Analyst Revision Breadth
In plain terms
Approximates analyst-revision breadth from the direction of consensus EPS-estimate changes over the last four quarters (we lack per-analyst revision data). Long when estimates have been rising in most recent quarters by a meaningful amount; short when falling. A simplified stand-in for the paper's per-analyst up/down revision counts.
How it works
Paper mechanism (Brown 1991 / Givoly-Lakonishok 1979): count of upward vs downward per-analyst EPS revisions over a trailing window, normalized by total analyst count; breadth (proportion of analysts moving the same direction) predicts post-revision drift better than magnitude. IMPLEMENTED AS A DOCUMENTED v1 PROXY: no per-analyst estimate panel is available (FMP exposes only the rolling consensus), so the generator uses the sign of quarter-over-quarter changes in the consensus EPS estimate, sign-summed over the trailing 4 quarters (net_breadth_4q in -4..+4) plus an average percent-change filter. Caveat: delta_est compares estimates for two different fiscal quarters, so the proxy partly reflects sequential earnings growth, not pure analyst revision activity.
Data dependencies
- Earnings history
A data feed this strategy reads, refreshed on its normal schedule.
Expected edge
- Reported return
- ~3-6% ann. L/S (Brown 1991 / Givoly-Lakonishok 1979)
- Tested over
- 1976-1988
~3-6% annualized long-short, lower in high-VIX regimes.
Related families
A big overnight gap NOT preceded by an analyst revision is mispriced — the revision arrives ~5 days later and the price drifts further in that direction.
When analysts disagree widely on a stock's earnings, it's overpriced (pessimists can't short-sell in size). Short high-dispersion, long low-dispersion.
Explore Analyst Revision Breadth on alphactor.ai
See which tickers this family is currently firing on, with live signals and rankings.