Almost no performance claim in crypto can be checked.
A screenshot of a winning trade, a win rate in a bio, a backtest in a deck. We show, with simple math and real data, why these records can look excellent while saying nothing about skill, and why a reader holding only the record cannot tell the difference.
In 2025 the FBI recorded $11.366 billion in reported losses from cryptocurrency-related complaints, including $7.228 billion from crypto investment fraud. Not all of it involves a fake track record. But an investment scam sells one thing: an expectation of returns.
Selection, editing, timing, and the wrong metric.
Selection
~10perfect 10-for-10 records expected from 10,000 people with no skill at all.
Editing
83%published win rate of a coin-flip trader who quietly deletes 4 of every 5 losing calls.
Timing
0ways to tell a prediction from hindsight without a timestamp the trader doesn't control.
The wrong metric
90%win rate that still loses money when the average loss dwarfs the average win (illustrative numbers).
The paper also covers why honest traders lose out too: if skill can't be proven, everyone is paid like the average, and the best stop selling (Akerlof's market for lemons). And why backtests mislead: try enough versions and one looks brilliant by chance.
The week's best traders, over their whole history.
median 7-day return of the top 100 wallets of the week
of those 100 still down over their whole history
were losing money before that month began
Source: Hyperliquid public leaderboard, snapshot of Oct 28, 2026, active wallets only (11,445 of 47,760). One exchange, one snapshot, aggregates only, no names or addresses. Free tool to check any Hyperliquid wallet: wallet check.
Commit before. Grade by rule. Keep everything.
Commit before
The exact call is fingerprinted and recorded on Ethereum before the window opens. The timestamp is the block, which the trader doesn't control.
Grade by rule
When the window closes, the call is graded by the rule in the locked text, on a price the trader doesn't choose.
Keep everything
Every call stays on the public record: correct, wrong and abandoned. Three per 30 days per address, and a score only after five graded calls.
The paper proves this removes editing and timing, and states plainly what it does not fix: one person running many wallets, trust in the grader, real trade execution, small samples. It is not an audit, and we don't call it one. Our own first calls were wrong, and they stay on our record.
Five questions to ask before you trust any record.
- Was each call written down before the move, somewhere the trader can't edit? If not, you can't tell a prediction from a story.
- Is this the whole record, or a selection? Ask how many calls were made in total.
- How many calls is it? Three out of three is noise. So is ten out of ten, from a crowd of thousands.
- What are the average win, the average loss and the worst trade? A win rate alone says almost nothing.
- Who graded it, and on what price? A result the trader computed on a price they chose is a claim, not a record.
If a trader declines all five, that is also an answer.
Quintero, J. (2026). Did They Call It, or Did They Just Post It? A Formal Account of Why Crypto Track Records Fail, and a Commitment-Based Remedy. Aevum Protocol.
FBI IC3, 2025 Internet Crime Report, pp. 52–53 · Chainalysis, 2026 Crypto Crime Report: Scams (Jan 13, 2026) · Bailey, Borwein, López de Prado & Zhu, Notices of the AMS 61(5), 2014 · Akerlof, QJE 84(3), 1970 · Aevum wallet study, Hyperliquid leaderboard snapshot, Oct 28, 2026.