The claim
The claim, in its generic recruiting form: “Anyone can learn to day trade for a living. It takes screen time, discipline, and a proven system — most people fail because they quit before the skill compounds.”
Steelmanned properly, this is a respectable-sounding argument. Trading is a skill domain with fast feedback: you place a trade, the market grades it within minutes. Most skills with fast feedback — chess, poker, typing — improve with deliberate practice. Profitable traders demonstrably exist; institutional desks pay them salaries. And the failure statistics quoted against day trading usually come from surveys or broker marketing, both of which have obvious sampling problems. The claim’s defenders can argue, fairly, that most failure data captures dabblers: people who tried for two weeks, lost their first stake, and left before any learning could occur. To test “learnable,” you need to isolate the people who actually put in the hours.
There is also a definitional point to concede up front: “learnable” does not require that everyone succeeds. Driving is learnable even though some people fail the test. So a fair test of the claim cannot demand a 100% success rate — it should ask whether the success rate among people who genuinely put in the practice looks like a skill being acquired (most committed learners reach competence) or like a lottery being played (a fixed, tiny fraction wins regardless of commitment).
That is precisely why the study we examine here matters. It does not sample. It follows everyone — and it lets us look specifically at the people who persisted.
The evidence
Researchers at the São Paulo School of Economics analyzed the complete trading records of every person who began day trading in the Brazilian equity futures market between 2013 and 2015 (SSRN abstract 3423101, accessed July 2026). Note the design: not a survey, not a broker’s self-selected client list, not a sample. A population. Every entrant, tracked through their actual trades. Survivorship bias — the standard defense against bad trading statistics — is structurally excluded, because nobody could drop out of the dataset.
The results for the persistence cohort, the very people the “you just quit too early” argument says we should examine:
| Outcome, among traders persisting more than 300 sessions | Share | Source |
|---|---|---|
| Lost money | 97% | SSRN 3423101 |
| Earned more than the Brazilian minimum wage | 1.1% | SSRN 3423101 |
| Earned more than a bank teller’s salary | 0.4% | SSRN 3423101 |
Read the table against the claim. The claim predicts that persistence converts into profitability: put in your 300 sessions and the skill compounds. The 300-plus-session cohort is the claim’s best-case population — the disciplined ones, the ones who did not quit. Ninety-seven percent of them lost money.
The two lower rows measure the “for a living” half of the claim. Losing money is one failure mode; the subtler one is earning less than the boring alternative. Of every persistent day trader in an entire national market over a three-year entry window, 1.1% did better than minimum wage, and 0.4% did better than an entry-level bank job. The relevant comparison for “day trading for a living” is not zero — it is the wage you gave up to sit at the screen. By that benchmark, roughly 99 in 100 persistent traders would have been financially better off taking the job the trading course told them to escape.
A note on what “300 sessions” means in lived time: at the market’s daily rhythm that is well over a year of showing up. These are not tourists. The dataset’s persistent cohort is the population that trading educators describe as doing everything right — consistent screen time, sustained commitment — and the outcome distribution barely moves in their favor.
We flag one documentation caveat in the interest of forensic honesty: we cite the study via its SSRN record and attribute it to São Paulo School of Economics researchers; readers wanting the full author list and journal status should open the SSRN page directly. The headline figures above are as stated in that record.
What the evidence supports — and what it does not
The evidence supports this statement: in at least one complete, population-level dataset, day trading persistence did not convert into profitability for 97% of those who persisted, and converted into a living wage for roughly 1 in 100. If day trading were learnable by most people the way its marketing implies, the persistent cohort is exactly where the learning would show up. It does not show up.
What the evidence does not support: a universal claim about all markets and all periods. This is Brazilian equity futures, 2013–2015. Fee structures, market microstructure, and competition differ across venues and eras. The study also cannot rule out that some tiny subpopulation has durable skill — 0.4% of the cohort did out-earn a bank teller, and the study cannot tell us whether that group is skilled or lucky. Nor does this study cover swing trading, long-term investing, or index accumulation, which are different activities with different evidence bases; nothing here is a claim about buy-and-hold investing.
Return to the driving-test standard we set in the claim section. If day trading were learnable the way driving is learnable, the 300-plus-session cohort should look like a population of licensed drivers: most of the committed reaching functional competence, with a failing minority. Instead the distribution is inverted — a 97% failure rate among the committed, with a winner fraction so small it is indistinguishable, from outside, from a lottery’s. The data cannot tell us whether the 0.4% hold skill or a winning ticket. It can tell us that “put in the hours and you will get there” describes, at best, 1 in 100 of the people who put in the hours.
But observe what the claim needs versus what it has. “Learnable by most people” is a claim about the distribution, not the existence of winners. The best available population-level distribution says: 97% of the committed lose. A course seller who shows you three profitable students has shown you the 1% — or the 0.4% — and left the denominator in the drawer. The denominator is the autopsy finding.
One further pattern is worth naming because it explains why this claim survives contact with these numbers year after year. Everyone selling the learnable-living story has a visible incentive to show you winners: course sellers monetize enrollment, brokers monetize order flow, and profitable-looking traders monetize attention. Nobody in the distribution chain has an incentive to publish the denominator — which is why it took academic researchers with exchange-level records to produce it, and why this study, and not any practitioner’s track record, is the primary source on this table.
Cause of death: the denominator.
What would change this verdict
- A comparable population-level study — complete broker or exchange records, all entrants, no sampling — in a different market and era, finding that a majority or even a substantial minority (say, over 20%) of traders persisting 300-plus sessions are profitable after costs.
- Evidence of a learning curve: a population dataset showing per-trader performance improving materially with accumulated experience, such that later-career profitability is meaningfully higher than early-career profitability for the typical persistent trader.
- Audited, multi-year track records for a sizable cohort of graduates of any retail day-trading education program, showing median earnings above the local wage benchmark. Audited is the operative word; screenshots are not records.
Any of these would force a rewrite of this page. We are not holding our breath, but we are holding the standard.
Source docket
- São Paulo School of Economics researchers, day trading population study, complete trading records of all entrants to Brazilian equity futures day trading 2013–2015 — SSRN abstract 3423101 — https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3423101 (accessed July 2026)
A short docket, deliberately. One population-level primary source outweighs a stack of vendor surveys, and we prefer not to pad the docket with secondary write-ups whose additional figures we could not verify against primary records.
Related autopsy: 70 MLM income disclosures, read so you don’t have to.