- Q2 2026: puts $2.58B across 9 of 147 funds (6.1%) — 0.13% of the $2,009.21B long book.
- New puts scored: 253 events, average +2.49 points vs SPY at 63 days, median 0.00, 47.8% negative — cluster CI [−0.82, +6.26] includes zero.
- Controls: new calls −1.15 points (n=454), new long stock −0.20 points (n=24,221) — differences from puts include zero.
- Holdouts: new puts, surge puts, and combined puts all failed — validation intervals include zero or flip sign.
- Peak quarter: Q1 2026 puts hit $17.87B, then collapsed to $2.58B in Q2 — a one-off concentration, not a regime.
What a 13F put row actually is
A 13F put appears as an option row: the filing reports the underlying issuer, a PUTCALL flag, and a VALUE in dollars. That VALUE is the value of the underlying shares the options reference — it is not the premium the fund paid, not the maximum loss, and not a measure of conviction. The filing reports no strike price, no expiry, and no delta, so two $1B put rows can mean economically different things. A fund that buys puts as insurance against a long portfolio and a fund that buys puts as a directional bet file rows that look identical. Keep that ambiguity in mind for everything below: we are scoring disclosed rows, not identified intentions.
One more scope note. A 13F is a long-positions filing. Short stock, swaps, and over-the-counter derivatives do not appear, so a fund’s net directional exposure cannot be reconstructed from this data. What we can see is the disclosed put book, and that book is small.
How small puts are — and the one quarter they weren’t
In the latest quarter, 2026Q2, the 147 tracked funds reported $2.58B of puts in total, held by just 9 funds (6.1%), across 25 fund-underlier pairs and 21 underliers. The same funds reported $3.20B of calls and $2,009.21B of long stock. Puts were 0.13% of the long book. Over the full period, the share of funds reporting any puts in a quarter ranged from 3.7% to 11.7%, peaking at 15 funds (10.2%) in 2025Q3. The majority of tracked funds report no puts in any given quarter.
| Quarter | Funds with puts | Put notional | Long stock | Puts as share of long |
|---|---|---|---|---|
| 2024Q3 | 8 | $1.19B | $1,795.22B | 0.066% |
| 2024Q4 | 10 | $3.97B | $1,777.79B | 0.223% |
| 2025Q1 | 6 | $3.74B | $1,758.94B | 0.213% |
| 2025Q2 | 11 | $3.53B | $1,820.99B | 0.194% |
| 2025Q3 | 15 | $6.88B | $1,883.05B | 0.365% |
| 2025Q4 | 13 | $8.51B | $1,897.04B | 0.449% |
| 2026Q1 | 10 | $17.87B | $1,753.56B | 1.019% |
| 2026Q2 | 9 | $2.58B | $2,009.21B | 0.128% |
Put notional is the reported VALUE of underlying shares referenced, not premium paid. Source: 13F put panel, 147-fund roster; see Methodology.
The Q1 2026 spike was three positions
The highest put quarter on record, 2026Q1 at $17.87B, was not a broad defensive shift. It was concentrated in a handful of rows: Carronade Capital’s SPY put at $5.90B (its implied price of $650.34 matches the March 31, 2026 SPY close exactly), Situational Awareness’s SMH put at $2.04B, and its NVDA put at $1.57B. The next quarter, total puts fell to $2.58B. A series that multiplies sevenfold on three rows and then collapses is a concentration story, not a signal that the cohort turned bearish.
What the underliers say: mostly index insurance
Across the recent eight quarters, the largest put underliers are index and sector ETFs, not single stocks. SPY was the top put underlier in six of the eight quarters — 2024Q3, 2024Q4, 2025Q3, 2025Q4, 2026Q1, and 2026Q2 — with IWM, QQQ, HYG, XBI, IEF, and SMH also recurring. That composition fits the hedge reading better than the directional-bet reading: an index put is portfolio insurance, and it dominates both the notional totals and the event counts. Any analysis that treats all put rows as bearish single-stock bets misreads most of the data.
Scoring: new puts did not predict weakness
We scored every new put — a fund-underlier pair that was not a put holding in the prior quarter, for funds that filed in both quarters — from the close of the first trading day after the fund’s actual filing date, over 63 trading days, against SPY over identical dates. A put “works” if the underlying trails SPY (excess below zero). Of 476 classified new puts, 253 could be scored; the rest fell out for missing price or entry data (123 unmapped, 56 without prices, 39 without a valid entry, 5 without a full horizon — a 46.8% dropout rate, disclosed here because it limits what the average can claim).
| Group | Classified | Scored | Average excess | Median | Share negative | Cluster CI |
|---|---|---|---|---|---|---|
| New puts | 476 | 253 | +2.49% | 0.00% | 47.8% | [−0.82, +6.26] |
| Surge puts | 29 | 22 | −0.05% | 0.00% | 63.6% | [−6.70, +5.56] |
| Combined puts | 505 | 275 | +2.29% | 0.00% | 49.1% | [−0.77, +5.68] |
| New calls (control) | 823 | 454 | −1.15% | −1.50% | 55.1% | [−3.04, +0.90] |
| New long stock (control) | 40,323 | 24,221 | −0.20% | −1.00% | 53.1% | [−0.57, +0.19] |
The average new put was followed by the underlying beating SPY by 2.49 points — but every confidence interval includes zero, so the honest statement is not “puts were wrong.” It is: new put additions did not predict underlying weakness over 63 days; no edge was established. The differences against the controls also include zero (new puts minus new calls: +3.64 points, CI [−0.00, +7.98]; new puts minus new longs: +2.70 points, CI [−0.33, +6.68]). Surge puts — existing puts at least doubled with a $10M+ increase — averaged −0.05% on just 22 scored events; the sample is too thin for any directional claim in either direction.
The median of exactly 0.00% is structural, not a finding: 31 of the 275 scored combined puts are SPY-based, and their excess return is zero by construction (the benchmark is itself), and 55 events are index-ETF based overall. Excluding index ETFs, new puts still averaged +2.95% (n=209, cluster CI [−1.21, +7.38]). The conclusion does not depend on the index rows.
Holdouts: nothing survived
The holdout rule was fixed before scoring: derivation is event quarters 2016Q4–2021Q2, validation is 2021Q3–2026Q2, and a group survives only if the validation interval excludes zero with the derivation’s sign.
| Group | Derivation avg (n) | Validation avg (n) | Validation cluster CI | Verdict |
|---|---|---|---|---|
| New puts | +0.09% (141) | +5.52% (112) | [−0.23, +13.34] | Fail — CI includes zero |
| Surge puts | +4.09% (12) | −5.02% (10) | [−17.87, −0.75] | Fail — sign flip |
| Combined puts | +0.41% (153) | +4.65% (122) | [−0.51, +11.88] | Fail — CI includes zero |
The validation average for new puts (+5.52%) looks large, but it is concentrated in two quarters — 2025Q3 (+17.71%, n=18) and 2026Q1 (+8.01%, n=18) — and the interval still includes zero. Remove those two quarters and most of the validation strength disappears. That fragility is itself the finding: there is no stable put signal here to trade on.
Robustness: the answer does not move
Trimming the top and bottom 1% of new-put outcomes gives +1.70% (n=249, CI [−0.73, +4.32]). Entering on the filing day itself instead of the next trading day gives +2.57% (n=253, CI [−0.69, +6.09]). Excluding index ETFs gives +2.95% (see above). In no variant does the interval exclude zero or change sign. The null result is not an artifact of one scoring choice.
What to do with a put disclosure
First, check the underlier. If it is SPY, QQQ, or another index ETF, the default reading is insurance, not a view on a company — and index puts dominate this data. Second, check the size against the fund’s long book: the cohort’s entire put book is about a tenth of one percent of its long stock, so even a “large” put row is usually small relative to what the fund owns. Third, remember the lag: the row describes a position held at quarter-end and disclosed up to 45 days later; it is not evidence about what the fund holds today. None of that makes puts useless — concentration spikes like Q1 2026 are worth noticing — but a new put row, on its own, is not a short signal. In 476 tries, it wasn’t one.
Methodology
Data. 13F filings of the 147 tracked funds, 2016Q3–2026Q2 (40 quarters). The roster is the current 147-fund list backfilled, so funds that existed historically but are not on the current list are absent throughout — a survivorship bias that affects level comparisons across time. In 2016Q3, 111 roster funds filed. Prices are split- and dividend-adjusted closes; entry is the first trading day’s close after the fund’s actual filing date; horizon is 63 trading days; excess is the underlier minus SPY over identical dates. Uncertainty is a ticker-cluster bootstrap. Full scored event lists and the quarterly trend are downloadable below; every waterfall closes exactly in the package integrity file.
Definitions, fixed before scoring. Put notional is the reported VALUE of underlying shares, not premium paid. A new put is a (fund, CUSIP) pair that was not a put holding in the prior quarter, for funds filing in both quarters. A surge put is a put held in both quarters whose notional at least doubled with an increase of $10M or more. Controls (new calls, new long stock) were computed from the same panel by the same rules. A put “works” if excess is below zero.
Limits. No strike, expiry, or delta is reported, so hedging and directional intent cannot be separated. No claim about periods outside 2016Q3–2026Q2, about the 46.8% of new puts that could not be scored, or about tradability after costs and liquidity. No causal claim. Past patterns are not a promise of future results. For education only — not investment advice.
Data download
Quarterly put trend (funds, notional, calls, long book): quarterly_put_trend.csv. Scored put events with anchors, entry dates, and 63-day excess returns: scored_put_events.csv
Revision history
Oct 27, 2026: First published.