it kept going, for a while. The lowest close came on 30 March at 6,343.72, which is 9.10% under the record, 43 sessions after it. The 10% line sat at 6,280.74, so the index stopped 62.98 points short of being called a correction.

On 15 April it closed at 7,022.95, a new record, 54 sessions after the old one.

Now compare 2025. The index closed at a record 6,144.15 on 19 February 2025 and reached the 5% mark much faster, on 4 March at 5,778.15, nine sessions later.

On 13 March it closed 10.13% down at 5,521.52. The lowest close was 4,982.77 on 8 April, 18.90% below the record, and the next record did not come until 27 June, 89 sessions after the old one.

At the 5% mark the two could not be told apart by depth. Both had just crossed the same line. From that close, 2025 fell another 13.77% before its low, and 2026 fell another 4.24%. One stopped 1.10 percentage points short of the 20% bear-market line; the other is remembered, if at all, as a dip.

What a trader could do with that was place a line, not make a call. Someone treating the 2026 dip as a pullback in a rising market needed a level where the idea would be wrong, decided before the low, and a position size that made being wrong affordable.

That is risk management, and it is the same whether the dip turns at 6% or 19%. The step-by-step version, with triggers and stops, is on how to trade a pullback.

The original data

The rule, exactly as the code applies it: a record is the S&P 500’s highest daily close so far, price only, with history from December 1927 so the 1929 peak still counts in 1950. A decline is counted on the first close at least 5% below a record, from 3 January 1950 to 30 October 2026, and it ends on the first close above that record.

Its depth is the lowest close in between. Sessions are trading days counted from the record close.

That finds 67 declines, and all 67 have recovered. The live correction page counts 68 to 18 September 2026; the difference is the fall from the 1929 peak that was already under way when 1950 began, which it includes and this count leaves out.

Most 5% falls were pullbacks in the news sense, and most were short. For the 43 that stopped short of 10%, the median extra fall after the first 5% close was 0.50%, and the deepest of them reached 9.76%. The deeper groups did not just go further; they took several times as long to get back.

A common belief is that a fast drop warns of a deep one. The median decline took 14 sessions to reach 5%. Of the 35 that got there in 14 sessions or fewer, 15 (42.9%) went on to 10% or more; of the 32 slower ones, 9 (28.1%) did.

That looks like a gap, but with groups this small a Fisher exact test gives p = 0.31, which is well within chance. The 2026 dip was slow and stayed shallow, the 2025 fall was fast and went deep, and the data says that pairing is not a reliable pattern.

The check uses a second price series. Running the same rule on SPY’s daily closes from 29 January 1993 finds 34 declines, 22 of which stopped short of 10%; the index has 31 in that window, 20 of them shallow.

Of the 34 SPY declines, 31 fall inside one of the index’s, and 29 of those 31 land in the same depth group. The two that differ sat right on a line: in 1997 SPY fell 10.38% against the index’s 9.63%, and in 2018 it fell 20.18% against 19.78%. The other three are SPY dips of about 5.1% that the index did not quite reach.

Read the distribution as a base rate, not a forecast. Sixteen of the 67 stopped between 5% and 6%, 10 between 6% and 7% and 7 between 7% and 8%, but eight stopped between 9% and 10% against two between 8% and 9%. Past the 10% line, 9 stopped between 10% and 15% and 4 between 15% and 20%.

Counts that small move around with a single decline, so no level in that range is special. The full list of all 67 gives each one’s dates, closes and session counts.

What traders watch while it is happening

The previous low is the one hard line. A dip that holds above it is still a pullback on the structure test; a close below it means the higher lows have stopped, and the reversal page is the one that applies.

Depth against the trend’s own history is the next. A pullback much deeper than the earlier ones in the same trend is new behavior, and the retracement page covers measuring it. Fixed fractions such as 38.2% and 61.8% come from Fibonacci and are a description of depth rather than a reason for price to stop.

The level, marked beforehand, is the third. It is what gives a stop somewhere to sit and turns a dip from a feeling into a price.

When it fails

It was the start of something bigger. Eleven of the 67 index declines went past 20%, and at their first 5% close they were the same kind of event as the 43 that did not. The median for those 11 was 513 sessions, roughly two years, before a new record. Nothing at the 5% mark separated them.

There was no trend to pull back from. In a sideways market every drop is a dip from the last high, and buying it is buying one side of a range. The structure test needs a trend before the word means anything.

The stop sat inside the dip. A pullback can run a few percent further than expected and still be a pullback: the 2026 dip went another 4.24% after its 5% close. A stop placed inside that normal noise gets taken by a move that then recovers.

The 10% line was treated as a wall. It is a label. The 2026 low sat 62.98 points above it and the 2025 fall went straight through it on its way to 18.90%.

Questions people ask

How much is a pullback in stocks?

In market news, usually a fall of about 5% to 10% from a recent high. On a chart it has no fixed size; it is any move against the trend that leaves the trend’s last low intact. Since 1950, 43 of the 67 S&P 500 falls of 5% from a record stopped short of 10%.

How long does a pullback last?

For S&P 500 falls of 5% to under 10% from a record, the median was 19 trading sessions from the record to the lowest close and 48 sessions to a new record. Individual ones ranged widely; the 2026 dip took 54 sessions record to record.

What is the difference between a pullback and a correction?

Size, by convention. A correction is a fall of 10% or more from a high; a pullback is the smaller version. Neither label tells you what happens next, because every correction was a pullback first.

Is a pullback a good time to buy?

It is a lower price, not a cheaper asset. Whether it was a good entry depends on the trend resuming, which is only known afterward, so the useful question is where you would be wrong and how much that costs, decided before you buy.

Related

Market structure is what makes a dip a pullback at all: no trend, no pullback. Retracement measures how far one came back, and reversals covers the point where a pullback turns out to be the end of the trend.

For the procedure, how to trade a pullback walks through the entry, trigger and stop.

This page is educational, not financial advice. Test every idea on your own charts before risking money. Some links on this page earn a commission if you buy through them. It costs you nothing and it does not decide what appears here or in what order — how these pages are made is set out in our methodology.