| Ticker | Company | Sector | Streak | Latest surprise |
|---|---|---|---|---|
| Loading… | ||||
| Ticker | Company | Sector | Streak | Latest surprise |
|---|---|---|---|---|
| Loading… | ||||
| Ticker | Company | Sector | Reported | Est. | Surprise |
|---|---|---|---|---|---|
| Loading… | |||||
| Ticker | Company | Sector | Reported | Est. | Surprise |
|---|---|---|---|---|---|
| Loading… | |||||
Each quarter, an S&P 500 company either beats, meets or misses the consensus analyst EPS estimate. The gap between reported and estimated EPS, as a percentage of the estimate, is the surprise. This page pulls Alpha Vantage's quarterly earnings history for every S&P 500 company (sector and name come from Sector Beeswarm) and works out three things: how often companies beat, by how much, and whether beating (or missing) tends to repeat.
Filters. Two filters apply before a quarter counts. A quarter with a consensus estimate under 5¢ is dropped, because dividing by a number that close to zero turns a penny of upside into a triple-digit "surprise." Any surprise beyond ±200% is also dropped, for the same reason. This keeps a few near-zero-EPS names from dominating a mean or leaderboard meant to describe the other 490-odd companies.
Calendarized quarters. A fiscal period ending in a given three-month calendar window (Jan–Mar, Apr–Jun, Jul–Sep, Oct–Dec) is grouped together, so a September-year-end company like Apple lines up with a December-year-end peer reporting a similar period. It's an approximation. The "latest quarter" for the headline stats, sector breakdown and biggest-beat/miss tables is the most recent calendarized quarter where at least half the mapped universe has reported, because early in a reporting season the first few dozen reporters aren't representative.
Persistence. The test pools every consecutive pair of a company's own filtered quarters across the whole panel and asks: given this quarter was a beat, was the next one also a beat? How does that compare with the same question after a miss, and with the overall beat rate? The z-statistic and p-value test whether the after-a-beat and after-a-miss rates differ by more than sampling noise. It's a standard two-proportion z-test, not a claim about causation or a trading signal, and it says nothing about price reaction. That would need a price series around each report date, which this page doesn't include.
Recomputed weekly (Saturday) across all ~500 constituents. There's no page history to accumulate, because Alpha Vantage's EARNINGS endpoint returns years of quarterly history in one call. The trend chart is historical data recomputed each run.