Every few weeks someone sends me a screenshot. A brokerage app, a green number, a percentage with too many digits in front of the decimal. Sometimes it’s a stranger online announcing that an AI made it rich. Sometimes it’s a friend asking whether he should believe the stranger. My answer is always the same question: what happened the day before this was taken, and the day after?
A screenshot is a single frame, chosen by the one person with every reason to choose it. It has no start date, no denominator, and no record of the accounts that never got posted. It is a highlight reel, and I’ve written before about what highlight reels do to judgment. So it’s fair to ask what makes Obsidian’s record any different. Obsidian has traded a $100,000 paper account since February 17, 2026. That is a simulation. I’m saying so up front, because the people who doubt it are half right, and the half they’re right about deserves to be taken seriously.
What Paper Can’t Tell You
Start with the obvious. A live-market paper account trades on real market data at real prices, but the fills are simulated by the broker. Three gaps fall out of that, and all three are disclosed on the performance page rather than buried. The simulator fills at the quoted price regardless of size, so market impact — the cost of your own order pushing the price against you — isn’t in the numbers. The account is charged no margin interest, which over this period is worth less than one percentage point. And nobody’s money is on the line. No client capital has been deployed to date.
That last gap is the one the title is about. Paper money doesn’t sweat. It doesn’t check the account at two in the morning, and it doesn’t remember that the rent is due. The strategy wins a minority of its trades by design — 39.8% of 398 closed trades so far — and makes its money because the average winner, about $1,209, is roughly twice the average loser, about $598. On paper, a long run of losers is a few lines in a log. In a real account, it’s the exact moment most people reach for the off switch, usually just before the winners that pay for the losers arrive. A paper record can’t tell you whether the person holding the account will let the system finish its sentence.
“A simulation can tell you whether the machine follows its rules. It can’t tell you whether the human will let it.”
What It Can
This is where the skeptics overshoot. “It’s only paper” gets used as if it meant “it’s only a story,” and those are very different claims. A live paper account is not a reconstruction of the past. Every decision in it was made in real time, on data the system had never seen, before anyone knew how the trade would end. You can’t tune a rulebook to prices that haven’t printed yet. That single property is what separates evidence from a story told after the fact, and it is the property a screenshot never has.
The account also has a fixed start date and runs continuously, which means the bad days stay in. You can’t crop a ledger. As of the October 12 close, the account stood at $163,351.84, a 63.35% return since inception, against 13.95% for SPY, 24.07% for QQQ and 5.65% for IWM over the identical window. Those figures are only worth anything because the drawdowns, the defects and the losing streaks sit in the same record, timestamped right next to them.
The Day the Paper Account Broke
The most useful thing the paper account has produced isn’t a return. It’s a bug. On August 19, orders generated while the market was closed were accepted and queued by the broker instead of being rejected. A queued order creates no position, so the reconciliation logic read that absence as a completed trade, cleared its own record and allowed the same order to be placed again. At the next open the orders filled together as one oversized position, which was closed the same day at a realized loss of $17,678. Earlier, in June, a separate defect in protective stop-order handling had pushed the drawdown against daily account equity to 17.1%.
Neither was a market call. Both were plumbing. And plumbing is exactly what a screenshot hides and a paper account exposes, because the fills may be simulated but the orders, the timing and the software sending them are entirely real. Both defects were corrected. The August fix added checks that each fail to the safe side: orders are now refused when the exchange is closed, refused again if an order is already working in the same security, and refused again if the resulting position would exceed a fixed share of account equity. Better to find that on a paper account than in someone’s retirement savings. Better still to publish it, because a record that only shows the good days is just a screenshot with extra steps.
What Eight Months Can’t Prove
Even an honest record has limits, and the honest move is to name them before anyone else does. The record covers less than eight months, and it spans a generally rising market. The strategy has changed along the way, one tested change at a time, so the history covers more than one configuration and the release running today is newer than much of it. The book is long-only and modestly levered, so a broad, sudden decline would hurt, and an overnight gap can jump straight past any stop. None of that is fine print. It is the list of things that would have to hold for these numbers to keep meaning something, and anyone reading the record should read that list first.
What the record does prove is narrower and much harder to fake. A fixed rulebook, applied without discretion, made every decision in real time for the better part of a year, and every one of those decisions is on the ledger — including the ones that cost money for reasons that had nothing to do with the market. That is less than “this will make you rich.” It is a great deal more than a green number in a cropped image.
Why It Was Built This Way
I traded my own capital by hand for years, and the failures that end discretionary traders are almost never analytical. They are operator failures: forgetting a position, trading bigger after a loss, changing size because a bill is coming. I built Obsidian Quant so the results would stop depending on the operator, which is also why it started on paper. It is a systematic long-only U.S. equity system that enters every position with a predefined maximum loss and no ceiling on the gain, runs inside the licensee’s own brokerage account under their own keys, and is licensed as software for a flat monthly fee. The full record, disclosures included, sits on the public performance page, updated as the account trades.
Paper money doesn’t sweat. That is its weakness as evidence, and it is why the test that matters most is still ahead: real capital, held by people who will feel every losing streak. Until then, I would rather show you an honest simulation with its scars than a perfect screenshot with none.