When to Buy After a Stock Market Correction

For each index: its fund (SPY or QQQ) as an ordinary price chart, then the index against its highest close of the past 252 sessions, then its volatility index. The shaded zone is a correction; each dashed line marks a day the re-entry signal fired (the index 5% off its correction low with its volatility index 25% off its peak), labelled Failed when the index went on to a new correction low. The signal fires once per correction, so after a failed signal the chart shows no second one at the real bottom (2022 is the clearest case); a correction without a line ended before both conditions were met.

Each rule buys once per correction, on the close, and is measured to the close a year (252 sessions) after the correction’s low. Across 15 S&P 500 corrections since 1990 and 23 Nasdaq-100 corrections since 2002, buy at −10% had the highest median on both. It also had a new low after 80% of S&P 500 buys. Every rule that waits for confirmation is wrong less often and pays more for it. Rebound with fading volatility passed the volatility test set in advance: within two points of the best price-only rule’s median in both markets, with new lows after at least ten points fewer of its buys.

Median return from the entry close to the close 252 sessions after the low (price only). “New low after” is the share of corrections in which the index later closed below the correction’s low at the time of the buy. A rule that never fired buys when the correction ends. The volatility-term rule needs VIX3M, so it starts in 2006.

Newest first. For each rule: when it bought, how far above the eventual low, and what it earned a year after the low. An arrow marks a buy that a new low followed.

A trader who buys a rally and sees a new low usually sells and waits for the next signal. Tested that way, with a stop at any new correction low and unlimited retries, no rule improved on buying once and holding. A new low tends to come near the bottom of a leg, so each failed rally is sold low and bought back higher.

The longest record uses the S&P 500’s all-time high instead of the 52-week high (10% below the record until the record is regained), price-only rules, and one more breadth rule from Ken French’s industries: 26 scored corrections since 1927. The leading rule changed by era (1928–1949: new 20-day high; 1950–1989: 10% off the low; 1990 onward: buy at −10%), so under the rule set in advance no rule dominates.

• Buy at −10%. The first close 10% below the 52-week high: the correction's first day.
• Follow-through day. A rally attempt starts on the first up close after a new low; on day 4 or later, a gain of at least 1.25% confirms it. A close below the attempt's low resets it. Price only.
• 10% off the low. The first close 10% above the lowest close of the correction so far.
• Rebound with fading volatility. The index at least 5% above its correction low and its volatility index (VIX or VXN) at least 25% below its correction peak, on the same close.
• VIX back below VIX3M. VIX closes back below VIX3M after closing above it during the correction (the volatility curve returns to normal).
• Every rule and pass mark was written down before its test was run, in three tests: the S&P 500 since 1928, the volatility rules, and stop-and-retry. One coding error was found and fixed after the volatility test (the Nasdaq-100 term rule had compared VXN with VIX3M instead of VIX); no verdict changed.
• The samples are corrections, not days: 15 for the S&P 500 with VIX and 23 for the Nasdaq-100 with VXN. Medians over so few cases move when one correction is added.
• Returns are index prices without dividends, measured to a shared date a year after the low, which only hindsight knows. They compare rules with each other; they are not a trading record.