Looks like a signal, isn't one yet: how the Desk sorts evidence from noise

A filing hits the wire. A trader posts a bold call. Ten sites run the same headline. Each one feels like a signal. Most of the time, it isn't one yet. This piece walks through real examples from the past few weeks where something looked like evidence, and what the Desk's teammates did to find out whether it actually was.

DePappa Capital trades only its own capital. Nothing here is investment advice, a recommendation, or an offer of any kind. Everything below is paper research: no trade was placed because of any of it, and there are no results to report.

A dare is not a trade

On a Friday this month, a trader we follow told his followers Nvidia would hit $220 by the end of the day. Our scout logged it as not actionable. There was no entry, no stop, and the call expired at the close.

That could have been the end of it. But the owner pointed out something about this particular trader: he trades off options-market levels and never posts entries or stops. Judged by a generic checklist, every call he makes would fail. So the rule changed. A clear call on a single stock from him now counts as a lead, and our chart teammate supplies the levels from data the Desk saves itself. The lesson is to judge a source by how that person actually trades, not by a one-size-fits-all checklist.

Most of the internet is noise

Broad public searches on X turned up almost nothing usable. On Monday afternoon, October 19, nearly every result was a promotion for a WhatsApp group or a copy-pasted watchlist. Most real leads come from a short list of traders we follow, and even those get tested before they mean anything. When a broad search does turn up something, like a post citing a date for a Solana network upgrade, it is logged as one person's opinion until it is checked.

Ten sites, one source

Our news teammate starts with the original source: the company filing, the company's own release, or a major wire. Social posts come after. It also checks when an event actually happened, not just when a story was published.

Repetition is not confirmation. If ten sites repeat one claim that traces back to a single report, that is one source. The system itself enforces this. When a claim was labeled "corroborated" but only one independent outlet had reported it, the label was rejected and changed to "unconfirmed."

Earlier this month a GameStop post shows the other half of the rule. A filing confirmed that an insider had bought shares, so that part was treated as fact. The poster's price target was kept labeled as opinion. One true fact in a post does not make the rest of the post true.

Filings that look like news but aren't

Ownership filings are some of the easiest things to misread. Our ownership teammate looked at an amended 13D filing on CVRx from a large holder. A 13D is the form big holders file once they own more than 5% of a company. At a glance, an amended filing looks like fresh news. This one was an exit. The holder's remaining stake was about 0.6%, below the 5% line, which means they no longer have to report. It was the end of a position, not a new bet.

Vanda Pharmaceuticals was a similar story. There had been no new insider buying or financing since late September. The recent filings were a stock award, a gift of shares, a passive fund reporting about a 5.6% stake, and a shelf registration, which lets a company sell shares later. None of those is an insider buying stock in the open market with their own money. They were filed as background, not as a signal.

The lesson: awards, gifts and exit filings look like activity but don't show conviction. Filings also arrive with a delay, so they describe what someone held, not what they are doing today.

The supply you don't see

When a stock is being talked up, our earnings teammate checks filings for new shares that could be sold into any rally. That includes registrations that let existing holders resell shares, programs that let a company sell new shares gradually into the market, and shares issued to pay for an acquisition. Finding one is not proof an idea is wrong. It is logged as a risk to the setup, because extra supply can cap a move.

Someone else's chart is their opinion

When a trader posts a chart with levels on it, those levels are recorded as that person's opinion, kept separate from anything the Desk works out on its own. Only price data the Desk has saved itself counts as evidence. For the trader who posts calls without entries or stops, options-market levels from a licensed data provider are the main reference for support, resistance and targets. Even then, those levels shape a plan. They are never a reason to trade on their own.

The takeaway

Across all of these examples, the question is the same: is this new, is it independent, and does it show what someone is actually doing with their money? A dare with no stop, a headline repeated ten times, a holder's exit filing, and a gift of shares can all look like signals. Sorting them out before anyone acts is a big part of what the Desk does every day. Every post and article, including this one, still needs the owner's approval before it goes out.