Earnings Implied vs Realized

Your agent translates this into per-ticker running hit rates, implied moves measured against realized ones, and the event counts behind each number.

What the market charged, and what the stock did.

Before a company reports, the options market puts a price on how far the stock is about to move. Buying the call and the put at the strike closest to the current price, a pair known as a straddle, costs roughly what a move of that size is worth, so the price of that pair is the market's implied move. Each row captures that price on the last session before the report, then measures what the stock actually did on the session that prices the reaction.

One row is one earnings event for one company. Alongside the event it carries that ticker's record up to that date: how many events have been observed, how often the straddle turned out to be too expensive, and the trailing averages of implied and realized moves.

Trade the gap between the implied move and the real one.

The reason to care is the gap. Across 3,554 earnings events since 2024, the straddle overpriced the realized move 64% of the time, worth roughly 1.1% of spot per event before costs. That is the volatility risk premium showing up somewhere you can date precisely, one report at a time.

The market-wide average is the starting point, not the trade. The per-ticker fields are what turn a statistic into a screen: names whose straddles have been consistently expensive, names that consistently break through them, and the ones whose behavior changed partway through. Because the statistics are point-in-time, you can rank on what was knowable the morning of the event instead of on the full-sample answer.

• Screen ahead of earnings season on hit_rate_to_date, with n_events_to_date as the sample-size guard.
• Compare implied_move_pct against avg_realized_abs_to_date to find names priced above their own history.
• Study the distribution of move_ratio rather than its average, since the tail is what sizes the position.
• Split results by time to keep before-the-open reports separate from after-the-close ones.
• Join eps_actual to eps_estimated to test whether surprise size explains the events that broke through.

Priced at the close before, measured at the close after.

The straddle is priced from the option chain on capture_date, the last trading session before the announcement, using the strike nearest spot and the expiration in exp_date. The realized move is measured on reaction_date, the session that prices the reaction. time records whether the company reported before the open or after the close, which is what decides which session is which.

Rows are written daily after the close. An event whose reaction session has not closed yet arrives with the straddle fields filled and the realized fields empty, so no outcome is ever stamped onto a date where it was not known. Once the reaction session resolves, the values are fixed and are not revised. The running statistics count only events up to and including their own row, which is what makes them safe to rank on inside a backtest.

The fields that separate the price from the outcome.

The headline fields below are a subset. Every field, with exact types and semantics, is documented in the API reference.

One ticker, its whole earnings track record.

One key works across the REST API, the hosted MCP server, and this dashboard. Every response is JSON with a { count, data } envelope.

import requests

url = "https://api.alphanume.com/v1/earnings-move-history"
params = {
    "ticker": "AAPL",
    "date_gte": "2024-01-01",
    "api_key": "alp_abc123"
}

r = requests.get(url, params=params)
print(r.json())

curl "https://api.alphanume.com/v1/earnings-move-history?ticker=AAPL&date_gte=2024-01-01&api_key=alp_abc123"

{
  "count": 1,
  "data": [
    {
      "date": "2026-01-29",
      "ticker": "AAPL",
      "time": "amc",
      "capture_date": "2026-01-29",
      "reaction_date": "2026-01-30",
      "spot": 232.14,
      "atm_strike": 232.5,
      "exp_date": "2026-01-30",
      "days_to_exp": 1,
      "atm_call_px": 4.85,
      "atm_put_px": 4.70,
      "straddle": 9.55,
      "implied_move_pct": 4.11,
      "implied_move_dollars": 9.55,
      "atm_iv": 0.68,
      "realized_return_pct": -2.34,
      "realized_abs_move_pct": 2.34,
      "over_under_pct": 1.77,
      "move_ratio": 0.57,
      "overpriced": true,
      "eps_estimated": 2.35,
      "eps_actual": 2.41,
      "notional_volume": 184320000,
      "n_events_to_date": 18,
      "hit_rate_to_date": 0.67,
      "avg_implied_move_to_date": 4.42,
      "avg_realized_abs_to_date": 3.61,
      "avg_over_under_to_date": 0.81
    }
  ]
}

What this data does not claim.

• Realized fields stay empty until the reaction session closes, so an event pulled the morning of the report has the straddle priced and the outcome blank.
• The straddle is a mark at the close, not a fill. Spreads, commissions, and slippage all come out of any edge measured from it, and the published figures are gross of costs.
• hit_rate_to_date on a name with a handful of events is close to noise. The overpricing tendency is market-wide first and name-specific second, so check n_events_to_date before ranking on it.
• Once an event resolves, its values are fixed and are not retroactively altered. That is what keeps backtests clean, and it also means the row will not quietly change under you later.
• Pro includes full history and the latest session.