Citron Research went quiet in March. We graded its last fifteen calls

For twenty five years, a Citron Research post could move a stock by double digits in a morning. On June 1, a federal jury in Los Angeles convicted its founder, Andrew Left, on 13 counts of securities fraud. The account's last post in our feed landed on March 13, eleven weeks before the verdict, and it has been silent since.

The verdict has been covered everywhere. The record has not. We have tracked @CitronResearch since March 2023, and 134 of its posts became scored signals in our system. Today we are publishing the graded record of the account's final stretch.

The case, in one paragraph

Prosecutors charged Left in July 2024 with a scheme that worked like this: publish a market-moving recommendation, watch the stock jump or drop (12% on average, by the government's math), then quickly trade against the stated position. In one charged example, Citron told followers it would stay long a stock until $65 while it was already selling at $28. The indictment quoted him calling it "taking candy from a baby". After a 15 day trial in which Left testified in his own defense, the jury convicted him on a scheme count and 12 fraud counts, acquitted him on four others, and the court set sentencing for August 31. The charged conduct ran from 2018 to 2023, and prosecutors put the profits at $21 million. Left denies wrongdoing and has signaled an appeal.

One date matters for everything below: the charged conduct ends in 2023, around where our tracking begins. What follows is the part of the story the indictment does not cover, the account's final fifteen positions.

The scorecard

We grade every public call the same way, whether it comes from a professional short seller or an anonymous Reddit account: what did the post say, and what did the price do over the next 7, 30, 90, and 180 days. A move under 2% counts as flat. Repeated posts about the same name collapse into one position, so pounding the table does not inflate a record.

A note on scope. Our price history for Citron's earliest tracked calls (2023 and most of 2024) is too thin to grade fairly, so we are publishing only the stretch where our coverage is complete: January 2025 through the final post. That stretch contains fifteen distinct positions, graded here at 90 days:

Twelve of fifteen right at ninety days. The same calls, graded at each window:

At 30 days he was wrong more often than right. At 90 days, twelve of fifteen landed. His own Teladoc post described the account's style as "early but not wrong", and on this stretch of data that reads less like bravado and more like a measurement.

Three calls

Duolingo. Citron shorted it at $382 in April 2025 and the stock ran 37% against the call within a month. The account said so in public: "We were short $DUOL before last earnings and got it wrong. But the thesis hasn't weakened". By November, Duolingo traded at $178, down 65% from the May re-short at $514. The position lost for a month and won for the rest of the year.

Rocket Companies. In this stretch the famous short seller graded better as a bull: his longs went 6 of 7 at ninety days, his shorts 6 of 8. Rocket was the standout ("Elizabeth Warren and Bernie Sanders have NEVER made Citron want to buy a stock... until now"), bought at $12.61 in May 2025, up 36% at the 90 day grade and 65% within six months of his last repost of the thesis.

Sandisk. This one went the other way. On February 24 Citron announced a short at $638.52 under the line "They don't ring a bell at the top". The first week went his way, down 11%. Then the memory rally arrived. Ninety days after the post, Sandisk closed at $1,589.55, up 149% against the short, the worst graded call on the account in our covered stretch. Early was this account's usual failure mode. On Sandisk, early was just wrong.

The account's final act was a disclosed reversal: after two decades of publishing, it flipped publicly from short to long on Credit Acceptance with a $714 fair value, reposted the thesis on March 13, and went dark. That last call is up 23% at its 90 day grade.

So what was the crime?

The jury did not convict Andrew Left because his research was wrong. Per the charges, it convicted him because what he told the market about his own positioning was not what his book was doing. The public thesis said one thing; the private trades, prosecutors showed, said another. The words can be right about the stock and still be bait.

A graded track record measures one thing: whether the public call was right about the market. It cannot see a private book. That second kind of trust is what disclosure rules and courtrooms are for. The first kind can be measured by anyone who keeps records, every day, for every account. The Citron story is what happens when the two point in opposite directions: an analyst good enough not to need the scheme, convicted for running it anyway.

Caveats

Fifteen positions is a small sample. The graded stretch begins in January 2025, after the conduct the jury ruled on, so nothing above supports or contradicts the verdict; it is a different question answered with different data. Our grading is mechanical (direction versus price over fixed windows) and does not capture position sizing, options, or exits. An appeal is pending. And if the account posts again, its record picks up exactly where it left off.

Keep the receipts

For years, the market's reflex to a Citron post was to trade first and check later; by the government's math, the average move was 12%. Almost nobody wrote the calls down and graded them afterward. The receipts were public the whole time.

Quantral does this for every account we track, with the same thresholds and the same windows whether the caller is famous or anonymous. The records live on the leaderboard, the method is documented in how a track record is graded, and the reason it decides everything is in what is a credibility score.

Trust the record you can verify. For everything else, there are juries.

---

Quantral surfaces signals and context from public sources to support your own research. Nothing here is financial advice or a recommendation to buy or sell. Facts about United States v. Left are drawn from public reporting and court records; the graded results describe only posts observed in the accounts we track. Past signals are not indicative of future results.