Education
Contents
By Arizmic LLC10 min readPublished Last reviewed

Dead-cat bounces: rebound rules and trend failure

Make a post-selloff rebound testable without confusing a later failure with information available at the time.

A dead-cat bounce is a label for a path, not a causal signal

A dead-cat bounce describes a rebound that follows a sharp decline but later fails to establish a durable recovery. It is a useful research label only when it is made operational. In live time, nobody knows that a rebound is a dead-cat bounce until later price action has supplied the answer.

That distinction prevents a common error: selecting declines that eventually continued downward, then describing the visible rebound as though the rule had known its future failure. A tradable study must instead identify a candidate shock and rebound using only information available then, state whether it is testing a tactical long or a failed-bounce continuation trade, and retain every candidate that later recovers as well as every candidate that fails.

Unlike QFL, this is not organized around a prior support base. It is organized around a fast decline, the rebound that follows it, and the question of whether the earlier trend failure still dominates the next holding window.

Rule design: choose one thesis before choosing a pattern

The same price path can support opposite hypotheses. A countertrend long asks whether urgent selling has temporarily overshot available liquidity. A failed-bounce short asks whether the rebound is too weak to reverse a broader repricing. They need different entries, borrowing assumptions, and risk limits.

DecisionTactical rebound longFailed-bounce continuation
EventSharp completed decline from a declared peak or prior referenceSame shock, followed by a bounded rebound
EntryReversal confirmation after the lowRejection below a declared recovery level
TargetPartial recovery or a short timeoutReturn to the prior low or a short timeout
Main invalidationNew low or continued volatility expansionReclaim above the recovery level or a squeeze
Implementation riskGaps lower and further liquidationBorrow, financing, gap, and squeeze risk

Do not merge the two in one backtest and credit whichever direction worked. Choose the thesis before observing the later path, then report both opportunity sets separately if both are researched.

Define the shock and rebound in point-in-time terms

A usable candidate definition might contain four pieces:

  1. Shock reference: a previously known high, close, or session reference.
  2. Decline threshold: a fixed percentage or volatility-scaled fall observed at a completed event.
  3. Rebound definition: a completed-bar recovery of a stated size, or a recovery that fails at a stated reference.
  4. Decision and expiry: the first eligible execution after confirmation, plus a short horizon after which the pattern is no longer considered active.

The reference must be chosen carefully. A running high from the start of a dataset is not automatically a session high, and an overnight gap may need a separate prior-close input. If the rule uses intraday data, keep the session, timezone, and opening/closing convention fixed across the sample.

A worked shock-and-rebound sequence

Illustrative example. Price falls from a known reference of 100 to 80, a 20% decline. It then rebounds to 86, a 7.5% recovery from the low while still 14% below the original reference.

Those numbers do not reveal the next move. They support two predeclared, mutually exclusive experiments:

  • A tactical long can wait for the rebound at 86 to be confirmed at a completed event, enter only at the next eligible price, and use a fixed partial-recovery target with a tight expiry. Its risk is that the same liquidation resumes.
  • A continuation short can wait for a rejection below a declared recovery threshold, enter only after that rejection is known, and treat a reclaim above the threshold as invalidation. Its risk is a violent recovery, borrow change, or gap.

Record the event time, decision time, order time, and fill. The apparent low at 80 is context. Assigning every trade a fill at that low would make a confirmation rule look better than it could have been.

Experiment design: measure paths, not just the final label

Research should begin with event studies rather than a verdict on individual charts. For each eligible decline, align the path at the causal decision event and measure:

  • maximum continuation below the entry and maximum rebound above it;
  • time to the recovery threshold, prior low, invalidation, or timeout;
  • spread, trading volume, and realized or conservative fill assumptions;
  • whether the decline followed scheduled information, an overnight move, or a normal-session move; and
  • overlapping signals created by the same selloff.

Compare those paths with a matched set of sharp declines that did not receive a rebound confirmation. If the candidate label adds no information beyond the initial shock, the recovery pattern has not earned a separate rule.

The short side requires its own feasibility record: point-in-time borrow or an explicit shortability proxy, financing, hard-to-borrow changes, and a fill model that recognizes gaps. A continuation result without those inputs is a price-path observation, not a deployable short strategy.

Why the pattern can fail in either direction

Temporary selling pressure can produce a fast rebound once liquidity returns. But severe declines also occur when information changes valuation, leverage is unwound, or market makers reduce the risk they are willing to carry. Under those conditions, an early rebound may merely pause a larger repricing. The strategy cannot infer which explanation is true from the shape alone.

That is why the test should vary the initial decline size, rebound size, volatility state, market trend, and event timing out of sample. It should also separate assets that can be shorted realistically from those where the short leg is unavailable. A high hit rate for small countertrend recoveries does not offset a rare gap or continuation event if position sizing ignores that tail.

Failure modes to keep visible

  • Post-hoc labeling: calling a bounce "dead" only after the later decline.
  • Intrabar hindsight: crediting the low, high, or rejection price before a completed rule could observe it.
  • Selection on survivable names: excluding delisted securities, halted instruments, or unavailable borrow.
  • Overlapping shock events: counting several bars from one selloff as independent evidence.
  • Unmodeled discontinuity: a stop can be crossed by a gap without filling near its trigger price.
  • Direction laundering: combining long-bounce and short-continuation outcomes until the best-looking interpretation emerges.

A useful dead-cat study is willing to conclude that the label carries no actionable information once timing, cost, and tail paths are included.

Try it in Arizmic

Strategy composition

Build the premise with shipped signals

A dead-cat bounce is a post-shock rebound hypothesis, not a label that can be known in advance. Separate a tactical rebound long from a failed-bounce continuation trade, preserve the causal decision time, and make gap, liquidity, and shortability assumptions visible.

These are starting structures, not presets or evidence of an edge. Choose one, replace the bracketed decisions, and keep the signal roles separate as you test it.

Composition 01

Shock drawdown with rebound confirmation

Tactical-long recipe
  • DrawdownSharp-decline state
  • SMA RSIRebound confirmation
  • Relative VolumeParticipation context
  • Wilder ATRTail-risk scale

Data: OHLCV Bars

View configuration and complete rule

Drawdown

Sharp-decline state

Drawdown makes the severity of the preceding decline explicit before any rebound is interpreted.

Configure

No configurable parameter is required for this role.

Use the output

drawdown
Create a candidate only after the declared drawdown threshold has been reached.
running_high
Inspect the cumulative peak that the drawdown state references.

SMA RSI

Rebound confirmation

SMA RSI prevents the rule from treating the first sharp low as a fillable reversal by default.

Configure

window · [rebound window]
Set the short horizon used to require a completed recovery in momentum.

Use the output

rsi
Enter long only after SMA RSI has recovered through [threshold] while the shock state remains active.

Relative Volume

Participation context

Relative Volume makes the participation context available for both the shock and the rebound.

Configure

window · [same-time volume window]
Compare the selloff and rebound with a declared participation baseline.

Use the output

relative_volume
Filter or partition candidates by [minimum relative volume] rather than assuming every rebound has the same liquidity.

Wilder ATR

Tail-risk scale

Wilder ATR prevents a fixed price stop from silently changing its risk meaning across volatility states.

Configure

window · [ATR window]
Scale the initial loss limit and unit size to the range prevailing during the shock.

Use the output

atr
Set [invalidation distance] and report maximum adverse excursion in ATR units.

Assemble the rule

Entry
After a completed drawdown shock, enter a tactical long only at the next eligible event after the declared SMA RSI rebound confirmation is complete.
Exit
Exit at [partial recovery target], [ATR-based invalidation], or [short timeout], whichever occurs first.
Decision time
The shock state and rebound confirmation are observed on completed events; do not assign the entry to the earlier low.
Sizing
Scale a single long position to the ATR-based invalidation distance under a fixed risk budget.

Useful variations

  • Compare immediate post-shock entry with completed-bar rebound confirmation and retain delayed or missed trades.
  • Partition outcomes by drawdown severity, participation state, and time to confirmation.
  • Compare a partial-recovery exit with a strict time exit while keeping the entry and risk budget fixed.

Keep in view

Drawdown is cumulative from the first bar and SMA RSI differs from Wilder RMA RSI. Relative Volume needs a declared same-time-of-session baseline; none of these signals can tell whether a rebound will later fail.

Composition 02

Opening shock with failed-rebound continuation

Continuation-short recipe
  • Overnight GapOpening-shock definition
  • SMA RSIWeak-rebound state
  • Relative VolumeExecution context
  • Wilder ATRSqueeze and gap-risk scale

Data: OHLCV Bars

View configuration and complete rule

Overnight Gap

Opening-shock definition

Overnight Gap distinguishes a discontinuous opening shock from an ordinary intraday decline.

Configure

No configurable parameter is required for this role.

Use the output

gap_pct
Create a candidate only after a declared negative prior-close-to-open gap threshold is observed.
gap
Review the raw gap alongside the normalized percentage for high-priced and low-priced instruments.

SMA RSI

Weak-rebound state

SMA RSI operationalizes a weak recovery without claiming that any low reading predicts continuation.

Configure

window · [rebound window]
Define the short horizon used to test whether recovery momentum remains bounded.

Use the output

rsi
Require the completed rebound state to remain below [recovery threshold] before a short candidate is eligible.

Relative Volume

Execution context

Relative Volume keeps the price-path thesis separate from whether the modeled execution environment was active.

Configure

window · [same-time volume window]
Separate active opening conditions from thin or abnormal participation.

Use the output

relative_volume
Require or report [minimum relative volume] at the attempted entry.

Wilder ATR

Squeeze and gap-risk scale

Wilder ATR provides a consistent scale for a continuation thesis whose failure can be discontinuous.

Configure

window · [ATR window]
Set a predeclared invalidation and maximum loss budget in current-range units.

Use the output

atr
Express the short invalidation and adverse excursion in ATR units, while retaining any gap through the stop separately.

Assemble the rule

Entry
After a declared negative opening gap, enter short only at the next eligible event after a completed rebound remains below the recovery policy and liquidity conditions pass.
Exit
Exit at [prior-low target], [ATR-based reclaim invalidation], or [short timeout], and record any gap through the intended exit.
Decision time
The opening gap is observable only after its stated prior-close and opening references exist; the failed-rebound state is a later completed-event decision.
Sizing
Use a fixed risk budget only where shortability, borrow, financing, and gap-loss assumptions are available; otherwise keep this as a price-path study.

Useful variations

  • Compare the failed-rebound short with the tactical-long formulation on separate, predeclared opportunity sets.
  • Partition candidates by gap size, time of day, and relative volume without changing the recovery rule per bucket.
  • Use conservative fills and borrow assumptions, then report the difference between price-path and deployable results.

Keep in view

Overnight Gap requires an explicit prior-close input and session definition, SMA RSI is not Wilder RMA RSI, and a price-only short result omits borrow, financing, halts, and gap-through-stop risk.

Ask the AI Companion

Draft this strategy

Strategy draft

Turn a sharp-decline and rebound idea into one directional draft with causal entry timing and explicit gap or shortability boundaries.

I want to research [instrument] after a sharp selloff and a partial rebound. Help me choose one starting hypothesis: a tightly risk-capped tactical rebound long or a failed-bounce continuation short. Define the shock, rebound confirmation, decision timing, exit, timeout, and risk boundary using only information available at the time. Then build the strategy and explain what data or execution assumptions would be required before treating a short result as deployable.

Extend it in Marimo

Study review

Begin from a retained Study so shock events, rebound confirmations, orders, fills, gaps, and unresolved candidates remain linked to evidence.

Align every eligible selloff at its causal decision event and distinguish the initial shock, rebound, continuation, invalidation, timeout, and execution path.

Bring in
engine-reported Drawdown drawdown and running_high, SMA RSI rsi, Relative Volume relative_volume, Wilder ATR atr, Overnight Gap gap and gap_pct, decisions, and fills, declared shock, rebound, target, invalidation, session, shortability, and expiry policies, retained prices, costs, order attempts, fills, gap events, and any available borrow or financing records
Build
event-aligned shock, rebound, and continuation paths from the actual decision time, long and short candidate table with recovery, invalidation, timeout, and maximum-excursion outcomes kept separate, price-path versus implementation-feasible comparison by gap size, participation state, and execution assumptions

Interpretation: Check whether the rebound definition adds information beyond the original selloff and whether any short-side result survives explicit borrowing, financing, fill, and gap assumptions.

Value origin: Signal values, decisions, order attempts, fills, costs, and retained outcomes are engine-reported. Event alignment, path labels, and implementation-feasibility partitions are notebook-derived; borrow and financing analysis is only as complete as the retained source data.

With Companion: Ask Companion to draft event-aligned shock-path and feasibility-comparison cells, inspect decision timing and missing borrow fields, then apply the diff explicitly.

Further reading