13F Options Explained: Puts, Calls, and the Notional Trap

How options actually appear in a 13F

Institutional managers with over $100 million in covered US equities must file Form 13F within 45 days of quarter end, and the form covers more than common stock. Exchange-traded options on 13(f) securities are reportable too, with one rule that surprises almost everyone the first time: per the SEC's instructions, a manager holding puts or calls reports the number of shares of the underlying security the options represent and the market value of those underlying shares. Not the option premium. Not the contract's market price. The notional value of the stock underneath.

Concretely: a fund that paid a few million dollars in premium for puts covering 1,000,000 shares of a $100 stock reports a row valued around $100 million. The economic exposure, the capital at risk, and the delta are all invisible. The SEC's official Form 13F FAQ and the form instructions spell out the mechanics.

Two asymmetries are worth tattooing somewhere visible. First, only held (long) option positions are reported. Written options, short stock, and most other short exposure never appear in a 13F at all. Second, the same underlying can occupy three rows in one filing: a share row, a call row, and a put row, each with its own value. Naively group by issuer and sum, and you triple-count notional that may partly hedge itself.

In Arkolith's Q1 2026 dataset (1,824 institutional filers, 1.87 million long positions, $53.7 trillion in reported value), the long book is tracked separately from option legs for exactly this reason. If the form itself is new to you, start with how to read a 13F filing.

The notional trap: puts counted as longs

The standard failure mode looks like this. A pipeline parses the 13F information table, runs a SUM over the value column grouped by issuer, sorts descending, and publishes "Fund X's top 10 holdings." If the fund's largest row is a put, the table now presents a bearish position as the fund's number one long. The sign of the position is not merely off in degree. It is inverted.

A well-known case made this trap famous: a prominent contrarian manager's index put options were headlined across financial media as a billion-dollar-scale bet against the market, when the filing disclosed only the notional value of the underlying shares. The premium actually at risk was never disclosed and was plausibly a small fraction of the headline number. The mirror-image error happens constantly at smaller scale, where a put leg quietly ranks as a fund's "top holding" in scraped aggregates, and the tell is usually a share count or value that looks implausible next to the fund's real size.

AI agents are unusually exposed here. An LLM summarizing a raw filing, or reading an aggregator that never segregated option rows, will state with full confidence that "the fund's largest position is X" when X is a hedge. The model has no way to know the data was mis-signed upstream, and no prompt fixes a schema problem. This is one of the cleaner arguments for grounding agents in structured, labeled filings data instead of letting them free-read EDGAR text: every Arkolith datapoint carries provenance back to the SEC accession number, and option legs arrive already labeled.

Reading the putCall field correctly

Each row of the 13F information table carries nameOfIssuer, a CUSIP, value, sshPrnamt (the share or principal amount), sshPrnamtType, and optionally putCall. The decision rule fits in one table:

Three working rules follow:

1. The long book is rows where putCall is empty. Every portfolio weight, concentration figure, and "top holdings" list should be computed on that subset only.
2. Never net option notional against share positions. A put on shares the fund also owns might be a protective collar leg, an outright bearish bet sized in premium, or one side of a spread whose written leg is invisible. Without strike and expiry you cannot compute delta, so netting by notional produces a precise-looking number with no meaning.
3. Treat calls as a soft signal and puts as a softer one. A held call at least agrees in direction with a long. A held put could be insurance on a position you can see in the same filing, or a standalone short thesis. The form cannot distinguish them, so neither can you.

What option rows still cannot tell you

Even read correctly, 13F option data is coarse. No strike, no expiration, no premium, no Greeks. You cannot tell a deep in-the-money call used as a stock substitute from a cheap out-of-the-money tail hedge, even though they imply very different convictions. Confidential treatment can delay some positions entirely. And the 45-day lag (2026 deadlines: Feb 17, May 15, Aug 14, Nov 16; the full deadline calendar is here) means listed options, which often run on monthly or quarterly cycles, may have expired or been rolled before you ever see the row. Academic work on 13F-based strategy replication has generally either dropped option legs or modeled them separately, precisely because notional values wreck portfolio weights.

What helps is triangulation. Insider activity on Form 4 must be filed within two business days of the trade, so an insider page like TSM's gives a far fresher directional read than a stale option leg; Arkolith tracks 51,000+ insider transactions alongside the 13F spine. On the ownership side, a stock page like NVDA shows institutional holders with put and call legs broken out, so a put-heavy quarter reads as what it is rather than inflating the apparent long base. For a fuller treatment of staleness, amendments, and the other structural limits of the form, see how accurate 13F data really is.

Querying option legs without the trap

Arkolith's API keeps the two books separate so the failure mode above cannot happen silently. Fund holdings endpoints return the long equity book by default, with option positions surfaced as a distinct labeled set, each row carrying its put or call direction and provenance to the accession number it came from.

Find a fund, then pull its holdings:

curl -H "Authorization: Bearer YOUR_KEY" "https://arkolith.com/api/v1/search?q=berkshire"

curl -H "Authorization: Bearer YOUR_KEY" "https://arkolith.com/api/v1/funds/0001067983/holdings"

Long rows and option legs come back as separate structures, so a sum over holdings is a sum over actual longs. The same separation applies in the MCP server: an agent asking "what are this fund's top positions" gets the long book, with puts and calls labeled rather than laundered into it. Setup takes a few minutes via the quickstart.

If you build your own parser instead, the contract is simple: split on putCall at ingest, store direction as a first-class column, and never let an aggregate query touch both books at once. The notional trap is not subtle once you know it exists. It persists because the raw form makes the wrong aggregation the easy one.

Frequently asked questions about 13F options

Are options included in 13F filings?

Yes, but only held (long) puts and calls on Section 13(f) securities. Written options, short stock, futures, and most swaps are not reported at all. Each option row is designated "Put" or "Call" in the filing's information table.

Does a put in a 13F mean the fund is short the stock?

Not necessarily. A held put is bearish in isolation, but it may be insurance on a long position visible elsewhere in the same filing, or one leg of a spread whose written side is invisible. Without strike and expiry you cannot compute the net exposure, so treat puts as a hedge-or-short ambiguity rather than a confirmed short.

Why do 13F option values look so large?

Because the form requires reporting the market value of the underlying shares, not the option premium. A modest premium outlay controlling a large share count produces a row that can dwarf the fund's real capital at risk, which is why option legs distort any naive value ranking.

Can you see option strikes and expiration dates in a 13F?

No. The filing discloses only the underlying security, the share count the options represent, the value of those underlying shares, and the put or call designation. Strike, expiry, premium, and whether the position even survived the 45-day reporting lag are all undisclosed.

This article explains public filings and data concepts. It is not investment advice.