Intraday Mean Reversion Strategies

A stock jumps 2% in fifteen minutes on heavy volume with no news. Some of that move is a genuine repricing — someone knows something. But a large chunk of it, historically, is temporary price pressure: a large order needing immediate execution pushed the price away from where informed traders think it belongs, and liquidity providers who stepped back during the rush haven't yet stepped back in to absorb it. That temporary component tends to fade within the same day. Trading that fade, systematically and at small size across many names, is intraday mean reversion.

Why the reversal exists

The mechanism is liquidity provision, not valuation. When a large order — a fund unwinding a position, an index fund rebalancing, a panicked stop-loss cascade — hits the market faster than resting liquidity can absorb it, market makers widen spreads and pull size rather than absorb the full flow at the old price. The price moves further than the information content of the trade justifies. Once the imbalanced flow passes, liquidity providers who were paid to take the other side during the move start unwinding their own inventory back toward a fair price, and the price partially reverts. This is the same economic logic as a market maker's inventory-risk premium: they demand a worse price to take on unwanted inventory, and give some of that premium back once the inventory is no longer unwanted.

Worked example

A $100 stock jumps to $102 in a ten-minute window on volume five times its normal rate, with no company-specific news and no move in its sector peers or index. A mean-reversion model flags the move as likely pressure-driven, based on the volume-to-news mismatch and the absence of a matching sector move, and shorts the stock at $102. Historically, moves of this shape retrace roughly 40-50% of the spike by the close. If the stock drifts back to $101, the strategy covers for a $1 gain per share before costs. Because the position is held only hours, not days, the strategy relies on running this pattern across hundreds of names a day rather than any single trade being large — the edge per trade is small and the win rate needs to clear transaction costs by a comfortable margin, since round-trip costs on an intraday trade eat a much larger share of a $1 move than of a multi-day trade.

What this means in practice

The hardest part of running this strategy is distinguishing pressure from information in real time, not after the fact — a move that looks like pressure can be the first leg of a real repricing if the news simply hasn't hit the tape yet. Desks filter aggressively: checking sector and index co-movement, cross-referencing news feeds and social sentiment for a lag, and sizing down or skipping names with active corporate events, where a "spike" might be entirely justified.