Pre-Earnings Drift as a Signal
The 'pre-earnings drift' effect is one of the more robust empirical findings in the literature: stocks tend to drift in the direction of the actual earnings surprise in the 5-10 trading days before the print. Not every time, and not for every stock — but enough on average that it shows up in long-run studies.
INTC is not a clean pre-earnings drift name historically. Its drifts have been noisy, often moving opposite the eventual print direction. That's consistent with a stock where the buy-side consensus is well-dispersed and the news-flow between quarters tends to dominate the pre-earnings tape.
What This Cycle Looks Like
In the 5 trading days ending today, INTC drifted modestly negative on mixed volume. The historical analogs that match this setup — mid/large-cap semi names drifting negative into earnings on mixed volume — split 9 of 20 positive at the 5-day post-earnings horizon, 11 of 20 negative.
That's a mild negative skew, which is consistent with the pre-earnings drift literature. But the magnitude is small and the sample size is limited. The more useful output from the analog set is the dispersion, not the central tendency: 5 of 20 analogs produced 10-day moves greater than ±8%, so the expectation is more 'outlier move likely' than 'expect a specific direction.'
- Match set n=20, 5-day median post-earnings return: approximately -0.6%
- Tail events (|return| > 8%) in 5 of 20 cases (25%) — outsized vs the unconditional INTC base rate of 14%
- Semi-specific analogs (AMD, MRVL, NVDA pre-print): modestly better than cross-sector matches
How to Use This
A mild negative skew with elevated tail risk is not a directional trade. For an agent, the actionable read is position-sizing: INTC's analog set suggests the realized move tonight is more likely to be large than small, in either direction. The specific direction is close to a coin flip on this pre-earnings setup.
Contrast with a stock like LLY or NOW where pre-earnings drift has historically been a much cleaner signal. Chart Library's pattern search exposes that distinction by returning different analog clusters for different drift profiles — the embedding captures the pre-event shape, not just the ticker.