Residual Reversion Strategies

A stock can move for two very different reasons: because the whole sector or market moved, or because something specific to that company happened. Trading the raw price mixes both together. A residual reversion strategy explains away the first part with a factor model, and trades only what's left — the piece of the move that's specific to the stock and, empirically, tends to partially reverse.

Building the residual

Fit a regression of a stock's daily return on a set of explanatory factors — say a sector ETF and the broad market:

$$r_t = \alpha + \beta_1 \, r^{sector}_t + \beta_2 \, r^{market}_t + \epsilon_t$$

In words: today's stock return is decomposed into a piece that moves with the sector, a piece that moves with the market, and whatever is left over, $\epsilon_t$, the residual. A large positive residual means the stock moved up more than its usual relationship with sector and market explains — some stock-specific news, flow, or noise pushed it. Residual reversion strategies bet that outsized residuals tend to partially fade over the next few days, because much of what drives a short-term idiosyncratic spike is order-flow pressure rather than new fundamental information.

The explorer above fits a line through scattered points and shows the residuals — the vertical gaps between each point and the line — directly. In a residual reversion strategy, those vertical gaps are exactly what gets traded: not the raw $y$ values, but how far each point sits from what the factor model predicted.

Worked example

A stock returns +3.2% on a day when its sector ETF is up 0.9% and the market is flat. A fitted model with $\beta_1 = 1.1$ (sector) and $\beta_2 = 0.3$ (market) predicts a return of $1.1 \times 0.9\% + 0.3 \times 0\% = 0.99\%$. The residual is $3.2\% - 0.99\% = 2.21\%$ — over two points of move the factor model cannot account for. A residual reversion strategy would short a fraction of that idiosyncratic excess, expecting some of it to unwind over the next few sessions as whatever specific pressure caused the spike fades, while the sector and market exposure stay hedged out through the same factor weights used to build the residual in the first place.

What this means in practice

Because the systematic exposure is explicitly hedged, residual reversion is a market-neutral strategy: it lives or dies on whether idiosyncratic residuals actually mean-revert, not on market direction. It scales naturally to hundreds of names at once by running the same regression cross-sectionally each day.