Index dip reversion is horizon and regime dependent

A laboratory with one entry

Editorial reading: teach a single long-side dip rule as a laboratory. Keep the 10-day-low entry fixed and vary only the hold. Treat the result as a classifier for whether that market’s stop, liquidity, and short-covering environment still favors continuation or has flipped to mean-reversion.

Continuation is follow-through in the same direction after a short-term extreme, here a fresh multi-day low that keeps falling. Mean-reversion is a tendency for price to bounce toward a recent average after a short-term decline rather than to extend the move. Short-covering-pressure is buy-to-cover demand that can lift a falling price back toward its recent mean when short interest is large.

What the archive held fixed

The long-side procedure bought the S&P 500 close on a 10-day low, exited after a fixed 10-session hold, and used a constant 100000-dollar ticket with no costs from 1970 through 2013.

The same study focused on long-side reversion: an upward move after a short-term decline, not a two-sided fade.

Mean-reversion is instrument-dependent

Mean-reversion is treated as instrument-dependent. Commodity and currency series are described as more continuation-prone, while the US equity daily series is described as more reversion-prone.

Change the hold, not the entry

Shortening the hold to two sessions after the same 10-day-low entry split the path: follow-through into further weakness through 1987, then bounce behavior thereafter. The two-session path’s inflection is dated to 19 October 1987.

A short-term drop in the index was described as more often followed by a bounce than by immediate further decline in the first few sessions, in both rising and falling market stretches.

Reversion that took several days in the 1970s and 1980s is described as completing on a shorter clock in the later sample.

Stops and the pre-automation cascade

Before the mid-1980s, stop-loss fills often waited until the next session and could stack selling across days. That stacking is a stop-cascade: a sequence in which delayed stop-loss fills generate further stops and stretch a decline across sessions. Same-session automated stops later compressed that cascade into the trading day.

Named drivers of the short-horizon flip

Four candidate drivers of the short-horizon flip are named: exponential volume growth, instant stop execution, broader electronic short covering, and automated buying of oversold conditions.

Transaction volume is described as growing exponentially from the early 1960s and reaching a high in the fourth quarter of 2008, offered as one reason later markets could absorb selling without multi-day runs.

Editorial use of the split

Editorial reading: once the entry is fixed, the two-session versus 10-session contrast classifies the stop, liquidity, and short-covering-pressure setting. If the short hold still extends the low, continuation remains the relevant market-regime for that clock. If the short hold bounces, mean-reversion has taken the first few sessions.

The dated flip on the two-session path sits on the pre- versus post-automation line. The archive separately described a bounce bias in the first few sessions in both rising and falling stretches, so that bull-or-bear contrast is not the split being dated.