What Is a Downtrend? Lower Highs, Lower Lows and 32 Years of SPY Swings

Michael Whitman · Trading since 2017, full-time since 2020 · YouTube creator, 137K subscribers · Published 2026-09-06 · Updated 2026-11-02 · Reviewed 2026-11-01

Covered on this page: TradingView.

A downtrend is a market making lower highs and lower lows, one after the other: every bounce stalls under the previous bounce, and every drop digs below the previous low. Both halves have to be there. A falling chart with one half missing is something weaker, and the difference shows up in the 2022 SPY chart below.

Most definitions stop at that sentence and add a cartoon zigzag. This page uses real swings instead: SPY’s 2022 decline marked point by point, the exact price that would have ended the downtrend, and 32 years of daily data on how often the strict rule is actually met and what followed.

How it works: two sequences, read together

Start with swing points, not candles. A swing high is a peak that price later falls away from by a set amount; a swing low is a trough that price later climbs away from by the same amount. On this page that amount is 5%, measured on daily closes.

Then compare each swing with the one of the same kind before it. The highs form one list and the lows another. A downtrend is the state where the latest high is below the high before it and the latest low is below the low before it.

The filter is a choice, and it changes the answer. A 2% filter on the same chart marks many more swings, so the sequence breaks and re-forms more often; a 10% filter marks only the big turns. Neither is wrong. What matters is stating which one you used before calling a market down.

Every swing is confirmed late. A peak only becomes a swing high once price has dropped 5% from it, so the reading always trails the chart. The 3 Jan 2022 high in SPY was not confirmed until 19 Jan. That lag is built in, not a flaw in the method.

A worked example: SPY through 2022

SPY closed at $477.71 on 3 Jan 2022 and at $382.43 on 30 Dec, a fall of 19.9%. Almost anyone would call that year a downtrend. The swing-by-swing reading is less tidy, and the reasons are instructive.

Three separate downtrend stretches came out of one falling year. Counting the session on which each swing was confirmed, the strict reading held from 17 Mar to 11 Apr, from 10 Jun to 25 Aug and from 4 Oct to 15 Dec: 123 of the year’s 251 sessions.

The rest of the time one list was falling and the other was not.

The March rally is what broke the first stretch. SPY climbed 10.9% from the 8 Mar low to the 29 Mar high and finished above the February peak. Nothing about the bigger picture had changed, yet by the rule the sequence was no longer intact.

The summer rally was larger. From $365.86 on 17 Jun to $429.70 on 16 Aug, SPY rose 17.4% while a downtrend was in force, and it went through the previous lower high on the way.

The exit level in October was specific. Once the 4 Oct high at $377.97 was confirmed, that close was the line: a close above it means a lower high has been exceeded. SPY closed at $378.87 on 24 Oct, eight sessions after the low. A short held with its stop above the last lower high would have been out there, at the first sign the sequence had failed.

The confirmed-swing reading did not flip until December, which is the lag described above, and the reason a stop works off price rather than the label.

Where the rallies come from

A downtrend is mostly made of the moves against it. Since 1994, 34 rallies of 5% or more were under way while SPY was in the strict downtrend state (counted by the state on the day each rally peaked). Their median size was 11.1%, and 19 of them reached 10% or more.

Several were very short. In March 2020 three separate 5%-plus jumps each took one session. In October 2008 SPY rose 14.5% in a single session. A rally that size inside a falling market is not a rare exception; on this count it is the normal texture.

This is why the size of a short matters more than the call. A stop placed tight against a falling chart sits inside the range of an ordinary counter-rally, so it gets hit on a move that later reverses.

The original data

The full sample is SPY’s daily closes from 29 Jan 1993 to 30 Oct 2026. Swing points use the 5% close-to-close filter described above (241 swings in all). A session is labeled a downtrend when the two most recent confirmed highs are falling and the two most recent confirmed lows are falling, an uptrend when both are rising, and mixed otherwise.

Labels start on 29 Mar 1994, once two of each exist: 8,203 sessions.

Downtrend sessions were the minority. 1,552 of the 8,203 sessions, 18.9%, met the strict downtrend rule, against 4,477 (54.6%) for an uptrend and 2,174 (26.5%) mixed. There were 25 separate downtrend stretches, with a median length of 40 sessions.

What followed a downtrend session was close to the base rate. Twenty sessions after a downtrend day, SPY closed higher 61.4% of the time (953 of 1,552 sessions), with a median change of +1.20%. From all sessions the figure was 64.3% (5,263 of 8,183) with a median of +1.33%.

That 2.9-point gap is within chance once the overlap is allowed for. A simple test gives a z-score of about -2.4, just past the usual 2, but it treats all 1,552 sessions as separate evidence.

They are not: each 20-session window shares 19 days with the next. Keeping only every 20th downtrend session (about 78 windows that do not overlap), the z-score ran from -1.8 to +1.1 depending on the starting day, inside 2 every time.

The fair summary is that a downtrend label described where price had been and said very little about the next month.

The deepest example in the record is still 2007 to 2009. The S&P 500 index fell from 1565.15 on 9 Oct 2007 to 676.53 on 9 Mar 2009, a 56.8% drop, and did not close above that peak again until 28 Mar 2013.

When it fails

A range can pass for a downtrend from its top. Two lower highs inside a trading range look like a sequence starting, until the lows refuse to break and price turns up from the floor.

The filter can manufacture or hide a trend. A small filter finds a downtrend in a quiet pullback; a large one misses a real decline that never bounced 5%. Change the filter and the label changes.

Late confirmation costs money. By the time a lower low is confirmed, part of the fall is already over, and the first counter-rally may arrive soon after.

Thin participation misleads. A lower low printed on light volume carries less weight than one where selling expanded, and the swing rule alone cannot see the difference.

Timeframes disagree. An hourly downtrend inside a daily uptrend is a pullback on the bigger chart, so the word needs a timeframe beside it before it means anything.

And the gap risk is real. A stop loss on a short above the last lower high does nothing against an opening gap that starts beyond it.

Questions people ask

How do you identify a downtrend?

List the swing highs and swing lows on one chart with one swing filter. If the latest high is below the previous high and the latest low is below the previous low, the chart is in a downtrend by this rule.

What ends a downtrend?

A close above the most recent lower high breaks the highs sequence, and a higher low breaks the lows sequence. In SPY in 2022, the close above $377.97 on 24 Oct was the first break.

How long does a downtrend last?

There is no fixed length. On SPY with a 5% filter, the 25 strict stretches since 1994 had a median of 40 sessions, about two months, and many were far shorter or longer.

Is a downtrend the same as a bear market?

No. A bear market is usually defined as a 20% fall from a peak. A downtrend is a sequence of swings, so it can appear in a 6% pullback and can be broken during a bear market, as it was twice in 2022.

Related

Market structure is the broader framework of swing highs and lows that this sequence belongs to. Trend following is the method built on riding a sequence until it breaks. And short selling covers the mechanics and borrow costs of taking the other side of a falling market.

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.