How Do You Spot a Fake Track Record?
Assume the trades are real and ask how they were chosen. Almost every misleading track record is genuine trades filtered after the fact — one account of several, one period of many, the winners of a larger set — which is why the fix is completeness, not authenticity.
Why is authenticity the wrong question?
Because outright fabrication is rare and easy to catch, while selection is common, legal, and produces a record where every single trade is real. Asking "did these trades happen" gets a truthful yes. The question that separates records is "were these all of them, and were they named before they resolved".
What are the four ways selection happens?
1. Multiple accounts. Ten accounts, one wins, that one is the track record. You are shown the survivor and never told about the other nine, and from inside the winning account it is indistinguishable from skill.
2. Period selection. The record starts at the beginning of the good run. Two years of results starting from the exact month a strategy began working is a claim about the strategy and also a claim about the start date, and only one of them is being advertised.
3. Trade selection. Publishing the trades worth publishing. No individual entry is dishonest; the set is. This is the most common form by a wide margin.
4. Backfill. Posting the call after the move, or editing an old post to match what happened. Both are available on every chat platform, and neither leaves evidence a reader can find.
What checks are worth running?
Count the losses. A real discretionary strategy resolves against the trader 30% to 60% of the time. Twenty published trades should contain six to twelve losses. Zero to two is a filter, not an edge.
Match entries to exits. Of 20 entries, at least about 16 should have a published resolution. Trades that stop being mentioned resolved badly.
Ask what the record excludes. Other accounts, other strategies, the period before this one. The answer to "is this all of your trading" is more informative than any number on the page.
Check the drawdown against the return. A 40% annual return with a 4% worst drawdown is a claim about the risk-return relationship that would be world-class. A 20% drawdown needs a 25% gain to recover, a 50% drawdown needs 100% — so deep holes that vanish quickly are themselves a claim.
What are the tiers of proof?
| Evidence | What it establishes |
|---|---|
| Screenshots | Nothing — selected and editable |
| A self-maintained results page | Nothing — selected by the author |
| A broker export | The fills happened; not that they were called in advance |
| A third-party verified account | Real fills, real returns; selection across accounts remains |
| Every trade committed before the fact, on a record the trader cannot edit | The whole question |
What does the last row look like in practice?
Every trade, including the losers, timestamped before the fact, on a record the person being evaluated cannot edit. A broker import cannot produce it, because an import runs after the fact over data the trader controls; the record is assembled from outcomes that already exist.
None of this is a novel problem, and it was worked out decades ago. The GIPS standards exist precisely because self-reported performance could not be compared or trusted, and they turn on fair representation and full disclosure — every portfolio in a composite, not the ones that worked.[1] Outside the institutions that adopt them there is no equivalent standard, and investment fraud is the largest single reported loss category in the country: US consumers reported $5.7bn lost to investment scams in 2024.[2]
The mechanism underneath all four forms of selection is survivorship bias — the failures never enter the count — measured on fund performance in 1992: look only at the ones still running and the average looks like skill.[3] The published literature holds itself to a stricter test than any of this. Harvey, Liu and Zhu argued that so many predictors had been tried that a new one should clear a t-statistic above 3.0 to be believed,[4] a bar essentially no retail track record is presented against.
kappi inverts the order the selection depends on: the trade is committed from a Chrome extension before it resolves, sealed for a time-capsuled delay, then published on a Merkle-anchored log, with PnL, RME, correlation to SPX, mean R:R and trade count over 30, 100 and 200-day windows. Completeness stops being a promise and becomes a property of how the record is made. $15/month.
Sources
- CFA Institute, Global Investment Performance Standards (GIPS) for Firms, 2020 edition read 2026-08-16
- FTC Consumer Sentinel Network Data Book 2024 (published March 2025) — $12.5bn reported fraud losses, investment scams the largest category at $5.7bn read 2026-08-16
- Brown, Goetzmann, Ibbotson & Ross, 'Survivorship Bias in Performance Studies', Review of Financial Studies 5(4), 1992, 553–580 read 2026-08-16
- Harvey, Liu & Zhu, '… and the Cross-Section of Expected Returns', Review of Financial Studies 29(1), 2016, 5–68 — argues a newly claimed factor should clear a t-statistic above 3.0 read 2026-08-16
Frequently asked questions
How do you fake a trading track record without faking trades?
By selecting real ones: trade one of ten accounts and show the winner, start the record at the beginning of a good run, publish only the trades worth publishing, or post the call after the move.
Does a broker export prove a track record?
It proves the fills happened. It cannot prove they were named in advance, and it cannot show trades in accounts the export does not cover, which is where selection usually lives.
How many losses should a real track record contain?
A discretionary strategy typically resolves against the trader 30-60% of the time, so twenty published trades should contain roughly six to twelve losses. Zero to two indicates filtering.
What single property makes a track record checkable?
That every trade was committed before the fact was known, on a record the trader cannot revise afterwards. Everything else depends on the trader's restraint at the moment of publishing.