What is SPX option flow actually made of?
Execution channels from the tape's own condition codes, 2022-2026
The tape does not just say what traded — every print carries an OPRA condition code naming the pipe it came through: the electronic book, the complex order book, a price-improvement auction, a multi-leg order legging into the simple book, or a human being on the physical floor in Chicago. Nobody appears to have published the composition for SPX, the one product where the answer is unusual. So, tabulated, with cancels excluded and every code mapped:
The near book, by year
| share of volume | 2022 | 2024 | 2026 |
|---|---|---|---|
| electronic simple book | 34.2% | 34.8% | 26.0% |
| complex order book + complex auctions | 36.1% | 39.1% | 44.7% |
| physical floor | 20.8% | 19.8% | 20.6% |
| AIM price-improvement auctions | 2.9% | 2.9% | 6.7% |
| legged / late / other | 6.0% | 3.4% | 2.0% |
Three things stand out. First, the floor is not dying: a steady fifth of the near book trades by open outcry, five years running — which independently matches the one recent academic estimate (Galati, Hendershott, Khan and Riordan: over 20% of U.S. options floor volume, "almost exclusively" S&P 500 index options). The dominant floor code is the proprietary-product multi-leg print — big complex structures negotiated by humans. Second, the complex book keeps eating the simple book: 36% to 45% while simple electronic fell to 26% — which is why this terminal signs packages as packages. Third, the auction share doubled, and that one deserves its own section.
The auction slice is retail, and it is measurable
Cboe publishes a free trade-by-trade sample carrying, for each execution, the participant's capacity. Joined against our tape it says: the initiator of an AIM price-improvement auction is a Customer 97.8% of the time, the contra side is a market maker 95.6% of the time, and 100% of the initiating volume is 10 contracts or smaller — the mechanism's size cap, visible in the data. Small-lot auction internalization against a market maker is the closest thing the SPX tape has to a printed retail label. That slice has doubled since 2022, and on the 0DTE expiry it reached almost 10% of volume this year. It is narrow — about 4% of all customer volume routes this way — so it is a pure tracer of retail direction, not a census of it.
One more composition fact
The tape-signed dealer book adds a fact no condition code shows: at the close, the measured 0DTE book leaves dealers net LONG gamma on 83% of sessions. The crowd's structural habit is selling same-day optionality into the close, and the dealer side of the tape wears it every afternoon.
Honesty section
The five-year series is the NEAR book (the three expiries behind today's) because that is the tape we hold in full; the 0DTE columns are spot-checked on the sessions held locally and the shares are similar in shape (floor much smaller, auctions larger). The capacity numbers come from one truth session — the sample is free, more days are not. "Retail" here means small-lot auction internalization; a retail order routed straight to the book wears no label and is not counted.
Reproduce it
The truth file is Cboe's C1 trade-by-trade 3% sample (free from Cboe DataShop). The per-print tape with condition codes is licensed data — ThetaData in our stack. The channel map is the OPRA condition-code enum; cancels (codes 40-43) are excluded; every unmapped code is tallied rather than dropped, and the unmapped residual is zero.
Part of gex.live research. Measured on the free session archive; every session is free to replay.