You watch a company beat earnings by 30%. The stock gaps up 6% at the open. By the close it has given back half the gap. Three weeks later it's 12% above the earnings-night close. The late chasers bought the open and got stopped on the fade. The patient ones waited for a pullback and caught the drift.
That's the trade most equity traders keep missing.
Earnings drift is one of the best-documented anomalies in equity markets. Academics first published it in 1968. It should have been arbitraged away decades ago. It wasn't. The drift still runs, and the reason it runs tells you how to trade it.
I want to walk through what the drift actually is, why it survives, what a clean setup looks like on the tape, and how I size it without betting the house on a single print.
What earnings drift actually means
Post-earnings-announcement drift is the tendency of a stock to keep moving in the direction of the earnings surprise for 30 to 90 days after the report. A big positive surprise tends to outperform the market for the next two to three months. A big negative surprise keeps underperforming. The move doesn't happen all at once on earnings night. It leaks out over weeks.
Ball and Brown first published the effect in 1968. Bernard and Thomas followed up in 1989 with the study that still gets cited in every finance textbook. Their top-decile positive-surprise stocks beat the market by roughly 3-5% over the 60 trading days after the report. The bottom decile underperformed by a similar amount. This held up across decades, across industries, across rising and falling markets.
Recent academic work confirms the effect hasn't vanished. It's smaller in mega caps, where earnings get front-run by event-driven hedge funds and systematic flows. But in small and mid-caps the drift is still between 2% and 4% over 60 days on the extreme decile. That's the part most retail traders can actually capture. You aren't competing with CTA models or earnings-day vol sellers down there.
Here's the thing. The effect isn't really about the beat itself. It's about what happens to analyst estimates after the beat.
Why the drift refuses to die
Three structural forces keep this alive.
First, sell-side analysts are slow. When a stock beats by 30%, analysts don't revise their forward estimates by 30% on day one. They revise up by 5%. Then, over the following four to six weeks, they revise up again. And again. Each revision pushes the stock a little higher. The drift is just the slow adjustment of forward estimates toward reality.
Second, institutions can't move in size without tipping their hand. A hedge fund that decides the beat changes the thesis takes three to five days to build a position. A long-only that rebuilds its model needs two weeks. That buying is patient, disciplined, and spread thin enough that the tape doesn't scream it.
Third, retail flow runs backwards. Retail tends to sell into strength after a gap-up. They're taking profits. Those sellers meet patient institutional bidding, and price grinds higher while the volume print looks unremarkable.
So you have slow analyst revisions, patient institutional accumulation, and misdirected retail selling. Put them together and you get a 60-day uptrend that isn't a trend in the technical sense. It's an estimate-revision cycle hiding inside the tape.
What a drift-worthy beat looks like
Not every beat is the same beat.
A company that beats the top line by 1% and the bottom line by 6% because of a tax benefit is not drifting. The quality of the surprise is in the drivers, not the headline number. A beat that comes from higher unit volume and expanding margins gets drift. A beat that comes from buybacks reducing share count does not. A beat driven by a guidance raise gets drift because guidance is where next quarter's consensus lives. A beat with no guidance change, or a quiet guidance cut, often fades inside a week.
Look at the next-quarter guide first. Not the headline. Not the EPS surprise. The guide. If the company raised its full-year revenue guide by more than 3-4%, you're looking at a drift candidate. If they reaffirmed the old guide, the report is probably dead money for this trade regardless of what the beat looked like.
Next, check analyst revision velocity in the five trading days after the print. Three or more upgrades plus a cluster of estimate hikes is the tell. The institutions haven't finished buying yet. Mean price-target bumps of 6% or more in a week after a beat usually mean drift is loading.
And watch the tape. If the stock held the post-earnings high during the first week, if every dip got bid within a session, and if the opening range of the first three sessions kept extending higher, the institutional flow is there. If the stock puked back through the earnings-night low in the first three days, the drift isn't happening. Skip it.
Now reduce all of that to three filters every setup I take has to pass before I put on size.
One: size of the surprise.
The effect concentrates in the extreme deciles. A 2% EPS beat is noise. I want a top-decile surprise, which in practice means EPS beat of 10% or more and revenue beat of 4% or more, both together. The standard academic measure is SUE (standardised unexpected earnings). If your data vendor publishes it, anything above a SUE of 4 is where the drift lives.
Two: guide revision direction. A beat with a raised full-year guide is a drift setup. A beat with a reaffirmed guide is probably not. A beat with a lowered guide is a fade, regardless of the headline number. This single filter kills more than half of the candidates I screen every quarter.
Three: market cap and float. Drift works best in $2-20 billion market caps. Mega caps get front-run by quant funds that already own the SUE literature. Micro caps are too thin for real institutional flow to show up on the tape. The middle is where the analyst-revision-to-price-action delay is widest. That's my hunting ground.
Pass all three and the trade has edge. Fail any one and you're coin-flipping.
How I size the trade
I want to be honest about something. Most of my earnings drift trades don't work inside the first 48 hours. Wait, actually, let me restate that. Roughly half of the candidates that pass my three filters work as expected. The other half chop for a week before the drift starts, or the drift never shows up because something in the macro changed.
So I size small, and I wait for confirmation. Here's the sequence.
Day 0 is earnings night. I don't trade the reaction. The opening range on day 1 is almost always too wide and too emotional. I watch. If the stock closes day 1 near the post-earnings high on above-average volume, the setup is live. If it closes near the low, the setup is dead. Binary.
Day 2 through day 5 is my entry window. I'm looking for a shallow pullback (3-5% off the day-1 high) that holds above the pre-earnings close. That reclaim is the structural retest. I enter there with a stop below the pre-earnings close. Roughly 1% of portfolio risk per position. Never more.
Target one is 6-8% above entry, trimmed on strength. Target two is 14-18% above entry, which lines up with the typical 60-day drift magnitude in small-mid caps after a strong surprise. I don't aim for target two on every trade. I aim for target one and let the position run when it acts right. Most of my P&L on this strategy comes from the two or three trades a quarter that reach target two.
Don't size for the home run. Size for the base hit and let the home run happen when it does.
What kills the drift
The drift fails in three scenarios, and I've been burned by all three.
A hostile macro tape drowns the signal. When the index is puking into a VIX spike, individual-name drift gets overridden by de-risking flows. I stand down when the SPX 5-day range is over 3% on rising volatility. The signal is still there, probably, but the noise-to-signal is too high to trade cleanly.
A sector-wide narrative reversal kills it. If three names in the same sector guide down after your drift candidate reported strong, the market reprices the whole sector. Your good report gets dragged along. I watch for peer reports in the following two weeks. If the sector breaks, I exit regardless of the individual chart action.
A buyback-driven beat collapses on the next report. I mentioned this in the guide-revision section but it's worth repeating. Beats that came from share-count reduction and not unit-volume expansion have no follow-through. The next quarter disappoints and the stock gives back everything. My rule: if the EPS beat is larger than the revenue beat by more than 5 percentage points, I don't touch it. That's usually a non-operational beat.
The trade I'm watching right now
I won't name a specific ticker here. Current setups turn over every week and this piece is meant to age for months, not days. But I can tell you the shape of the trade I'm watching.
Mid-cap industrial. Beat on both lines by more than 10%. Raised full-year revenue guide by 4%. Two analyst upgrades inside three sessions. The stock pulled back 5% on day 4, then reclaimed the day-1 high on day 6 with volume expanding. Pre-earnings close held throughout.
That's the setup I spent this morning writing into the order book. My guess is the entry gets tagged within two sessions, and the stop survives the first week. If I'm wrong about the second part, I'm out with a small loss and I move on. That's the job.
The work nobody sees
The reason this drift keeps paying is simple. Most of the market doesn't read the release. They read the headline, watch the open, and trade the reaction. The structural drift hides in the footnotes and the guide. It's slow, boring, and takes reading hundreds of releases a season to find ten clean setups.
That's the work Kunkel Capital Research puts in every earnings season. Screen the releases. Check the guides. Watch analyst revision velocity. Build the setup list. The signals our members get are filtered out of that same funnel.
You can do the work yourself. Most won't. The ones who do usually find that the hardest part isn't the math. It's the patience to let a shallow pullback develop before putting on size, and the discipline to skip 80% of the prints that don't pass the filters.
