To build a dealer book from the options tape you must decide, print by print, who bought. For simple orders the quote rule does that. But on the day we measured, 40.2% of all SPX option prints were legs of multi-leg packages — and on those legs the quote rule is invalid by construction: an exchange fills a spread at a net price and allocates it across legs by convention, so a leg can print anywhere inside (or outside) its own market regardless of who initiated.
Our terminal signs the package: legs printing on the same millisecond are reassembled into their parent order, the package's net price is compared to a net bid/ask derived from the leg quotes, and the quote rule is applied at that level. The idea is not unique to us. What has never existed — for ours or anyone's scheme — is an accuracy number.
The truth set. Cboe publishes a free trade-by-trade sample: 3% of executions, each carrying the side, the capacity (customer / market maker / firm), the trade type, and an id linking legs of one complex execution. That is per-leg ground truth. Joined to our tape: 12,422 complex legs carry a verdict.
The number:
| rule, graded on customer legs | per print | volume-weighted |
|---|---|---|
| package signer, as shipped | 80.4% | 79.1% |
| excluding legged-in contras (unrecoverable in principle) | 87.9% | 84.7% |
| single-leg quote rule on the same legs | 75.1% | 75.1% |
| coin flip | 50% | 50% |
Iron condors grade at 88%, strangles at 91%. Even the legged-in category — which Cboe itself declines to attribute — comes out at 69% rather than the coin flip we budgeted.
The mirror check. A rule that claims to find the customer should be reliably wrong about the market maker on the other side. It is: on legs where the sampled participant is a market maker, the rule agrees with that side only 10.9% of the time.
The semantics finding. Against time-priority truth the rule fails (16.9% vs true liquidity removers) while scoring 90.4% against auction initiators. Both make sense at once only one way: the package quote rule identifies the side that paid the spread against the derived net mid, not the side that arrived second. For a customer-versus-dealer book, that is not a bug — it is the reason the rule works.
One truth day, binomial SE ~0.7pp; the signer abstains on ~45% of complex volume rather than guessing. Full study: How accurate is complex-trade signing?
