Realistic Returns in Retail FX: An Unflinching Reality Check
A reality check for anyone considering a systematic FX system before committing weeks of work.
Verdict in one line: The base rate for retail FX is "lose money." The headline win-rate fantasies are dead on arrival. A defensible target is 10–25 % annual on a properly funded account, with 10–20 % drawdowns and a multi-year apprenticeship.
1. The Losing-Trader Statistics (Mandatory ESMA Disclosures)
European, UK, and Australian regulators force CFD/FX brokers to publish the percentage of retail accounts that lost money over the prior 12 months. Brokers must show this number in every ad and on every landing page. The figures are the most honest number in retail finance because they are legally compelled, recomputed quarterly, and audited.
Current disclosures (snapshots from broker sites and Good Money Guide's Dec 2025 aggregation):
| Broker | % retail accounts that lost money |
|---|---|
| eToro | ~46–61 % (varies by entity) |
| IG Group (UK) | 67–71 % |
| OANDA | 73.5 % |
| CMC Markets | 76 % |
| Pepperstone | 78.6 % |
| Plus500 | 76–82 % (improving from 81 % a year prior) |
| Industry-wide ESMA range | 68 %–89 % |
eToro's lower number is the outlier — driven by its CopyTrader business, where most "traders" passively mirror others rather than make discretionary trades. The honest spread for actively trading retail accounts is 70–85 % losers in any rolling 12-month window.
This is the short-window number. Over multi-year horizons it gets worse — Heimer & Simon's NBER work (NBER WP 22146; ~3,100 retail FX accounts, 110k+ transactions) found the median retail FX trader loses approximately 3 % per week of equity once frictions are accounted for. That implies near-total ruin in well under a year for anyone who keeps trading. Barber, Lee, Liu & Odean (2014), the canonical day-trader paper, found only about 15–20 % of Taiwan day traders were net profitable over short windows, and fewer than 1 % were reliably profitable net of costs over long windows. FX is harder than equities, not easier.
2. What Top Retail Traders Actually Earn
Forget Instagram. The only data worth respecting comes from:
- Myfxbook / FX Blue verified accounts (broker-API linked, can't be faked easily)
- Prop firm payout reports (audited, but with caveats)
- Hedge fund track records (the institutional ceiling)
Institutional ceiling (the very top of the curve):
- Renaissance Medallion Fund: ~66 % gross / ~39 % net annualised over 30+ years — and closed to outside money.
- Soros Quantum Fund: ~30 % annualised over ~30 years (multi-asset, not pure FX).
- Top CTAs and macro funds in publicly visible track records: 15–25 % annualised is elite; 30 %+ is exceptional and usually doesn't last.
Verified retail reality:
- The realistic Myfxbook benchmark for a good discretionary or systematic retail trader with a 2-year-plus track record is 15–25 % annual with 2–8 % max drawdown. Anything above 50 % annual with low drawdown on Myfxbook is almost always either martingale/grid (will blow up), curve-fit backtests that don't live-trade well, or short sample size.
- FundedNext's Feb 2026 payout report: median win rate on CFD payouts is 50 %, with 41 % of paid accounts winning under 50 % of trades (the R:R math works). Median futures win rate is 63 %.
Where the curve drops off: anything claiming >30 % annual for multiple years with single-digit drawdowns should be treated as fraudulent until proven otherwise via brokerage statements. The drop-off from "real top retail" (~25 % / yr) to fantasy land (~100 %+ / yr) is where 99 % of the influencer market lives.
3. Win Rate vs R-Multiple Math (Expectancy is Everything)
Expectancy per trade = (Win Rate × Average Win) − (Loss Rate × Average Loss).
A few realistic combinations:
| Win Rate | R:R | Expectancy per trade | Notes |
|---|---|---|---|
| 70 % | 1:0.5 (risk 1R to make 0.5R) | (0.70 × 0.5) − (0.30 × 1.0) = +0.05 R | Death by 1,000 cuts; one bad streak wipes you out |
| 50 % | 1:1 | 0.00 R | Pure coin flip after costs → net negative |
| 50 % | 1:1.5 | (0.50 × 1.5) − (0.50 × 1.0) = +0.25 R | Reasonable mean-reversion target |
| 40 % | 1:2 | (0.40 × 2) − (0.60 × 1) = +0.20 R | Classic trend-following profile |
| 35 % | 1:3 | (0.35 × 3) − (0.65 × 1) = +0.40 R | Top-decile CTAs live here |
| 25 % | 1:5 | (0.25 × 5) − (0.75 × 1) = +0.50 R | Long-vol / breakout (brutal psychologically) |
Two non-obvious truths:
- High win rates are a trap. Selling premium / mean-reversion / scalping systems with 70–80 % win rates almost always have left-tail blowups embedded. The smooth equity curve hides hurricane risk. Most retail accounts get killed by "high win-rate" strategies, not low ones.
- A higher R-multiple is a much more powerful lever than a higher win rate. Trend-followers averaging 38–48 % win rates with 2:1+ payoff have the most durable real track records. This is why almost every CTA in the BarclayHedge top-50 list lives in this zone.
Costs matter more than you think. A typical EUR/USD spread is 0.5–1 pip. If you risk 20 pips per trade and need a 2R win for breakeven net of slippage and spread, the spread alone is 5 % of your edge. Scalp at 5-pip risk and you've ceded ~20 % of your edge before you've even placed the trade.
4. Drawdown Reality
Drawdown is the variable that ends careers. A 50 % drawdown requires a 100 % recovery just to break even.
Realistic drawdown ranges by trader profile:
| Profile | Max DD typical | Time to recover |
|---|---|---|
| Top systematic CTA (e.g., Mulvaney, Dunn) | 25–45 % | 12–36 months |
| Strong retail systematic (verified Myfxbook) | 10–25 % | 6–18 months |
| Renaissance Medallion (gold standard) | ~10 % | <12 months |
| Most "successful" retail discretionary | 15–35 % | 6–24 months |
| Typical prop firm funded trader before blowup | 4–10 % then terminal | N/A |
The Sharpe-to-drawdown intuition: a strategy with Sharpe 1.0 typically experiences a max drawdown of 1.5–2.5x the annual return. A Sharpe-1.0 system targeting 20 %/yr will commonly drawdown 30–50 % over a multi-year sample. To get drawdowns into the single digits, you need either Sharpe > 2 (institutional/rare) or you have to dramatically de-lever, which means accepting low single-digit returns.
Implication for the AI system: if backtests show 80 % annual with 5 % max drawdown, the backtest is wrong. Either lookahead bias, survivorship bias, or curve-fit. Real edges don't look like that.
5. Survivorship and Selection Bias
This is the most important section for anyone consuming social-media trading content.
The coin-flip thought experiment. Start with 1,024 traders flipping coins. After 10 months, ~1 trader has a 10-month winning streak by pure chance. That trader sells a course. Run the experiment with 100,000 retail traders, and you produce roughly 100 traders with 10-month track records, plus a handful with multi-year "legendary" runs — driven by nothing but Bernoulli arithmetic.
Heimer & Simon (2016) studied exactly this dynamic on the social trading network myForexBook: winners broadcast their results loudly (about 50 % more than losers), and losers go silent. When a trader posted a win, peers in their network started trading ~20 % more frequently the following week. This is the engine behind retail FX: the visible population is the loudest, not the median. The median is in a quiet 3 %/week bleed.
Sample-size math. A trader with a 60-trade track record and a 55 % win rate has a 95 % confidence interval roughly from 42 % to 68 %. You literally cannot distinguish them from a coin flip at that sample size. Most "I made $10k last month" posts are operating in the noise band.
Selection bias in prop firm marketing. When FTMO advertises "$450M paid to traders," the relevant denominators (entry fees collected, total traders that paid for challenges, traders who blew up funded accounts) are not shown. The math suggests the firm collects more in challenge fees than it pays out in profit splits — which is exactly the B-book business model: the firm is the counterparty to most participants.
6. Prop Firms as a Path
Honest assessment, by the numbers:
- FTMO: ~10–12 % pass the two-step evaluation. Of those funded, ~50 % blow up or quit within 90 days, and a TradingView-cited industry analysis suggests ~98 % sever ties within 6 months. The average payout among the small surviving cohort is ~4 % of account size (e.g., ~$4k on a $100k funded account).
- FundedNext: $261M+ paid out, ~95,000 unique paid traders since launch, ~$8.8M/month payouts in 2025. Median paid trader's win rate is 50 % (CFD) / 63 % (futures). 50.1 % of payout recipients had been paid at least once before — the only durable cohort.
- MyForexFunds: collapsed under CFTC fraud charges Aug 2023 (~135,000 affected traders, $300M+ alleged fraud). The CFTC's case was later dismissed and sanctioned, and MFF is processing legacy payouts as of 2026, but the episode exposed the industry's simulated-account / B-book structural risk. Most "prop firms" never route retail flow to a real liquidity provider.
Sustainability question. The model is structurally a fee business, not a trading business. Most firms make their money on the $100–$1,000 challenge fees from the 90 %+ who fail, not on the 10 % who get funded. That's not inherently fraudulent — it is what it is — but it means prop firms scale by recruiting more failers, not by funding more winners. Treat them as a leverage product, not a career path.
When prop firms make sense: for a proven systematic trader who already has a real track record on a small live account but lacks capital. You're effectively renting leverage at the cost of an entry fee and a profit split. For an unproven trader, you're paying tuition.
7. The "How Much Capital?" Question
Position-size math at 2 % risk per trade and a realistic 15–25 % annual return target:
| Account size | Annual return @ 20 % | Monthly | What it covers |
|---|---|---|---|
| $1,000 | $200 | ~$17 | Hobby — pays for coffee |
| $10,000 | $2,000 | ~$167 | Pays a phone bill |
| $25,000 | $5,000 | ~$417 | Supplementary income |
| $100,000 | $20,000 | ~$1,667 | Meaningful side income |
| $250,000 | $50,000 | ~$4,167 | Mortgage-sized; approaches livable income |
| $500,000+ | $100,000+ | ~$8,333+ | Full-time income territory |
Honest brackets:
- <$10k: it's a hobby. Expected outcome = lose the account while learning. Treat it as tuition.
- $25k–$100k: edge can produce meaningful supplementary income, but you'll need to add capital to compound seriously. Be careful: this bracket is where most traders over-lever to "make it real" and blow up.
- $250k+: you can plausibly generate a livable income at realistic return rates without taking insane leverage. This is the bracket where "professional retail" begins.
- $500k+: full-time viable. Below this you are almost certainly better off with a salary plus index funds.
The leverage trap. A trader with $5,000 thinking "I'll just use 1:30 leverage to live on this" is mathematically doomed. Required monthly return becomes ~30–40 %, which forces enormous position sizing relative to stop, which guarantees ruin on the first adverse streak. The maths of compounding favours large slow capital, not small fast capital.
Bottom Line for an AI-Assisted System
- The win condition is not "80 % win rate, 5 % drawdown." It's "15–25 % annual, 15–25 % drawdown, Sharpe ~1.0, 2+ years of out-of-sample live track record." That is what "good" looks like.
- The realistic profitable population is ~10–15 % over short windows, ~1–5 % over multi-year windows. Assume you're not the exception until live PnL says otherwise.
- Spreads, slippage, and overnight financing eat retail edges fast. Bake them into every backtest at 1.5–2x their visible cost.
- You need either capital or a prop firm to scale, and you should not approach a prop firm until you have a real live track record. Do not pay challenge fees as a learning expense.
- Build the system to be killable. The right question is not "what's my max return?" but "what's my max drawdown before I shut it off?" Set that number in advance.
- Time horizon matters more than IQ. Plan on 18–36 months from "first backtest" to "consistent profitability on a real account, after costs, with a verified Myfxbook record." Anyone who tells you faster is selling something.
If after reading this the answer is still "yes, I want to build it," the work is worth doing. Just do it with eyes open to the base rates.
Primary Sources
- ESMA broker disclosures (IG, Pepperstone, OANDA, CMC, Plus500, eToro) — see Good Money Guide CFD risk warning aggregator (Dec 2025)
- Ben-David, Heimer & Simon — NBER Working Paper 22146, retail FX trader behavior
- Heimer & Simsek — NBER WP 24176, leverage restrictions and retail FX
- Barber, Lee, Liu & Odean — The Cross-Section of Speculator Skill: Evidence from Day Trading
- BIS — 2022 Triennial Central Bank Survey, global FX market structure ($7.5T/day)
- FundedNext — Feb 2026 Payout Report
- FTMO statistics — CoinLaw 2026 aggregator and QuantVPS prop firm stats 2026
- MyForexFunds / CFTC case — Finance Magnates coverage
- Renaissance Medallion / Soros Quantum returns — QuantifiedStrategies on Simons, TraderSlog top hedge funds
- Expectancy math reference — Trader's Second Brain expectancy formula