Options Data v2: the 2002–2022 Archive, Rebuilt
We are regenerating our entire end-of-day options archive with self-computed Greeks and IV, per-row quality flags, and a strict rule: a value we could not compute is left blank and flagged — never written as 0. This page explains what changes for you as a buyer, and lets you test real v2 files before it ships.
Why rebuild it? An honest accounting
The current archive's quotes, volume and open interest are solid — they are exchange records. Its derived columns are not. Like most low-cost vendors, our upstream source wrote a Greek it failed to compute as 0. A delta of exactly 0 looks legitimate, so it slips through null-filters silently: if you select strikes by delta, those rows drop out of your universe without an error.
Measured on the current archive, 2008-06-16: 10.1% of SPY option rows and 7.8% of QQQ rows carry a delta written as 0. During the November 2008 volatility peak the rates were 6.8% and 6.9%. That is what v2 eliminates.
What changes, from a buyer's perspective
| Problem in the current archive | v2 behavior |
|---|---|
Greeks/IV written as 0 when computation failed | Self-computed (Black-76 with parity-implied forward for European; 200-step binomial for American). Failures are blank + flagged, never 0 |
| No way to tell "missing" from "failed" from "fine" | New iv_flag column (0–7) states exactly why a row has or lacks IV. One buyer rule: iv_flag ≤ 1 means IV and all five Greeks are present |
| Single mid-quote IV, no reliability signal | Three-leg solve: iv_bid and iv_ask bracket the mid IV — a quoted, model-implied IV range whose width tells you per-row how much to trust it |
close carries stale prices forward on no-trade days | close only when volume > 0; otherwise blank. New last_trade_date tells you when the contract last actually traded |
| Settlement timing invisible; time-to-expiry over-counted near expiration | New settlement_time (AM/PM) column, measured empirically per root and expiration. Day counts correctly handle A.M.-settled index options and pre-2015 Saturday expirations |
| 18 columns | 33 columns — adds rho, trade_vwap, transactions, multileg_volume, active_minutes and more (later columns start where their source data begins; blanks before that are honest, never zero-filled) |
What stays the same: contract identifiers, quotes, volume, open interest and dates are byte-identical to the current archive — verified row-by-row on sampled days from every year 2002–2021 — so your existing backtests remain directly comparable. Two deliberate exceptions, stated plainly rather than buried: the style (exercise style) column is corrected for a handful of index roots the current archive mislabels, each change backed by the exchange's own contract specification; and the first half of 2022 is rebuilt from a different upstream source than the current files, so rows there are not expected to match one-for-one.
How it is validated
Every monthly file must pass a 14-point consistency gate before release — row conservation against the current archive, quote sanity, Greek domain checks, IV monotonicity (iv_bid ≤ iv ≤ iv_ask), and flag–value consistency on every row. So far, more than 40 distinct trading days spanning every year from 2002 to 2022 — over 30 million rows — have been generated and passed the gate. The issues that sampling uncovered (a 2009 symbol collision in the source data, sentinel zeros in underlying prices, missing index-root classifications) were each fixed and re-verified before we moved on.
Test it yourself — free v2 sample days
Two complete trading days in the exact v2 format (33 columns, gzip-compressed CSV; opens with 7-Zip, WinRAR or gunzip):
- v2sample_2008-06-16.csv.gz — 283,843 rows, ~10 MB. Financial-crisis era: includes 398 rows of Lehman Brothers (
LEH) and 178 of Washington Mutual (WM) chains, three months before either failed. - v2sample_2020-06-16.csv.gz — 1,163,146 rows, ~40 MB. Modern market breadth.
One honest note: last_trade_date in these preview files is cold-started on the sampled day itself. In the shipped archive it accumulates from 2002, so coverage there will be much higher.
Early access & upgrade policy
The full 2002–2022 regeneration is in progress; monthly files publish only after passing the gate, and we will not promise a date we might miss.
If you would like early access to finished years before the official release — or want to be notified when v2 ships — email info@historicaldata.net with subject "v2". Tell us which years matter to you; early-access interest directly decides which years we finish first.
Existing options-data customers: when v2 ships, you can obtain the full updated archive for 30% of list price with your original order.