A Complaint in Los Angeles, a Wrong Label on the Pitch: Decoding the USD 2.1 Million Figure

On 11 May, a civil complaint was filed at Los Angeles County Superior Court. The plaintiffs are Sheila Greenberg, 85, and Norman Greenberg, 97. The defendants are Rosland Capital LLC, a precious metals dealer, and Erik von Detten, a former American television actor. The complaint pleads five causes of action: violation of elder protection law, fraud, breach of fiduciary duty, constructive fraud, and conversion. The plaintiffs have requested a jury. As of the time of writing, no defendant has responded publicly. That is everything the file states. It is also everything the file does not state. What made me stop was the label, not the content. In the data system this item passed through, it was classified as football content. No club. No player. No competition, contract, transfer, or governing body. Twenty-six information points, and not one of them touches football. Yet the label stayed, and from that label a football data product will be generated. In my trade we call that a misdiagnosis. And a misdiagnosis, in any field, is more dangerous than the disease. Erik von Detten, 44, is known for The Princess Diaries, Brink!, So Weird and a voice role in Toy Story. He left Hollywood at 25 to move into finance. In a 2026 interview he said the reason was a desire for steadier employment to support his family. He is recorded as a broker at Rosland Capital LLC. The two plaintiffs are described as holding roughly USD 2 million across retirement accounts and as self-declaring limited knowledge of gold products. The transaction structure set out in the complaint has four steps, and all four matter. First, the clients arrived through a television advertisement. Second, they were assigned to an individual broker. Third, they bought roughly USD 1 million of gold, but the metal was never delivered to them — it sat at Delaware Depository, generating storage fees and recurring account maintenance charges. Fourth, a further purchase of roughly USD 1 million was executed on their behalf, and the markups were allegedly concealed until later. Total paid: about USD 2.1 million. The plaintiffs' assessed value of what they received: about USD 1.3 million. Damages claimed: USD 900,000, plus punitive damages. I have spent most of my career reading documents like this in exactly one way: strip out the storytelling, isolate the arithmetic, then ask whether the arithmetic stands on its own. In 2026, when Chelsea paid GBP 58 million for Alvaro Morata, I did precisely that with his injury record and found back-injury frequency rising 26% per season. In 2026, at the World Cup, I did it with Harry Kane's ankle and GPS data from sensor-equipped boots: shot power down 18%, sprint times markedly slower. The method does not change when the subject changes. Every transfer is a surgical case — outsiders see the scar, professionals see the blood supply. In this file, the blood supply is in three calculations. The first is the markup. Two point one million divided by one point three million gives a factor of roughly 1.615. In other words, the clients paid about 61.5% above assessed value, equivalent to roughly USD 800,000. In retail bullion, the spread between buy and sell is normally published and sits inside a narrow band. Sixty-one and a half percent sits far outside it. This is the spine of the entire complaint, and it is built entirely from arithmetic, not emotion. The second is the USD 100,000 gap. Assessed value is USD 800,000 below the purchase price, but damages claimed are USD 900,000. The two figures diverge by about USD 100,000. There is an innocent explanation that requires no allegation of fraud: that difference is accumulated storage fees, account maintenance charges and transaction costs over the holding period. But because it is plausible, it becomes a vulnerability. A competent defence lawyer opens cross-examination with exactly this question — if you can compute the markup to the dollar, why does your loss figure not reconcile? When a plaintiff's damages model does not close, the valuation standing behind it is dragged into doubt with it. The third is the least discussed. Two point one million is consideration. Nine hundred thousand is damages claimed. The ratio is roughly 43%. The largest number in the story is more than double the figure actually in dispute before a court. Over 32 years in this industry I have watched this mechanism repeat hundreds of times: people read the headline price and assume it is the loss. A headline price is what a seller wants you to see. A loss figure is what a court will test. They are never the same, and the distance between them is where my profession exists. One structural detail is worth more than all three calculations, and it is usually skipped because it is not shocking. The gold was never delivered. It sat with a third-party depository, and every month that passed generated a fee. This is the classic architecture of high-margin retail: the product is sold once, but the fee stream runs indefinitely. The USD 800,000 markup is a loss visible on day one. Storage and maintenance charges are an invisible loss, leaking month by month, and nobody totals them until somebody totals them. Football has an equivalent structure, and it is skipped in exactly the same way. A club pays EUR 40 million for a player its own valuation models price at EUR 25 million. The EUR 15 million gap is visible. The invisible part is wages, agent fees, instalments and amortisation spread across the contract. When the player fails, everyone argues about the 40 million. Nobody argues about total cost, because total cost was never written on a single line. Legally, the five causes of action do not carry equal weight. Elder financial abuse is the highest-severity count in California civil practice, because it typically opens the door to enhanced damages and fee-shifting. Fraud and constructive fraud require proof of misrepresentation and reliance. Conversion is a property-based claim, and it is notable because it suggests the plaintiffs may argue the metal never truly came under their control — it sat in a depository, in someone else's name. The decisive count is the third: breach of fiduciary duty. The whole case turns on one question about the nature of the relationship. If it is seller-to-buyer, fiduciary duties barely exist and this count collapses. If it is adviser-to-client, fiduciary duties attach, and with them a full disclosure obligation. The complaint is drafted to lean toward the second: it describes someone purchasing on the clients' behalf, not selling to them. That is a deliberate choice of words, and it is the real battleground. One point that coverage tends to compress: no defendant has responded, there has been no discovery, and no contemporaneous spot-price evidence from the defence exists on the record. The USD 1.3 million valuation is a one-sided assertion. It will be contested with the spot price of that exact day. Until that data appears, any conclusion about the true markup is a probabilistic hypothesis, not an established fact. There is a pattern here, and it is why I am writing this on a sports desk. The transfer market contains two kinds of premium. The first is a competition premium: two clubs want the same player, the price is bid up, and the excess is the cost of winning a public race. The second is an information-asymmetry premium: one side understands true value, the other has no tool to verify it, and the excess is the cost of not knowing. The first leaves traces — rivals, bids, records. The second leaves nothing but a signed contract. In this file there is no rival. No race. One buyer, one seller, one markup. That is the second kind, and it is the hardest to detect because it makes no noise. I have tracked many deals in the Premier League and Championship mid-market where the premium came not from an auction but from a buying club lacking the data function to push back on an agent. The price then reflects the payer's information quality, not the player's value. And when the player fails, the player is blamed. The process that produced the price rarely is. This is where I go against the consensus. The easiest reading is: former child star sued over defrauding an elderly couple. That reading has everything a short news cycle needs — 2000s film nostalgia, ages 85 and 97, and a number that sounds enormous. Remove those three elements and what remains is a consumer-finance complaint in the mid-six figures, with no response filed and no finding made. The heat is generated by film fame, not by evidence volume. The heat-to-fundamentals ratio here is severely divergent, and in my trade that divergence is a signal to read more slowly, not faster. The second point matters more: the plaintiffs have a USD 100,000 arithmetic gap inside their own filing. An outsider reads that as a technicality. A professional reads it as the opening of the entire defence strategy. In every case I have followed, the weaker party is not the one with fewer arguments. It is the one whose numbers do not reconcile. The third point I have to state plainly: it cannot yet be determined whether any regulator will open a parallel inquiry. The record does not say whether the broker held registrations, whether any agency is reviewing the matter, or whether this sales model has attracted prior complaints. In elder financial cases the single largest escalation variable is always a parallel regulatory action, which dwarfs a private claim many times over. At this stage, that variable sits outside the frame. An injury decoder is only worth something when he is willing to say so. And then there is the label. Taken alone, the classification failure is minor. A non-sports file gets tagged as sports, is removed from the corpus, and everything returns to normal. But it does not happen once. If a classifier assigns labels by entity name rather than by topic, then any famous name appearing in an article can drag the entire article into an unrelated corpus. A former Toy Story voice actor lands in a football database today. Tomorrow it is a singer, a politician, an athlete from another sport. Each time, a model learns a relationship that does not exist. I have spent a career arguing that players' bodies do not lie — only the people managing them lie on their behalf. The same principle governs data. A dataset does not lie. It only repeats what was assigned to it. When a football data product contains a lawsuit about gold, that product is lying to its users, and it is lying in our own voice. Rebuilding is not returning — it is re-reading the whole map and redrawing what has gone out of date. In 2026, when football stopped for 99 days, this industry learned that systems believed to be solid could not survive a single stress test. Our data systems are in that exact state now: large enough to look credible, and never tested where they are weakest. Believe me on this: true value is the one thing in football that cannot be bought through negotiation — everything else is smoke. And the thickest smoke is always generated by the first number anyone sees. The question I leave is not for a Los Angeles court but for people who do my job: if a file with no connection to football can pass through our systems unchallenged, then in what percentage of the data we use to price players, assess injuries and forecast seasons is there a wrong label sitting quietly — and who will be the first to add up the fees nobody has added up?