Plutux
Free tools

Market statistics

How often the S&P 500 sets a record high — and what happened next

About one session in seventeen has closed at an all-time high, and buying at one has not been the mistake it feels like: the year after a record close has averaged very slightly more than the year after an ordinary one, and finished positive slightly more often.

1,444 record closes · data through 28 Aug 2026

19282026

Record closes
1,444
5.8% of all sessions
Years with a record
48 of 99
busiest was 1995, with 77
Median gap
3 days
average 25 days
Longest drought
25 years
from 16 Sep 1929
Last record
13 Aug 2026
15 days ago · -1.12% from it
What the index did next, from a record close and from any session. Both columns are measured over the same series and the same horizons, so the pair is a like-for-like comparison rather than a figure against a remembered average.
HorizonFrom a record: averageFrom a record: medianFrom a record: positiveFrom any day: averageFrom any day: positive
1 month1,441 obs+0.39%+0.71%60.1%+0.67%60.3%
3 months1,439 obs+1.64%+2.46%69.0%+2.01%64.0%
6 months1,421 obs+3.83%+5.26%72.7%+3.98%67.2%
1 year1,397 obs+8.24%+9.81%70.9%+8.20%69.8%
Where the index closed relative to its own running high, counted in sessions.
Distance from the highSessionsShare
At a record high1,4625.9%
Within 1% of one2,47810.0%
1% to 5% below5,00120.2%
5% to 10% below2,94911.9%
10% to 20% below3,85615.6%
More than 20% below9,03536.5%
The longest stretches without a new record close, longest first.
FromToYearsSessions
16 Sep 192922 Sep 195425.06,249
11 Jan 197317 Jul 19807.51,898
24 Mar 200030 May 20077.21,803
9 Oct 200728 Mar 20135.51,376
29 Nov 19686 Mar 19723.3820
2 Aug 195624 Sep 19582.1540
3 Jan 202219 Jan 20242.0513
28 Nov 19803 Nov 19821.9488
25 Aug 198726 Jul 19891.9485
12 Dec 19613 Sep 19631.7434
Record closes by calendar year, newest first. A year with none is not a bad year — it is a year that started below a previous high.
YearRecordsSessionsShare of the year
20262716516.4%
20253925015.6%
20245725222.6%
20230250
202212510.4%
20217025227.8%
20203325313.0%
20193525213.9%
2018192517.6%
20176225124.7%
2016182527.1%
2015102524.0%
20145325221.0%
20134525217.9%
20120250
20110252
20100252
20090252
20080253
200792513.6%
20060251
20050252
20040252
20030252
20020252
20010248
200042521.6%
19993525213.9%
19984725218.7%
19974525317.8%
19963925415.4%
19957725230.6%
199452522.0%
1993162536.3%
1992182547.1%
1991222538.7%
199062532.4%
1989132525.2%
19880253
19874725318.6%
19863125312.3%
19854325217.1%
19840253
19832925311.5%
198222530.8%
19810253
1980242539.5%
19790253
19780252
19770252
19760253
19750253
19740253
197332521.2%
19723225112.7%
19710253
19700254
19690250
19683322614.6%
1967142515.6%
196692523.6%
19653725214.7%
19646225324.5%
1963112514.4%
19620252
19613925015.6%
19600252
19592725310.7%
1958222528.7%
19570252
1956152516.0%
19554125216.3%
1954222528.7%
19530251
19520250
19510249
19500249
19490250
19480251
19470250
19460250
19450246
19440251
19430250
19420251
19410250
19400251
19390249
19380250
19370250
19360253
19350250
19340249
19330242
19320250
19310252
19300251
19294324917.3%
19285325021.2%

Price returns from daily closing levels. Index price returns exclude dividends. Full provenance, method and a citation line are in Sources and method below.

A record is a common event that feels like a rare one

There have been 1,444 record closes in 24,781 sessions — about one in 17 — and they arrived in 48 of 99 calendar years. The median gap between consecutive records is 3 days and the average is 25, which is the shape of the whole phenomenon in two numbers: records come in clusters separated by long droughts, so the typical gap is short and the average is dragged out by a handful of enormous ones.

The forward-return table is the reason this page exists. Over the year following a record close the index averaged +8.24% and finished higher 70.9% of the time; over the year following any session at all it averaged +8.20% and finished higher 69.8% of the time. Whatever a record high is, it has not been a signal to wait.

The reason that feels wrong is survivorship of memory. Two records in this record were followed by catastrophe — September 1929 and March 2000 — and both are famous. The other 1,442 are unmemorable, which is exactly the problem — a base rate assembled from the two instances you can name is not a base rate.

The distance bands underneath say the other half of it. The index has spent 52% of all sessions more than 10% below its previous high, and the longest single drought — 16 Sep 1929 to 22 Sep 1954 — ran 25 years. Records are ordinary, and so is spending years without one.

Questions people ask about this

Is it a bad idea to buy at an all-time high?
Not on this record. Measured over the following month, quarter, half-year and year, an entry at a record close has done about as well as an entry on a randomly chosen day, and has finished positive slightly more often. The intuition that a high is dangerous comes from the two occasions it was — 1929 and 2000 — which are memorable precisely because they are rare.
Why does the index spend so little time at a high?
Because records cluster. A market at a high tends to set several more within weeks, and then goes years without one: the gap between the 1929 peak and the next record ran to twenty-five years, and the gap after 2000 to seven. The median gap between consecutive records is a few days and the average is several weeks, which is the same fact stated twice.
Would using intraday highs change these numbers?
It would raise the count. Every figure here is built from closing levels, so an intraday high that the index gave back before the bell is not a record on this page. Closing levels are what the index is officially recorded at, and mixing the two conventions is how two sources end up disagreeing about how many records a year contained.

Sources and method

Data
  • Financial Modeling PrepDaily adjusted closing levels and quotes, retrieved through Plutux's own data service.
  • S&P Dow Jones IndicesPublisher and methodology owner of the S&P 500 index itself.
How it was calculated
A record is a close strictly above every close before it in the series, so the first bar is not one and a repeat of a previous high is not a new record. The distance bands count sessions by how far the close sat below the running high. Forward returns are measured in trading sessions — 21 for a month, 252 for a year — from a record close, and compared with the same horizons measured from every session for which that horizon exists, so the conditional figure and its baseline are drawn from the same window. Closing levels, price returns, dividends excluded.
How often it changes
Regenerated from the full daily history about once a year; the date it runs through is at the top of the page.
Citing this page

Free to quote — please link rather than copy the table.

Plutux. "How often the S&P 500 sets a record high — and what happened next." Data through 28 Aug 2026. https://plutux.ai/resources/tools/stock-market-record-highs

Historical figures for information only — not investment advice, and not a forecast.

Plutux is not an investment adviser. Market data and AI-generated analysis are for information and education only, not investment advice. Disclaimer

© Plutux Technology Limited 2026
S&P 500 Record Highs: Frequency, Droughts and Forward Returns | Plutux