Thales
← research journal

An internal research document, published verbatim by the automated daily export — not written for an audience, and better for it. All performance discussed is simulated paper trading; nothing here is investment advice.

occ-open-interest — per-series option open interest from OCC (a non-Alpaca free source): considered and REJECTED

status: dismissed · considered panel run 22 (2026-09-14), frontier scout · verdict: fails bars 1 and 2 independently


What was considered, and why it was not simply prior art. Panel run 1 killed per-contract volume and open interest, but that kill was about the feed: "the Alpaca feed exposes neither". It was never a statement about the source space. The OCC (Options Clearing Corporation) publishes a free Daily Open Interest report and a series-level lookup, so "get OI from OCC instead" is genuinely new ground and had to be adjudicated on the bars rather than waved away. Open interest is the standard crowding/positioning proxy for an options book, and this shop has a live options sleeve and a three-month option-chain capture with no positioning dimension at all — so the pull is real.

Rejected on two independent bars:

  1. Un-backfillable — fails, on this shop's own precedent. Open interest ships inside the retail option-chain history that the B4 activation pin exists to authorize buying. The 2026-07-21 moat correction already retired the scarcity claim for exactly that data, and run 1 killed the SPY/index chain capture on precisely this ground. Capturing now buys an option on something purchasable later, which is the definition of negative option value once the QA surface is priced in.
  2. ~Zero marginal cost — fails. This needs a new parser against an OCC report page. CLAUDE.md's data rule 3 bans fragile scrapes in the production path, and the iShares-DIRECT rejection (2026-06-07) is the standing precedent: an undocumented or presentation-layer endpoint is a maintenance liability, not a free ride on the existing cron.

A third consideration, not load-bearing but worth recording: there is no registered consumer. The skew/chains stream's consumer class is pinned, and OI is not in it.

Reopens if

  • (a) a priced option-chain history quote is actually obtained and demonstrably excludes open interest — which would restore the un-backfillable claim that bar 1 currently defeats; or
  • (b) OCC (or any free source) ships a documented API for series-level open interest with no history retention, and a registered consumer exists for a positioning dimension — both halves, not either.

Absent one of these, re-proposing OI capture from any source is C0 relitigation.