Search for day trading statistics and you will mostly find the same recycled numbers, copied from blog to blog without a primary source. This page takes the opposite approach: every figure here is either published by a regulator, attributable to named academic researchers, or drawn from our own original dataset — a 12-month backtest of eleven popular strategies that we publish in full, equity curves and all.
Three kinds of evidence, in order of how directly you can check them yourself: regulatory loss disclosures, peer-reviewed research, and our 2026 strategy dataset of more than 42,000 simulated trades.
The short list: statistics with sources
If you only take away five numbers, take these — each one verifiable:
- 74–89% of retail CFD accounts lose money, according to the risk disclosures EU brokers are legally required to publish.
- Fewer than 1% of day traders are persistently profitable after fees, per Barber, Lee, Liu and Odean’s research on the complete trading records of the Taiwan Stock Exchange.
- 0 of 24: in our own 12-month backtest (twelve strategy readings × BTC and ETH, including two relaxed Silver Bullet variants), not one unfiltered mechanical strategy finished profitable after fees.
- The least-bad strategy we tested by profit factor — the NY opening range breakout — still lost, with a profit factor of 0.65 on BTC and 0.80 on ETH.
- Our 1-minute scalping baseline paid $10,850 in fees on a $10,000 account in one year — trading costs alone exceeded the entire starting balance.
The regulatory numbers: 74–89% of CFD accounts lose
Under EU product-intervention rules introduced in 2018, every broker offering CFDs to retail clients must display the percentage of its retail accounts that lose money. The disclosed figures cluster between 74% and 89%. This is the rare trading statistic with legal force behind it: a broker that misstates it answers to a regulator, not a community note.
Read it carefully, though: it covers leveraged CFD accounts (not all forms of trading), counts accounts over a disclosure period (not lifetimes), and says nothing about why the accounts lose. But as a floor on how hard leveraged retail trading is, it is the best-evidenced number in the industry.
The academic numbers: Barber, Lee, Liu and Odean
Taiwan’s stock exchange recorded every transaction with trader identity attached, which allowed Brad Barber, Yi-Tsung Lee, Yu-Jane Liu and Terrance Odean to study the entire population of day traders over many years — no surveys, no self-selection. Their conclusion: the vast majority of day traders lose money after costs, and fewer than 1% earn reliably positive abnormal returns net of fees. The profitable few were persistent from year to year, so skill exists — it is just vanishingly rare.
Barber and Odean’s US household-brokerage study, “Trading Is Hazardous to Your Wealth”, adds the dose-response curve: the more actively retail investors traded, the worse their net returns. Trading frequency itself was the strongest predictor of underperformance in their sample.
Our original 2026 dataset: the methodology
Aggregate statistics tell you that traders lose; they don’t tell you whether the specific strategies traders are taught actually work. So we tested them. We took eleven popular strategy recipes — ICT Silver Bullet, fair value gap entries, liquidity sweep reversals, opening range breakouts, EMA and MACD crosses, RSI mean reversion, 1-minute momentum scalping and more — and ran each one mechanically, exactly as commonly described, with no discretionary filter.
The setup: twelve months of Binance 1-minute data (June 1, 2025 – June 1, 2026) on BTC and ETH, a $10,000 starting balance, 1% risk per trade, position size capped at 10× notional leverage, and 0.05% commission per side. One deliberately conservative convention matters: when a single bar touches both the stop and the target, the engine counts the stop. Every run — stats, monthly breakdowns, equity curves and assumptions — is published at secuora.xyz/strategy.
The results, strategy by strategy
Across 24 strategy-symbol runs and more than 42,000 simulated trades, not a single unfiltered baseline finished the year profitable after fees. The BTC figures, with ETH in parentheses for selected runs:
- Opening Range Breakout (NY, 30m) — the least bad: 417 trades, 39.3% win rate, 0.65 profit factor, −44.8% net (ETH: 42.2% win rate, 0.80 profit factor, −26.9%).
- EMA 50/200 golden cross: 55 trades, 27.3% win rate, 0.54 profit factor, −19.8% net (ETH: −13.5%). The small loss is mostly because it barely trades.
- MACD zero-line cross: 293 trades, 29.7% win rate, 0.62 profit factor, −61.3% net (ETH: 0.79 profit factor, −35.2%).
- London session breakout: 465 trades, 34.2% win rate, 0.49 profit factor, −74.2% net (ETH: −40.2%).
- NY opening drive: 307 trades, 30.0% win rate, 0.33 profit factor, −86.2% net.
- RSI(14) mean reversion: 1,338 trades, 38.4% win rate, 0.54 profit factor, −98.9% net (ETH: 0.75 profit factor, −87.3%).
- FVG entry after 10:00 NY: 1,154 trades, 27.3% win rate, 0.34 profit factor, −99.3% net.
- Turtle Soup (intraday): 777 trades, 21.5% win rate, 0.22 profit factor, −99.9% net.
- Liquidity sweep reversal: 3,349 trades, 21.1% win rate, 0.22 profit factor, −100.0% net (ETH: 3,282 trades, also −100.0%).
- Momentum scalping (1m): 15,653 trades, 15.2% win rate, 0.05 profit factor, −100.0% net — with $10,850 paid in fees.
- ICT Silver Bullet (strict): exactly 1 trade all year, a loss (−1.2% net). The relaxed variant took 46 trades, roughly broke even before costs, and still lost 33.5% — almost all of it fees ($3,323).
What the dataset actually teaches
The headline is not “day trading is impossible” — it is that raw, unfiltered recipes carry no built-in edge, and costs decide what happens next. Fee drag is the clearest mechanism in the data: the strategies that traded most lost most, and the 1-minute scalper handed its entire account to commissions. The relaxed Silver Bullet is the purest case — on BTC it was essentially a break-even strategy before costs, and fees alone turned it into a −33.5% year.
Two more lessons travel well beyond crypto. Sample size: the strict Silver Bullet produced one trade in a year, which is why we published the relaxed variant alongside it — an untestable strategy is an unfalsifiable one. And conventions: on the 1-minute timeframe the worst-case same-bar rule (the stop is counted whenever one bar touches both stop and target) materially shapes the result, so any 1m backtest that doesn’t disclose its fill convention should be read with suspicion.
The “95% fail in 90 days” myth
The most quoted day-trading statistic of all — that 90–95% of traders lose their money within 90 days — has no traceable primary source. It appears in books, videos and forum posts, always attributed to “studies” that are never named. We looked: the chain of citations reliably ends at another blog. It is folklore with a percentage sign.
The irony is that the truth is bad enough without embellishment: regulators document 74–89% of CFD accounts losing, and the best academic dataset puts persistent day-trading skill below 1%. A claim that needs a fabricated source doesn’t make the case stronger — it teaches readers to dismiss the real evidence along with the fake.
How to read any trading statistic
A short checklist that filters out most of the nonsense:
- Demand a primary source — a named regulator, named researchers, or published raw data. “Studies show” is not a source.
- Check the denominator: accounts, people, trades and challenge attempts produce very different percentages.
- Ask whether costs are included — a strategy can be positive gross and ruinous net, as several of our runs demonstrate.
- Watch for survivorship: statistics drawn from visible, active or self-reporting traders skew rosy by construction.
- Prefer data you can re-run. Our dataset is reproducible by design — the same deterministic engine that produced it powers Secuora’s AI backtester at /backtest/ai, so you can test the same recipes, or your own variant, yourself.
Frequently asked questions
What percentage of day traders are profitable?
The best evidence — Barber, Lee, Liu and Odean’s research on the Taiwan Stock Exchange’s complete records — found that most day traders lose after costs and fewer than 1% are persistently profitable net of fees. Separately, EU CFD brokers disclose that 74–89% of retail accounts lose money. No credible universal figure exists beyond those.
Is the “95% of traders fail within 90 days” statistic real?
No traceable primary source exists for it — citations always lead to another blog, never to a named study. The documented reality (74–89% of CFD accounts losing; under 1% persistent day-trading skill in the Taiwan data) is sobering enough without invented numbers.
Does your dataset prove that day trading can’t work?
No. It shows that eleven popular recipes, traded mechanically with no discretion over this 12-month period on BTC and ETH, all lost after fees. Discretionary filters, regime awareness and cost control are precisely where any real edge has to come from — and the academic record confirms a small persistent minority does achieve it.
Where can I see the full backtest data?
The complete research — per-strategy stats, monthly breakdowns, equity curves and every assumption — is published at secuora.xyz/strategy. The runs were produced by the same deterministic engine that powers the in-app AI backtester.
Why test strategies you expected to lose?
Because nobody else publishes the baselines. Knowing that a raw recipe loses after fees — and exactly how it loses, whether to costs, low win rate or both — tells you what work a strategy still needs before it deserves real money. That is more useful than another highlight reel.