SPY Mean Reversion After 1% Down Days

Conclusion: Inconclusive

Out-of-sample Sharpe was 0.03, versus 0.05 in-sample. A bootstrap test returned p = 0.139 (not significant).

Short-term mean reversion in SPY is real in the literature and faintly visible here, but it is too small, too concentrated in crisis windows, and too fragile to the test setup to trade. It stays a backtest and is not promoted to paper. (Note: an earlier version of this experiment reported a 3.20 Sharpe at p=0.52 — both were artefacts of overlapping trades and a miscalibrated significance test, since corrected.)

Method

Rule / pseudocode

DATA ← adjusted observations for the stated instrument and horizon

RULE ← SPY generates positive 3-day forward returns after daily drops exceeding 1%, as short-term panic-selling exhausts.

METHOD ← Entry: Close on signal day (daily return < -1%) · Exit: Close 3 trading days later · Train/test split: 70/30 by trade count (459 IS trades through 2020-04-15, 197 OOS) · Costs (bps): 10

FIT ← estimate or select parameters on each training window only

TEST ← apply the frozen rule to the next unseen window after stated costs

REPORT ← return, Sharpe, drawdown, trade count, significance and fold stability

VERDICT ← prefer robustness and sufficient observations over the headline return

_Well-documented short-term mean reversion in broad market indices after sharp single-day selloffs. Avoids survivorship bias by trading the index ETF directly. The original post proposed buying individual momentum stocks but was rightly criticised for using a static 2026 stock list; buying SPY itself preserves the signal while eliminating selection bias._

Out-of-sample equity curve

Growth of $1 over the held-out test data the strategy never saw while it was being built, after 10 bps round-trip costs.

Out-of-sample performance

OOS Sharpe

0.03

Ann. return

Max drawdown

-25%

Win rate

57%

p-value

0.139

not significant

WF folds +

3/5

Measured on held-out data. A high Sharpe with a non-significant p-value means the result could still be luck — read them together.

Hypothesis

SPY generates positive 3-day forward returns after daily drops exceeding 1%, as short-term panic-selling exhausts.

Method

Entry: Close on signal day (daily return < -1%) · Exit: Close 3 trading days later · Train/test split: 70/30 by trade count (459 IS trades through 2020-04-15, 197 OOS) · Costs (bps): 10

Results

Out-of-sample Sharpe was 0.03, versus 0.05 in-sample. A bootstrap test returned p = 0.139 (not significant). Walk-forward Sharpe stayed positive in 3 of 5 folds. Worst out-of-sample drawdown was -25.48%.

Analysis

The rule is deliberately simple: when SPY closes down more than 1% on the day, buy at the close and exit three trading days later. We measured it on non-overlapping trades only (no new position opens until the current one closes) across 2005–2026 — 380 trades, split 70/30 in-sample/out-of-sample.

On the trades themselves the signal looks encouraging: a 57% out-of-sample win rate and an average trade of +0.36% after 10 bps round-trip costs. But that flatters the strategy, because it sits in cash roughly 88% of the time. Measured honestly on the capital it actually ties up — a daily mark-to-market equity curve — the out-of-sample Sharpe is 0.03 (in-sample 0.05). On a risk-adjusted basis, deployed capital barely beats sitting still.

Walk-forward analysis is no more reassuring. Three of five folds are positive (60%), but the fold Sharpes swing from -0.9 to +11, so the result leans heavily on a couple of volatile windows (2008, 2020) rather than a steady, repeatable edge.

Significance confirms the doubt. Under a corrected block-bootstrap test — which demeans the returns so the null genuinely has no edge, and resamples in blocks so overlapping holds cannot masquerade as signal — the out-of-sample p-value is 0.139, short of the 0.10 bar. Worse, the p-value swings from 0.001 to 0.139 depending on whether the out-of-sample period is cut by trade count or by calendar date. An edge that flips significance with the split is sample-window dependent, not dependable.

Verdict: INCONCLUSIVE. Short-term mean reversion in SPY is real in the literature and faintly visible here, but it is too small, too concentrated in crisis windows, and too fragile to the test setup to trade. It stays a backtest and is not promoted to paper. (Note: an earlier version of this experiment reported a 3.20 Sharpe at p=0.52 — both were artefacts of overlapping trades and a miscalibrated significance test, since corrected.)

Source

Originated from discussion on r/manual.

Evidence

Walk-forward Sharpe by fold

Each fold retrains on past data then tests on the next unseen window. Staying positive across folds is the real test of an edge.

---

_Generated by labs-algo-trading. Automated research — not financial advice. Backtests overfit; treat verdicts as hypotheses._