How to get historical options prices, expired contracts included

The short way

Ask for a past trade date, not for a chain. ThetaData's end-of-day endpoint takes expiration=* and returns a row for every contract on that underlying, for the date range you name, whether or not the contract still exists. Its REST API runs on the Theta Terminal, a Java program on your own machine that holds the login, so the request carries no key:

import io

import pandas as pd
import requests

BASE = "http://127.0.0.1:25503/v3"

params = {"symbol": "AAPL", "expiration": "*", "start_date": "20241104",
          "end_date": "20241104", "format": "csv"}
r = requests.get(f"{BASE}/option/history/eod", params=params, timeout=120)
r.raise_for_status()
chain = pd.read_csv(io.StringIO(r.text))

expired = chain[pd.to_datetime(chain["expiration"]) < pd.Timestamp("2026-10-09")]
print(len(chain), "contracts,", len(expired), "since expired")
print(chain[["expiration", "strike", "right", "close", "bid", "ask", "volume", "last_trade"]].head())

That is the vendor's own sample request. Each row has the day's open, high, low, close, volume and trade count, and the last NBBO bid and ask with sizes, exchanges and conditions. Read the created and last_trade columns before you use close: OPRA publishes no national end-of-day report, so ThetaData generates its own at 17:15 Eastern, and close is the last trade, whenever that was. A free account gets this from 1 June 2023, one day late; Value, Standard and Pro reach 2020, 2016 and 2012.

To find what existed on a date without pulling prices, /v3/option/list/contracts/quote?symbol=AAPL&date=20220930 lists every contract quoted that day. The same call with trade lists every contract that traded.

What the options are

A contract index with a flag for the dead ones. Massive's /v3/reference/options/contracts lists active contracts by default. expired=true returns the expired ones, and as_of asks what the index looked like on a past date. Each result carries an OCC-style ticker such as O:AAPL211119C00085000, the shares_per_contract, and any additional_underlyings a corporate action added to the deliverable. Prices are then one request per contract:

import requests

BASE = "https://api.massive.com"
AUTH = {"apiKey": "YOUR_API_KEY"}

listing = requests.get(f"{BASE}/v3/reference/options/contracts", params={
    "underlying_ticker": "AAPL", "expiration_date": "2024-11-15",
    "expired": "true", "limit": 1000, **AUTH}, timeout=30).json()

for c in listing["results"][:5]:
    bars = requests.get(
        f"{BASE}/v2/aggs/ticker/{c['ticker']}/range/1/day/2024-10-01/2024-11-15",
        params={"adjusted": "false", **AUTH}, timeout=30).json()
    print(c["ticker"], c["shares_per_contract"], len(bars.get("results", [])), "daily bars")

The listing pages through next_url at 1,000 contracts. Bars are on every Options plan, built from trades, so a day with no trade has no bar. Historical quotes need Options Advanced, $199 a month. History runs two years on Basic and Starter, four on Developer and five-plus on Advanced.

The OPRA record, billed by the byte. Databento's OPRA.PILLAR dataset has trades, OHLCV bars, statistics, instrument definitions and the one-minute consolidated BBO from 1 April 2013. Tick-level NBBO and quotes immediately before each trade start on 28 March 2023. Parent symbology (stype_in="parent", symbols="AAPL.OPT") fetches every contract on an underlying, and the definition schema says which contracts existed on a given day. There is no subscription needed for history: it is billed per gigabyte, which makes one underlying for one month cheap and the whole market for ten years not.

The exchange's own files. Cboe DataShop's Option EOD Summary covers every OPRA series from January 2012, as one file per day. Each row has two snapshots: the NBBO with sizes and the underlying's bid and ask at 15:45 Eastern, and open, high, low, close, volume, VWAP and open interest at the end of the day. IV and greeks are an add-on. In the cart on 13 September 2026, a full-market day was $50, $70 with IV and greeks, and capped at $300 for a calendar month. The whole 2012-to-date archive was $7,200.

A cleaned surface by trade date. ORATS' /datav2/hist/strikes takes a ticker and a tradeDate back to 2007. It returns every strike listed that day, calls and puts on one row, with bid, ask, the smoothed IV and greeks. /datav2/hist/strikes/options follows one contract by expiry and strike. Delisted tickers are kept. The price is ORATS's: quotes captured 14 minutes before the close. One-minute history from August 2020 is the $599 plan.

End of day by the file, with a subscription or without. IVolatility's API reaches January 2005 for over 5,500 underlyings, delisted names included. Its end-of-day price is the mid of the closing best bid and offer, published after 17:00 Eastern, with a 15:45 snapshot kept beside it. Without a subscription, its Data Download tool sells NBBO option prices at $0.40 per ticker per day.

Where this breaks

The chain forgets. A live chain, which is what getting an options chain covers, is the list of contracts that exist now. Anything that has expired is gone from it. The same is true of a contract list cached last month. Massive's index needs expired=true to show dead contracts. ThetaData's expirations and strikes lists are "updated overnight", and its contracts-by-date list is the reliable way to ask what traded on a past day.

A split rewrites the contract, and sometimes the symbol. OCC adjusts listed options after corporate actions, case by case, and announces each adjustment in an information memo. For NVIDIA's 10-for-1 split, effective 10 June 2024, memo #54623 kept the NVDA symbol, divided strikes by 10 and multiplied the number of contracts by 10. A contract you followed under one strike on 7 June trades under a strike one tenth as large on 10 June, so a join on the OCC symbol ends one series and starts another. For the 1-for-3 reverse split of QQQY on 1 August 2024, memo #54939 changed the symbol to QQQY1. The new deliverable was about 33 shares plus cash, and the memo priced QQQY1 at 0.333333 times QQQY. Treat a 1 root as an ordinary 100-share contract and its IV, delta and P&L are all wrong. Massive's shares_per_contract and additional_underlyings fields are there for this. IVolatility's one-minute data leaves these non-standard series out altogether.

The symbol is spelled four ways. OPRA's raw symbols are OCC's 21-character format: root right-padded with spaces to six characters, YYMMDD, C or P, and the strike times 1,000 in eight digits, as in AAPL 241115C00170000. Massive drops the padding and adds O:. ThetaData never uses the string at all, and takes symbol, expiration, strike in dollars and right as four fields. Index options add root splits. ThetaData's symbology page separates SPX from SPXW, RUT from RUTW and NDX from NDXP by settlement style, and notes that SPXPM stopped being used on 21 December 2018. Normalise to (root, expiry, right, strike in thousandths), and keep the root, because SPX and SPXW with the same expiry and strike are different contracts.

"End of day" is four different times. ThetaData's report is generated at 17:15 Eastern with the last NBBO at that moment. Cboe snapshots the quote at 15:45 and the trades at the close. ORATS captures quotes 14 minutes before the close. IVolatility uses the closing NBBO, with a 15:45 snapshot beside it. Put any of these beside a stock's 16:00 close and the implied volatility you solve for absorbs the gap; why implied volatility differs works through how much. Open interest has its own clock. ThetaData documents it as reported by OPRA at about 06:30 Eastern and representing the end of the previous trading day, so the figure on Tuesday's row is Monday's.

Most contracts did not trade. ThetaData's own sample row has a last trade at 15:48 and a report at 17:16. On a deep out-of-the-money contract, zero OHLC means no qualifying trades, not missing data, and Massive produces no bar at all. A daily close series built from trades is a series of stale prints. Use the bid and ask, and keep the timestamp of the price you used.

The volume is OPRA-sized, and so is the bill. Every strike of every expiry of every underlying, every day, is why the history is sold as files and per gigabyte. One trading day of full-market 1-minute quotes on Cboe DataShop was $740 in the cart on 13 September 2026. The personal licences on ThetaData, Massive, IVolatility and Databento's lower plans do not cover a product other people see. OPRA's own redistribution fee is the line that ends side projects, and why options data costs more than stock data and the options data cost calculator put numbers on it.

If you outgrow this

When you need the surface rather than the prices, the question becomes implied volatility at fixed tenors, not contract by contract. How to get historical implied volatility as data covers the vendors that sell it ready-made.

When you need ticks rather than end-of-day, the choice is between Databento's per-gigabyte OPRA record, ThetaData's tick tiers and Cboe's per-day trade files. Plan the storage before the download; storing tick data covers the shapes that work.

When the record is your own trades, the problem is the broker's statement, not an archive. Recording options trades in a journal covers assignment, rolls and expiry.

Everything in the catalogue that touches options is on options data, and the vendors bought for options first are in options data and flow analytics. OPRA and redistribution explain the licence terms behind every price on this page.