Exchanges, regulators and brokers have all measured what happens to ordinary people who try. In every dataset most of them lose, practice makes it worse, and the people most sure they need no special knowledge are the ones the data shows lack it.
A few days ago I got talking to a taxi driver who told me he is learning to trade. Not investing: trading, the kind where you open a position in the morning and close it before dinner. He was clear about one thing. He doesn't need maths, he said, or economics, or finance, or statistics. You learn the charts, you practise, you keep your discipline, and the money follows.
He isn't unusual. Versions of that sentence are the pitch of a whole industry: trading courses, signal groups, "funded trader" challenges, zero-commission apps, and the people on YouTube and TikTok showing their winning days. The promise is a way out of wage work that needs nothing but time and nerve.
It is also one of the few promises in personal finance that has been measured carefully. Brazil's securities regulator gave researchers the records of every person who started day trading futures over three years. Taiwan's exchange shared fifteen years of every trade by every day trader. Nine European regulators counted how many retail clients lost money on leveraged products before the EU capped them in 2018, and brokers in Europe are now required to print the share on their front pages. US researchers have traced billions of dollars of retail options trades.
I collected those studies, read them in full and checked their arithmetic where the counts allowed it. This essay goes through what they say, belief by belief, and ends with the one the taxi driver started with. The short version:
If trading were a craft like driving or cooking, the people who did it longest should be the ones most likely to be making money. The best test of that is a Brazilian study by Fernando Chague, Rodrigo De-Losso and Bruno Giovannetti. With data from the securities regulator, they followed all 19,646 people who started day trading mini-index futures between 2013 and 2015, and sorted them by how many days they kept at it.
It goes the wrong way at every step. Of the people who traded once, 29.8% came out ahead, which is roughly what luck gives you. Of the 1,551 who persisted for more than 300 days, 3.0% did. Their average result was a loss of $48.81 a day. Seventeen of them earned more than the minimum wage. Eight earned more than a bank teller's starting pay of about $54 a day. The single best trader averaged $310 a day, with a daily standard deviation of $2,560. Even the best result in the sample swung by eight times its average from one day to the next.
The Netherlands shows the same shape in a different product. The Dutch regulator looked at 35,000 people trading turbos, a leveraged certificate, and found that 68% lost money overall. Among those who made up to ten transactions, 64% lost. Among those who made 500 or more, 88% did. In the US, Brad Barber and Terrance Odean's study of 66,000 brokerage households in 1991–96 found that the busiest fifth earned 11.4% a year against 17.9% for the market, while the least active roughly matched it.
Taiwan has the largest dataset: every day trader on its stock exchange from 1992 to 2006, about 450,000 a year. In any single year about one in five is profitable after fees. But when Barber and his co-authors asked how many were predictably profitable, profitable often enough that it couldn't be luck, the answer was at most about 4,000 a year, under 1%. In the published version of their work, 97% of day traders are likely to lose money in future trading. Day traders as a group lost money in every one of the fifteen years.
That doesn't mean nobody can do it. The 500 best Taiwanese day traders of one year earned 0.38% a day net of costs the following year, which is real money. The trouble is that you can only find out whether you're one of them by losing money first, and the odds are about a hundred to one.
Verdict: false as stated. Practice moves most people further into loss, not out of it.
It would be easy to cherry-pick. So here is every loss rate I found from a study, a regulator or a broker, on one chart. The studies use different definitions (accounts or clients, one year or four), so the exact numbers aren't comparable. The direction is.
Three things about this chart matter more than any single dot.
First, the loss rate doesn't move with the market. Poland's regulator counted the share of active CFD clients losing money each year from 2012 to 2017: 81, 81, 80, 82, 79, 80%. Markets rose and fell in those years; the loss rate didn't. The European Securities and Markets Authority concluded that the persistence “indicates a structural feature of the return profile”. It's a property of the product, not of a bad year.
Second, the costs are enormous relative to the money involved. In Norway, transaction costs averaged 37% of clients' account equity. In Taiwan, day traders lost a little before costs (0.07% a day) and a lot after them (0.24% a day). Most retail trading is a game that is slightly negative before the house takes its cut, and clearly negative after.
Third, leverage does the damage, and regulating it works. When the US capped retail forex leverage in 2010, losses among highly leveraged traders fell by about 40%. The broker disclosures from 2026, in grey, are lower than the regulator studies from 2014–17, which is what you'd expect after the EU's 2018 caps. They're the brokers' own calculations and not directly comparable, but they're the one hopeful pattern in the data. Even so, the lowest of them is 55%.
Verdict on “losses come from bad discipline, not the game”: mostly false. Psychology is real, but it works through a product built to amplify it.
The usual answer to all of this is that losses are tuition. Everyone loses at first; you learn from it and get better. This, too, has been measured.
France's market regulator, the AMF, followed about 2,000 retail forex and CFD clients who were active in all four years from 2009 to 2012. If losing is how you learn, this is the group that should have learned most.
After one year more than 82% were down, after two more than 85%, after four nearly 88%, and the average loss per client grew from €4,989 to €26,745. The AMF's phrase, repeated by ESMA, was that “there is no learning curve”.
Brazil agrees. The persistent traders lost $47.34 a day on average over their first 250 days and $51.65 a day after that. The trend in profit over time was flat, statistically indistinguishable from zero. In Taiwan, people with a history of losses kept trading almost exactly as often as people with a history of profits.
People do learn something. More than three-quarters of Taiwanese day traders quit within two years, and losers quit faster. A study by Amit Seru, Tyler Shumway and Noah Stoffman found that a large part of what looks like “learning by trading” is in fact poor traders leaving, and that ignoring this badly overstates how fast anyone improves. The lesson most people draw from their losses, eventually, is the right one: stop.
Verdict on “I'll learn from my losses”: mostly false. Those who stay don't improve on average, and their losses keep adding up.
Commission-free apps changed who trades. The pitch is that the costs that used to protect professionals are gone, so the small trader finally has a fair shot. What happened instead is that the cost moved from a line you can see to one you can't.
Svetlana Bryzgalova, Anna Pavlova and Taisiya Sikorskaya traced US retail options trading from November 2019 to June 2021, a period when retail traders became about 62% of US options volume.
Commissions came to about $0.9 billion. The gap between what retail traders paid and the midpoint price came to $6.4 billion, seven times as much. Half of retail options trades were in contracts expiring within a week, where the average quoted spread was 12.6% of the price, and 23.5% on small trades of $250 or less. In aggregate, retail traders lost $2.1 billion.
The pattern repeats. Stocks that Robinhood users piled into most on a given day went on to underperform by 4.7% over the next 20 trading days. More than three-quarters of retail S&P 500 options trades are now in contracts that expire the same day. And in Brazil, as the authors of the day-trading study note, brokers cut day-trading fees to zero in late 2018 “to lure in new investors to their courses and other services”. Free trading is how the customer is acquired, not what makes the money.
To be fair, some retail traders do profit here. Retail sellers of short-dated options make significant profits after costs. They are on the same side as the wholesalers: selling people the lottery tickets they want to buy.
Verdict on “zero commission put the edge with small traders”: false. The edge is with whoever sells retail what it wants to buy.
The newest variant is the prop-firm challenge. You pay a fee, trade a test account, and if you hit a profit target without breaching loss limits, the firm “funds” you and shares the profits. The pitch: you trade the firm's money, not yours.
The evidence here is weaker than anywhere else in this essay, because almost all of it comes from the industry. With that warning, here is what the industry itself has reported.
In the larger dataset, about 6 in 100 ever received a payout. The average payout was 4% of the account size, while the average trader spent about $800 on challenges per account. Many who got paid will have paid more in fees than they received. At the second firm, 1 or 2 in 100 were paid, and when that was reported, 1,272 clients had been waiting over a year for payouts they were owed.
The “funded account” is often not a real account at all. Belgium's regulator, the FSMA, warned that consumers in these schemes “never execute real trades”: the trading is simulated and the firm controls it. What you risk isn't trading capital but the fee, paid again after each failure. It is closer to a skill-game entry fee than a job.
One case is often cited here and shouldn't be. In 2023 the US CFTC accused a firm called My Forex Funds of taking $310 million in fees from 135,000 customers while acting as counterparty to their trades. In May 2025 the case was dismissed with prejudice and the CFTC was sanctioned for its conduct. Those allegations were never proven.
Verdict on “you trade without risking your own money”: technically true, substantively misleading. You risk the fee, and more than nine in ten fee-payers never see a payout.
The strongest evidence most people have that trading works is that they can see people doing it: screenshots, live streams, cars. Here the data is unusually direct.
A team of finance researchers (Ali Kakhbod, Seyed Kazempour, Dmitry Livdan and Norman Schuerhoff) scored about 29,000 “finfluencers” on the platform StockTwits by whether following their advice made money. 29% were skilled, with followers earning about 2.6% a month above the market. 16% had no measurable skill. 55% were antiskilled: following them lost about 2.3% a month. The unskilled and antiskilled had more followers and were more active. The loudest voices are, on average, the worst.
Simple arithmetic does the rest. If 3% of persistent traders are net profitable, a platform with 100,000 aspiring traders still has about 3,000 winners to show. You will never run out of success stories, because selection guarantees a supply of them. They tell you nothing about the base rate.
Taiwan shows who the winners are and how little of the activity they account for.
Three-quarters of the volume comes from people who lost money the year before. The two groups that go on to earn net profits, experienced traders who were already profitable, do under 10% of it.
Some of the most visible success is the product itself. In 2022 the US Federal Trade Commission sued Warrior Trading, whose ads read “Learn How I Made over $101,280.47 in Verified Profits Day Trading Part Time in Under 45 Days”. The complaint quotes the company saying it does “not track the typical results” of its customers, and alleges that its own simulator data showed 74% of accounts losing money and only 10% making more than $90. Warrior paid $3 million in redress. RagingBull paid $2.4 million and, the FTC said, “did not track its customers' trading results”. IM Mastery Academy, which the FTC alleges took more than $1.2 billion from customers since 2018, including through college social-media pages, had salespeople flaunting lifestyles “purportedly funded by trading profits and multi-level-marketing commissions”. By the company's own data, 60% of customers quit within a month. Its lead defendants agreed in May 2026 to a $795.8 million judgment and a ban on selling trading training.
Verdict on “the traders I see online are proof it works”: false. What you see is filtered, often commercial, and if anything worse than average.
Which brings me back to the taxi. I think the belief that trading needs no particular knowledge is the one underneath all the others. If it needs none, then practice, imitation, discipline and learning from mistakes should be enough, and the data above says they aren't. So what does it take? Item by item: here is what a trader has to understand, and what surveys show people actually know.
Under EU rules, a retail trader can borrow up to 30 times their stake on the major currency pairs, which means putting down 3.33% of the position as margin. Brokers must close the position when losses eat half the margin. So a 1.67% move against you ends the trade. Before the cap, ESMA noted, some providers offered 100:1, where the same rule would close you out after a 0.5% move.
The US FINRA Foundation tests investors' knowledge every few years. In its 2021 survey of 2,824 people who hold investments outside retirement accounts, it asked: you put in $500 to buy $1,000 of stock on margin; the stock falls 50%; how much of your $500 is left? The answer is nothing. 23% got it right. Among people who had actually bought on margin, 76% got it wrong.
The same survey asked what a call option with a $50 strike is worth on expiry day if the stock is at $40. It's worth nothing. 13% of investors knew. People who had traded options were more likely to get it wrong than investors in general, 62% against 39%. FINRA's comment: “a little bit of knowledge can be dangerous”. ESMA, explaining the 2018 restrictions, said CFD prices and costs are too complex for most retail clients to assess, partly because fees are charged on the whole leveraged position rather than the money you put in. And the options spreads above, seven times the visible commissions, are exactly the cost that people without this knowledge don't see.
This is the part the taxi driver is most wrong about, and the one that matters most. Suppose you have a strategy that really does win 52% of trades where you risk and gain the same amount. How many trades before you can tell it apart from tossing a coin?
At 52%, about 2,400 trades. At 51%, nearly 10,000. Over a few hundred trades, a strategy with a real edge and one with none look the same, and a run of good luck looks like talent. That's why the Taiwan researchers needed years of data to find the under-1% who were predictably profitable when one in five looked profitable in any given year. And it's why one of the most damaging beliefs in the FINRA survey is a statistical one: 44% of investors think past performance is a good guide to future results.
Every trade has a counterparty. In Taiwan from 1995 to 1999, individual investors lost about $6.4 billion a year through trading, 2.2% of the country's GDP and 3.8 percentage points of their annual return. Institutions gained 1.5 points a year after costs, largely by supplying the other side of individuals' impatient orders. In Brazil, high-frequency trading firms went from 11.6% of mini-index deals in 2012 to 41.9% in 2017, and the more of them there were, the more individuals lost. The person across the table is a firm that hires the mathematicians and statisticians the taxi driver thinks he can do without.
Mostly not. On FINRA's ten-question quiz, US investors averaged 4.7 correct. Those under 35 got 4.0 right and 4.0 wrong, and those with under two years' experience got more wrong than right. But ask them to rate their own knowledge:
65% of under-35s rate their knowledge as high, the same as investors over 55, who score far better. FINRA's conclusion is the most precise description of the taxi driver I've read: younger investors are “not simply uninformed, but potentially misinformed”. A separate FINRA and CFA Institute survey of Gen Z investors found that 48% think they know more about investing than their parents.
Confidence costs money. Barber and Odean followed 1,607 investors who moved from phone to online trading in the 1990s. Before switching, they beat the market by more than 2% a year; afterwards they traded more and lagged it by more than 3%. The authors' explanation is an “illusion of knowledge”: more information on a screen makes people feel they understand more than they do. In another study, men, who are more overconfident about finance, traded 45% more than women and lost more to it.
And the skills show up in results. Mark Grinblatt, Matti Keloharju and Juhani Linnainmaa matched Finnish men's IQ scores from military service with their trading records from 1995 to 2002. Investors in the top IQ group earned 2.2% a year more than the bottom group, and 4.9% more once the timing of moving money in and out of the market is counted. They held on to losing stocks less stubbornly, sold for tax reasons when they should and placed orders when spreads were narrow. Their edge wasn't eaten by trading costs.
Knowledge is necessary, but it isn't enough. The Finnish study compares IQ groups with each other; it doesn't say the high-IQ group beat the market. And the professionals who have all of this knowledge, plus research teams and data budgets, mostly fail too. Over the 15 years to December 2024, 89.5% of actively managed US large-cap funds did worse than the S&P 500, and about 98% did after adjusting for risk. Fewer than half the funds even survived the period. (Those figures come from S&P, which sells index products and has an interest in the result, but the fund data underneath is the standard academic database.)
The taxi driver is right about one thing: you don't need a degree. You need a handful of concepts: expected value, variance, what leverage does, what things really cost, and who you're trading against. The data says most retail traders don't have them, and the ones least likely to have them are the most confident.
Verdict on “I don't need maths, statistics, economics or finance”: false, and it matters more than it looks. Not knowing the maths doesn't remove it. It just hides it.
None of this is new, and yet the belief is spreading among young people in particular. The surveys suggest why.
Young investors learn about money from social media first. In the FINRA Foundation and CFA Institute survey of US investors aged 18–25, social media was the top source of investing information, ahead of search engines and family. (Part of that sample was recruited through social-media ads, so the share is probably overstated.) Half said they had made an investment out of fear of missing out. Among those with a high appetite for risk, 45% gamble at least monthly, and the frequent gamblers were far more likely to trade options or buy on margin.
And the motive is telling. In the Canadian part of the survey, 42% of Gen Z investors ranked “a source of income outside my job” among their top goals, against about a quarter of older investors. Trading is sold as an exit from wage work. I've written before about why work hasn't got shorter even as output per hour tripled; it isn't surprising that people look for a way out on their own. It's just that this way out is, by the numbers, mostly a way of transferring their savings to the people who sell it.
A piece like this can turn into a sermon, so here is what I'm not claiming.
Not “don't”, since he's an adult and it's his money. But I'd ask him to do five things before he puts in more than he can afford to lose.
The odds of making a living this way are about the same as the odds of making a living playing poker, and for the same reason: a small skilled minority is paid by everyone else, and the house takes a cut of every hand. The difference is that nobody tells you poker needs no maths.