Dead Cat Bounce: Every Failed Rally Since 1950, Measured

A dead cat bounce is a short rally inside a falling market that gives way to a new low. The name is a trading-floor joke, that even a dead cat will bounce if it falls from a great height, and the joke hides a real question: when the index bounces off a low, how often is that the bottom? We measured it. In every S&P 500 decline of 10% or more since 1950, there have been 85 bounces of at least 5% off a closing low, and 59 of them, 69%, were undercut by a lower low before the market got back to a record. The size of the bounce is the one thing that has separated them: all 42 bounces that stopped short of 10% failed, while 15 of the 18 bounces of 20% or more turned out to be the real recovery. This page gives the definition, where the phrase comes from, every bear market's count of failed bounces, how long a dead cat bounce has lasted, and what a rules-based trader does with the information.

• Definition: a brief recovery in a declining market or stock that is followed by a continuation of the decline; it can only be confirmed in hindsight, once the price falls below the low it bounced from.
• Count since 1950: 26 declines of 10% or more, 85 bounces of 5% or more inside them, 59 failed (69%), 26 that marked the bottom. In 7 of the 26 declines the first 5% bounce was the bottom.
• Size is the tell: failed in 42 of 42 cases at 5 to 10%, 10 of 16 at 10 to 15%, 4 of 9 at 15 to 20%, and 3 of 18 at 20% or more.
• How long it lasts: a failed bounce peaked a median 15 trading days after the low, gained a median 7.8% at its peak, and was undercut after a median 35 trading days; the longest took 195 trading days to fail. The new low that followed sat a median 17.3% below the low it had bounced from.
• Now: as of the Nov 11, 2026 close of 7,619.38, the S&P 500 is 2.6% below its record close of 7,818.93 on Oct 6, 2026, so no decline of the kind measured here is in progress.

What a dead cat bounce is, and where the phrase comes from

The phrase is younger than most market jargon. Its first recorded use is from December 1985, when Chris Sherwell and Wong Sulong of the Financial Times described a rise in the Singapore and Malaysian stock markets during that year's recession as "what we call a dead cat bounce"; both markets kept falling (Wikipedia). The analyst Raymond F. DeVoe Jr. spread it in a San Jose Mercury News column on April 28, 1986, about oil prices that had recovered from under $10 a barrel to over $13: "If you threw a dead cat off a 50-story building, it might bounce when it hit the sidewalk. But don't confuse that bounce with renewed life. It is still a dead cat." The same idea goes by other names, a sucker's rally, a relief rally, a bear market rally, or a bull trap when the rally pulls buyers in before it fails. The Corporate Finance Institute's definition adds the part that matters for anyone trying to trade it: the pattern "cannot be identified with certainty until it is analyzed in hindsight" (CFI). A bounce is only a dead cat once the price has made a lower low, which is also why the term gets argued about in real time; on April 10, 2025, the morning after the Nasdaq's 12% tariff-pause rally, Fortune pointed out that the surge was "not technically" a dead cat bounce yet, because the label applies only once an index falls below its previous low (Fortune). It never did.

How we measured it

We took every daily close of the S&P 500 from January 3, 1950 to Nov 11, 2026 from Yahoo Finance and applied three rules, the same ones our correction page uses for its episodes. A decline opens the first day the index closes 10% or more below its record close and ends on the day of the next record close; its bottom is the lowest close inside it. Inside a decline, a bounce starts from the lowest close so far and counts the first day the index closes 5% or more above that low. A bounce has failed, and is a dead cat, the day a close undercuts the low it started from; the bounce that starts from the decline's bottom is, by construction, the real recovery. We ran the same rules with a 20% decline and a 10% bounce to get the classic bear market rally, and with a 3% bounce as a check. Every figure below is at the closing-price level of the index, with no dividends and no intraday highs or lows.

The size of the bounce is the tell

The table is the whole finding. A bounce that fizzles under 10% has never been the bottom of a 10% decline in 76 years; every one of the 42 gave way to a lower low. Past 10% the odds turn, and past 20% they invert: of the 18 bounces that reached 20%, only 3 failed, the 21.4% rally from September 21, 2001 that peaked in early January 2002, the 20.7% rally from July 23, 2002 that failed in October, and the 24.2% rally from November 20, 2008 that gave up in February 2009. The catch is the one the definition already contains: you do not know the size of a bounce until it is over. What you can measure as it happens is speed, and there the two groups also differ, though with far more overlap.

Real bottoms moved faster: a median 4 trading days to the first 5%, against 6 for the ones that failed, and by the twentieth session after the low the real recoveries were up a median 9.6% against 4.8%. The gap widens from there, because a dead cat is by then on its way back down: sixty sessions after the low the failed group's median gain was 3.7%, the real group's 12.2%. None of that is a signal you could have traded cleanly. The two most violent bounces in the data, 14.9% in two sessions after Black Monday in October 1987 and 9.3% in one session on March 13, 2020, both failed, and the real recovery of April 2025 cleared 5% in a single day, so speed sorts the groups on average and misleads at the extremes.

Every bear market's dead cats since 1950

Every bear market had at least one dead cat, and the deep, slow ones had the most. The 1973 to 1974 decline and the 2007 to 2009 decline each produced nine failed bounces of 5% or more; the 2000 to 2002 decline produced seven, the longest-lived of them the 21.4% rally that began on September 21, 2001, peaked 72 trading days later and was not undercut until July 2002, 195 trading days after it started. The worst outcome for anyone who bought a bounce as the bottom came in the same era: the first 5% bounce of the financial-crisis bear market started from a low on November 26, 2007, and the eventual bottom on March 9, 2009 was 51.9% below it. The fast crashes went the other way. The 2020 decline took 23 trading days from record to bottom and still fit three dead cats into them, a 6.0% bounce from February 28, the 9.3% single-day jump of March 13 and a 6.0% rise from March 16, each undercut within days before the March 23 low, in the same month the market-wide circuit breakers tripped four times. In 1987 the Black Monday close of October 19 was followed by a 14.9% two-session bounce that gave way to a lower close on December 4. And 2022, the most recent bear market, is the textbook case: five failed bounces, the biggest a 17.4% rally from June 16 to August 16 that convinced a lot of people the bear was over before the index fell to its October 12 low.

Measured the classic way, as rallies of 10% or more inside declines of 20% or more, the bear market rally looks like this: 27 of them since 1950, 16 of which failed (59%). Four bears (1962, 1966, 1970 and 2020) had no failed 10% rally at all, the 2000 to 2002 and 2007 to 2009 bears had four each. A failed bear market rally gained a median 13.5%, peaked 34.0 trading days after its low, failed after 66.0, and was followed by a low a median 16.9% lower. The size rule holds here too: all 9 bear market rallies that stopped between 10% and 15% failed, all 4 between 15% and 20% failed, and 11 of the 14 that reached 20% were the real recovery.

Corrections: when the first bounce was the bottom

Corrections are where the first bounce has a real chance of being the bottom. In 7 of the 15 corrections since 1950 the first 5% rally off the low was the recovery: 1950, 1955, 1984, 1997, 1998, the February 2018 drop and the tariff correction of 2025, which fell 18.9% from February 19 to April 8 and then rose 5% in one session, 23.2% in 54, and never looked back. The other eight corrections produced one or two dead cats each; the longest-lived of them was the 2015 to 2016 decline's single failed bounce, a 13.0% rally from the August 25, 2015 low that peaked 49 trading days later and was undercut in January 2016 on the way to the February bottom. Put together with the bear-market table, the pattern is simple: shallow declines tend to resolve on the first bounce, deep ones almost never do, and nobody knows which kind of decline they are in until it is over, which is the point our drawdown page makes with recovery math and the correction page makes with the 11 of 26 corrections that turned into bear markets.

How long a dead cat bounce lasts

Measured on the index since 1950, a failed bounce took a median 6 trading days to reach 5% above its low, peaked after a median 15 trading days (about three weeks) with a median gain of 7.8%, and was undercut a median 35 trading days (about seven weeks) after the low. The spread is wide. The March 2020 bounces peaked in one to three sessions and failed within a week; the September 2001 rally took 72 sessions to peak and 195 to fail; the June 2022 rally took 41 sessions to peak and 69 to fail. For single stocks the numbers will be different and usually faster, since an index bounce is the average of hundreds of them, but the shape is the same: a sharp first leg usually attributed to short covering and bargain hunting, a peak within a few weeks, and then the resumption of whatever was pushing the price down in the first place, which the bounce did nothing to fix.

What a rules-based trader does with a bounce

The honest conclusion of the data is that a bounce cannot be classified in real time, which is why rules beat judgment here. A trader who buys every bounce as the bottom is wrong 69% of the time inside a 10% decline and loses a median 17.3% from the entry to the next low if nothing stops the trade. A trader who waits for a 20% rally before believing it has been right 15 times out of 18 but has missed the first 20% every time. The usual answer is to stop deciding and let two settings do the work. A stop-loss turns a dead cat into a small, defined loss: the median failed bounce gave up its entire gain and more within seven weeks, and a stop placed a few percent under the entry is hit on the way, well before the median 17.3% new low. A trailing stop handles the other case: a rally that keeps making higher closes carries the stop up with it, and a bounce that rolls over hands back only the trail's width from its peak rather than the median 17.3% that followed a failed bounce. Our test of trailing-stop widths on every 10% mover in October 2026 shows what each width costs in whipsaws; the daily-drop page has the odds of another big down day after the first one; and the VIX page shows what the volatility readings around these lows have meant for the next month.

That is how JorgAI is built to behave. It does not try to call bottoms: it buys when a stock passes its screen and the confidence score clears the threshold the member set, and it sells by the rules the member set, a stop-loss, a take-profit and an optional trailing stop that are entered as percentages in Settings (among live settings the median stop sits at 5% and the median trailing stop at 4%, listed in the settings glossary). A stopped-out position is not re-bought because the market bounced; it comes back only if it passes the screen again on a later day. Daily spend and position caps keep a run of failed bounces from compounding, and the loss circuit breaker in Settings pauses buying after a losing streak. None of that predicts anything, which is the point: in 59 of 85 cases since 1950 the prediction would have been wrong. The free account lets you set those rules up and watch them run on the demo before any broker is connected.

Quick answers

What is a dead cat bounce?

A short recovery in a declining market or stock that is followed by a fall below the previous low. The phrase, first recorded in the Financial Times in December 1985, comes from the idea that even a dead cat will bounce if dropped from high enough. It is confirmed only in hindsight, once the lower low has printed.

How long does a dead cat bounce last?

On the S&P 500 since 1950, failed bounces of 5% or more peaked a median 15 trading days after the low and were undercut a median 35 trading days after it, with a range from a few sessions (March 2020) to 195 trading days (the rally that began in September 2001).

How can you tell a dead cat bounce from a real bottom?

Not in advance. After the fact, size has been the cleanest separator: every bounce under 10% inside a 10% decline since 1950 failed, while 15 of 18 bounces of 20% or more were the real recovery. Real bottoms also moved faster in the first few weeks, but the fastest bounces on record were dead cats, so speed alone is not reliable.

Was April 2025 a dead cat bounce?

No. The S&P 500 fell 18.9% from February 19 to April 8, 2025, rose more than 5% on April 9 after the tariff pause, and reached a new record on June 27 without undercutting the April 8 close. The 2025 decline is one of seven corrections since 1950 in which the first bounce was the bottom.

What is a bear market rally?

A rally of 10% or more inside a decline of 20% or more that fails. Since 1950 there have been 27 rallies of that size inside bear markets and 16 failed; the failed ones gained a median 13.5% and were followed by a low a median 16.9% lower.

Is the stock market in a dead cat bounce right now?

As of the Nov 11, 2026 close the S&P 500 was 2.6% below its record of Oct 6, 2026, so it is not in a decline of the kind measured here. The term only applies once a market has fallen hard, bounced and then made a lower low.

How we put this together. The counts come from Yahoo Finance daily closing prices of the S&P 500 index (ticker ^GSPC) from January 3, 1950 to Nov 11, 2026, pulled on November 12, 2026 and processed with the three rules described above; the computation is ours and every figure in the tables is reproducible from that series. The definition and history are from Wikipedia's entry on the phrase, the Corporate Finance Institute's definition and Fortune's April 10, 2025 explainer, all read the same day. The settings medians describe current live JorgAI settings, not results. Market history is not a forecast: 76 years of index closes contain 26 declines of this size, which is enough to describe what has happened and not enough to say what will.

JorgAI is software that carries out the rules you set in your own brokerage account. It is not investment advice and does not predict or promise results. Trading involves risk of loss, and past market behavior does not determine future behavior.