Forced Selling vs. Fundamentals: How to Tell a Positioning Event From an Information Event

By Jake, founder of StockResearch.app

Every violent drawdown in a stock is, at bottom, one of two things:

1. An information event. The market learned something that lowers the present value of the business: a lost customer, a guidance cut, a competitor breakthrough, a regulatory hit. The old price was based on beliefs that are now wrong. 2. A positioning event. Nothing changed about the business, but something changed about who owns it and how. A leveraged holder got a margin call. A fund faced redemptions. An index dropped the name. Risk limits forced a desk to cut. The selling itself sets the price.

The reason this distinction matters is that the two events have opposite implications for what the new price means. After a true information event, the lower price reflects new facts. After a pure positioning event, the lower price reflects a temporary imbalance between one motivated seller and the available buyers. History suggests those imbalances tend to close when the seller finishes.

The reason this distinction is hard is that the price chart of the two events is identical. A stock down 45% in three weeks looks exactly the same whether the cause was a broken thesis or a broken balance sheet at a hedge fund the company has never heard of.

July 2026 supplied a rare teaching case: the Situational Awareness fund liquidation, where a 4x-levered, five-name-concentrated AI book was force-sold into a falling market and the whole AI-infrastructure complex de-rated 40–60% in a month. Because we now know exactly what was in the book, when the margin calls came, and when the selling stopped, we can check every diagnostic against a known answer key. Here are the five checks I find most useful, with the July evidence for each.

Check 1: Guilt-by-association breadth

The question: are stocks falling because of what they do, or because of what category they belong to?

Information travels along fundamental lines. If the market learns something bad about GPU cloud economics, it should hit GPU clouds hardest, memory suppliers somewhat, and an unrelated bitcoin-miner-turned-datacenter less, in proportion to exposure. Forced selling travels along ownership lines: whatever the distressed holder owns (and whatever similar funds own alongside it) falls together, regardless of fundamental linkage. Then sector ETFs and correlated books transmit the pressure to everything that merely rhymes with the distressed book.

The July evidence: on July 30, when the forced seller's book was removed in a single block, Cipher Mining, a name the fund never owned, bounced +28%, harder than the names actually dumped (Nebius +27%, IREN +26.5%, CoreWeave +22%). Cipher had been dragged down by category membership alone, and it snapped back the moment the category-level selling pressure disappeared. Information events do not work like that.

Check 2: The shape of the recovery

The question: when the selling pauses, does price V-bounce with no news?

After an information event, a stock finds a new, lower equilibrium and tends to stay there until the facts change again. Rallies need catalysts. After a positioning event, the moment the motivated seller finishes, the imbalance reverses violently. Because the last leg down was never about value, buyers reappear well above the lows, on no news at all.

The July evidence: the July 30 session. Five names up 22–28% intraday against SPY +1.8%, on a day whose only "news" was that a seller had finished selling. No earnings, no product announcements, no estimate revisions. The absence of a fundamental catalyst for the bounce is itself the data point.

Check 3: News-to-move proportionality

The question: does the stated catalyst plausibly justify the magnitude, and does it survive a second reading?

Real information events usually have a legible chain from headline to cash flow. Positioning cascades, by contrast, tend to launch off ambiguous catalysts that get read in whatever direction the flow is already pushing.

The July evidence: the selloff's trigger was Meta announcing a commercial compute offering on July 17, read as "hyperscaler dumping capacity, GPU glut coming." Gavin Baker, in an Invest Like the Best episode recorded during liquidation week, argued the market had it backwards: Meta was responding to SpaceX successfully selling clusters into the spot market at a premium, a data point about compute scarcity, not glut. Similarly, the open-source model panic that month (GLM, Kimi) arguably described a margin shift from the model layer to the infrastructure layer ("a token is a token") rather than demand destruction. You do not have to agree with Baker's readings to notice the pattern: when a 40–60% sector de-rating rests on catalysts that bear two opposite interpretations, positioning is doing a lot of the pricing.

Check 4: Is the fundamental data flow confirming?

The question: while the price falls, what are the actual reported numbers doing?

This is the strongest check, because it is the hardest to fake. If a selloff is informational, the deterioration should start showing up in the data: guidance cuts, decelerating growth, falling utilization, pricing weakness. If the numbers keep accelerating through the drawdown, you are watching price and fundamentals diverge, which is the signature of a positioning event (or of a market correctly anticipating a deterioration that has not yet arrived; more on that honesty below).

The July evidence: in the same week the liquidation bottomed, Azure reported +43% growth and Nvidia's commentary remained supply-constrained. Baker's summary of his own checking that week: he could not find a single negative quantitative metric. GPU pricing, DRAM spot, token growth: all accelerating. The market fell 40–60% while the measured fundamentals sped up.

Check 5: The leverage context

The question: how much borrowed money is holding this asset class, and is it stressed?

Positioning events need fuel. A market where leveraged holders are concentrated in the same trades is primed for forced-selling cascades regardless of fundamentals. The margin call, not the news, becomes the transmission mechanism.

The July evidence: US margin debt stood at a record ~$1.5 trillion, up 77% since April 2025, and the epicenter fund ran 4x gross leverage with 76% of its long book in five names. Prime-broker margin calls from Goldman, JPMorgan, and Bank of America were the proximate cause of the final capitulation, not any company's report. Notably, the credit stress bit at the investor layer, not at the hyperscaler balance sheets the bears were watching.

The honest caveats

Three things keep this framework from being a cheat code:

Sometimes it is both. Positioning events cluster around genuinely ambiguous fundamentals, because that is when leveraged conviction gets tested. A forced seller can be wrong about risk management and the market can still be right about over-earning. The checks above tell you what is driving this week's price, not whether the three-year bull case survives. Anticipation looks like divergence. A market pricing a future deterioration will diverge from trailing fundamentals for a while, and be vindicated later. "The numbers are still accelerating" is evidence, not proof. You rarely get the answer key. July 2026 is pedagogically perfect because the seller's book became public. In most drawdowns you will never know for certain who was forced. The checks are probabilistic: five weak signals pointing the same way beat one strong opinion.

Making the checks measurable

Checks 1–3 you can run from the tape and the news flow. Check 4 is where a tool helps: the cleanest single view of "price versus fundamentals" is a company's valuation multiple against its own history. When a stock falls 40% on unchanged estimates, its multiple compresses toward the bottom of its historical range; when it falls 40% because estimates fell, the multiple barely moves. We wrote a guide to using historical valuation multiples without overfitting, and the comparison tools chart daily P/E, P/S, and EV/EBITDA history for exactly this purpose.

The question July leaves every investor with: next time a sector you follow drops 40% in a month, will you be able to tell, with data instead of vibes, which kind of event you are looking at?

---

Jake is the founder of StockResearch.app, where he writes data-first research on valuation dislocations. This article is for informational purposes only and does not constitute financial advice. Nothing here is a recommendation to buy or sell any security. Always do your own research before making investment decisions.