Reality Check / Trading courses / Outcomes
TRADING COURSE REALITY CHECK · READ 9 Oct 2026
How retail traders actually do
Before you judge a course, know the base rate. These are the official and peer-reviewed numbers on retail trading outcomes, with the sample and the year next to each one.
Read this first
What the numbers do and do not show
- They measure retail accounts and day traders, not course graduates. We found no study that compares buyers of paid trading courses with non-buyers. That absence is itself a finding: no seller can point to an independent study that proves their course works.
- CFD and forex statistics are not crypto spot statistics. They are the best official loss data that exists. Crypto figures from the BIS are simulations, and the page says so.
- Samples differ by country and period. Do not average them. Read each row with its year.
Evidence
Regulators: share of retail accounts losing money
SourceFigureYearLink
ESMA: national authorities' analysis found most retail CFD accounts lose money
74-89% of retail investor accounts typically lose money; average loss per investor EUR 1,600-29,000
2018 (27 Mar)
ESMA press release: same 74-89% range; the measures require a standardised risk warning stating the % of losing retail accounts of each provider
74-89%; leverage caps 30:1 to 2:1; 50% margin close-out; negative balance protection; ban on binary options
2018 (27 Mar)
AMF France: study of retail Forex/CFD clients of authorised providers
more than 89% of clients lost money; average loss about EUR 10,900; median loss EUR 1,843; 13,224 losing clients lost about EUR 175 million; remaining 1,575 clients gained EUR 13.8 million (about 11%, computed)
2014 (Oct); data 2009-2012
AMF restates the study: 9 of 10 clients lost, no learning effect over time, EUR 161 million net loss
about 90% losing; EUR 10,900 average; no learning effect observed; net EUR 161 million (175 minus 13.8)
2015 (8 Apr)
AMF 2020 alert on trading-training pyramid schemes also restates European CFD loss range
share of retail clients losing money on CFDs at European providers roughly 70% to over 80%
2020 (22 Jun)
ASIC: most retail CFD clients lose money
72% of CFD clients and 63% of margin FX clients lost money; 80% of binary-options clients lost money; total negative balances on CFD accounts 2018 = A$33m
2019 (22 Aug); client data from ASIC 2017 review, issuer data 2017-2019
ASIC product-intervention notice: period of the 72/63% figures and COVID-era data
63% (margin FX) and 72% (other CFDs) lost money over a 12-month period in the 2017 review; 16 Mar-19 Apr 2020 weekly share of losing retail accounts 56.9%-63.2%; net retail losses A$774,238,833 in five weeks; calendar 2018: about 5.7% of accounts (over 41,000) went into negative equity, over A$33m
2020 (22 Oct)
CONSOB 2024 survey context: 18% of Italian financial decision-makers hold crypto (8% in 2022); 36% use social media for investment choices (58% of 18-34s); 26% overconfident on basic financial knowledge; 41% interested in online trading
18% crypto; 36% social media; 26% overconfident; 41% interested in online trading
2024 (survey Q1 2024; press release 30 Jul 2024)
BIS Working Paper 1049 (retail crypto): simulation estimate of the share of crypto-app users who lost money on Bitcoin
73-81% of users had likely lost money; median simulated loss USD 431 = 48% of USD 900 invested. Simulations, not observed returns: the database holds no actual performance data
2022 (Nov)
BIS Bulletin 69 (retail crypto)
A majority of crypto-app users in nearly all economies made losses on their bitcoin holdings (no percentage on the page; the page gives the period as Aug-Dec 2022 in the abstract and Aug 2015-Dec 2022 in the takeaways)
2023 (20 Feb)
Evidence
Peer-reviewed day-trading and retail trader studies
SourceFigureYearLink
Chague, De-Losso, Giovannetti: day trading for a living in Brazilian futures. Abstract figures
97% of day traders who persisted 300+ days lost money; only 0.4% earned more than a bank teller (about USD 54/day); best individual about USD 310/day (SD USD 2,560); no evidence of learning
2019 (SSRN); FGV EESP Texto para Discussao 525 (2020)
Barber, Lee, Liu, Odean: skill among Taiwan day traders
Less than 1% of the day trader population is able to predictably and reliably earn positive abnormal returns net of fees. Top-ranked 500 traders: +37.9 bps/day net; bottom-ranked: -28.9 bps/day net
2014 (J. Financial Markets 18(1) 1-24)
Barber, Lee, Liu, Odean, Zhang: do day traders learn? WORKING PAPER (Oct 2017), not the final published version
more than 75% of day traders quit within two years; survival 44% at 1 year, 24% at 2, 15% at 3 (traders with 10+ day-trading days); aggregate day trading loses in every year but 1992, average net about -23.9 bps/day; unprofitable traders are over two-thirds of day traders and 72% of volume; profitable experienced trade
2017 (working paper)
Barber, Lee, Liu, Odean: aggregate losses of individual investors
Individuals' aggregate portfolios underperform by 3.8 percentage points per year; losses equal 2.2% of Taiwan GDP and 2.8% of personal income; institutions gain about 1.5 pp per year
2009 (Review of Financial Studies 22(2) 609-632)
Barber and Odean: 'Trading is hazardous to your wealth'
Households in the top turnover quintile earned 11.4% a year net of costs, versus 16.4% for the average household and 17.9% for the market (Feb 1991-Jan 1997)
2000 (Journal of Finance 55(2) 773-806)
Odean: 'Do investors trade too much?'
Stocks bought underperformed stocks sold by about 1.4 pp (4 months), 3.3 pp (1 year) and 3.3 pp (2 years); average round-trip cost about 5.9%; conclusion: these investors trade excessively
1999 (American Economic Review 89(5) 1279-1298)
Kogan, Makarov, Niessner, Schoar: retail crypto trading behaviour. It measures behaviour (momentum in crypto vs contrarian in stocks), NOT returns or losses, so it cannot support a loss claim
median holding period 12 days (crypto) vs 7 days (stocks); average account balance under USD 1,000
2023 (NBER WP 31317)
Evidence
Does paid trading education improve results?
SourceFigureYearLink
Meta-analysis of general financial education (not trading courses)
Interventions explain only 0.1% of the variance in financial behaviours; even intensive programmes have negligible effects on behaviour 20 months or more after the intervention
2014 (Management Science)
Chague et al.: no evidence of learning among Brazilian day traders; authors frame the paper against course providers' claims (see topic 2)
no evidence of learning; 97% lost
2019/2020
Escobar and Pedraza: random classroom assignment in a financial education programme; students whose classmates had more trading experience started trading more; groups with high peer returns earned lower trading profits (social transmission of biased success s
qualitative: lower trading profits in high-peer-return groups
2019 (World Bank WP 8767); published in Journal of Financial Economics 2023 (per ScienceDirect listing)
FTC v. Warrior Trading, Inc. (day-trading course seller). Allegation in the FTC complaint, resolved by stipulated order: most customer accounts lost money and many lost thousands on top of fees paid; earnings claims misleading
USD 3 million refund order; USD 2.9 million refunded to 20,402 people (10 Jan 2023); complaint 19 Apr 2022; D. Mass., Civil Action 3:22-cv-30048; final order 26 May 2022
2022-2023
FTC press release on the same case (April 2022)
most customer accounts lost money; misleading earnings claims; USD 3 million
2022 (19 Apr)
AMF warning on MLM/pyramid sellers of Forex and crypto trading training
about 100 reports in three months; training packs cost a few hundred euros plus subscriptions; AMF: trading skills cannot be learned in a few hours
2020 (22 Jun)
FINRA Foundation / Stanford IFDM: investment-literacy study. It does not cover trading courses
18% advanced, 66% basic, 16% low literacy; 43% of low-literacy investors follow social media personalities vs 13% of advanced; advanced-literacy investors 15 points less likely to carry costly card debt
2026
Evidence
Survivorship, selection, overfitting
SourceFigureYearLink
Deflated Sharpe Ratio: when many strategies are tried and only the best is reported, the reported performance is inflated; the DSR corrects for selection bias under multiple testing and non-normal returns
Journal of Portfolio Management 40(5) 94-107. Core claim: failing to account for the number of trials behind a discovery inflates expected performance
2014
Probability of backtest overfitting: standard safeguards such as hold-out samples are often unreliable for investment backtests; the authors propose estimating PBO via combinatorially symmetric cross-validation
Journal of Computational Finance 20(4) 39-69 (April 2017; online Sep 2016; SSRN 2015)
2015-2017
Minimum backtest length: with only 5 years of data, no more than 45 independent model configurations should be tried, otherwise an in-sample Sharpe ratio of about 1 with zero out-of-sample Sharpe is almost guaranteed; analysts rarely report how many configurat
5 years -> max 45 configurations; MinBTL < 2 ln(N) / E[max_N]^2
2014 (version of 1 Apr 2014; article in Notices of the AMS)
Harvey and Liu: the common 50% haircut to a backtested Sharpe ratio is wrong; the correct haircut depends on the number of tests (data mining)
haircut often above 50% when annualised Sharpe is below about 0.4; at most about 25% above 1.0; quote: 'a serious mistake to use the usual 50% haircut'
2015 (J. Portfolio Management, Fall)
Survivorship bias in performance data: analysing a sample truncated by survival creates the appearance of predictability, strong enough to account for the evidence of fund performance persistence
Review of Financial Studies 5(4) 553-580
1992
Applied to retail traders: most day traders quit, so visible long-running accounts are survivors (see Barber et al. working paper, topic 2)
over 75% quit within 2 years; 15% survive 3 years
2017
Evidence
Verification standards for track records
SourceFigureYearLink
CFTC Rule 4.35 (CTAs): actual performance of all accounts directed by the CTA must be disclosed, with prescribed capsule data and a 'past performance not necessarily indicative' legend
Capsule: largest monthly drawdown and worst peak-to-valley drawdown (5 years + YTD), annual returns for 5 years, accounts opened/closed; performance current within 3 months; supporting records kept (Rule 1.31); proprietary results labelled separately
current eCFR (read 2026)
CFTC Rule 4.41(b): simulated or hypothetical results must carry a prescribed cautionary statement in immediate proximity ('These results are based on simulated or hypothetical performance results that have certain inherent limitations...')
Statement may be the CFTC text or one prescribed by NFA (Compliance Rule 2-29(c)); cover-page-only statement insufficient. This is the 2007 adopting release; check eCFR for later amendments
2007 (72 FR 8106, 23 Feb; effective 26 Mar 2007)
SEC Marketing Rule (Advisers Act Rule 206(4)-1): general prohibitions; testimonials/endorsements need disclosures, written agreement and oversight; gross performance only with net of equal prominence; 1-, 5- and 10-year periods; hypothetical performance only u
Effective 4 May 2021; compliance date 4 Nov 2022. Applies to SEC-registered investment advisers, not to every trading educator
2021 (86 FR 13024, 5 Mar 2021, Release IA-5653)
Rule text (d)(1), (d)(2), (d)(6) quoted from the CFR
net with at least equal prominence to gross; one-, five- and ten-year periods with equal prominence, ending no earlier than the latest calendar year-end; hypothetical performance requires relevance policies and sufficient information on criteria, assumptions, risks and limitations
current
SEC staff compliance guide for the Marketing Rule
summary of the above; hypothetical performance requires relevance policies and supporting information
2021-2022
GIPS verification is firm-level and voluntary: it does not verify individual composites, performance results or the claim of compliance; accuracy of a specific composite needs a separate performance examination
'Verification does not provide assurance on the accuracy of any specific performance report'
GIPS 2020 standards (Q&A effective 1 Jan 2020)
GIPS template: verification is an independent third-party review of a firm, not of individual products; encouraged, not mandatory
firms are either verified or not; no levels
2021
GIPS composite reports: at least 5 years of annual performance building to a minimum of 10 years; time-weighted returns net of transaction costs
5 -> 10 years
2026 refresher reading
Myfxbook 'verified' = (1) trading history matches broker data via investor password and (2) the person controls the account (authorisation-password or pending-order test). It does not claim to verify profitability, absence of other accounts, or deposits/withdr
two checks: Track Record + Trading Privileges
current (read 2026)
Gaps
What we could not find
- AMF 'CFD 2014-2018' study requested in the brief: NOT FOUND. The only AMF retail Forex/CFD loss study located covers 2009-2012 (published 2014). No later AMF loss-percentage study on CFD or crypto found in 5 searches
- CONSOB: no loss-rate statistic of its own located. Its 2024 household report gives no loss figures and only cites ESMA warnings
- Peer-reviewed or regulator study measuring returns of buyers of PAID trading courses vs non-buyers: NOT FOUND (searched academic sources, FINRA Foundation, CONSOB, AMF, FCA/ASIC pages, CFTC)
- NFA Compliance Rules 2-29 and 2-36 govern promotional material, including hypothetical results. The NFA page read only lists the topics; the rule text was NOT read
Figures marked secondary in our source file are left out of this page. Education only, not investment advice. Report an error.