Three checkpoints in this guide
Follow the full walkthrough in order, or jump directly to one of its main sections.
Experience can improve a trader. It can also make a losing habit more practiced. The difficult part is telling learning apart from selection: when weak performers quit and stronger performers remain, the average experienced trader can look better even if simply surviving did not make any individual better.
The best available long-run studies do not support the slogan “trade long enough and you will become profitable.” They do show that some traders have skill, many lose after costs, and an experienced-looking group is partly a group from which many people have already exited. Those findings are about specific markets and periods—not a worldwide 2026 pass rate or a prediction for your next trade.
The short answer
Trading experience is useful only when it changes a repeatable decision and that change survives later, after-cost testing. Years on a platform, a long streak of activity, or a profitable month cannot establish that. In a study of Taiwan stock day traders from 1992–2006, losing traders were more likely to quit, yet many continued after extensive losses. A separate analysis of the same market found a small group whose prior results predicted positive net abnormal returns in the following year. That is evidence that skill can exist, not that time alone creates it. Learning study; skill study.
Three different things that people call “getting better”
| Observation | What it could mean | What it does not establish |
|---|---|---|
| A trader's recent P/L improved | Execution changed, conditions improved, exposure rose, or luck helped | That the process will work in the next market regime |
| Older accounts outperform newer accounts | Survivors are selected, some traders learned, or the groups trade different products | That simply remaining active causes better returns |
| A defined setup performs better in a later untouched period | More credible evidence that a process change helped | That all of the trader's setups or accounts have an edge |
The denominator matters. A study of everyone who started asks a different question from a survey of people still trading today. It is possible for a surviving minority to be skilled and for the average entrant to lose. Neither fact contradicts the other.
What the long-run evidence actually says
Barber, Lee, Liu, Odean and Zhang used the Taiwan Stock Exchange's complete transaction history from 1992–2006 to examine whether day traders learned about their ability. Their later published paper reports that unprofitable traders were more likely to quit, but also that losing traders generated a large share of day-trading volume and many persisted despite negative records. Its research question is about rational versus biased learning—not about modern US futures prop evaluations. Original paper.
In a separate analysis of Taiwan day traders, Barber and coauthors ranked traders by past performance and then tested the following year. In a typical year, roughly one in five of the more active day traders had positive abnormal returns net of the study's transaction-cost assumptions. Fewer than one in a hundred of the entire day-trader population showed reliably positive net abnormal returns in the following year. The difference is the point: a profitable observation is easier to obtain than evidence of repeatable skill. Do not convert that historical Taiwan finding into “99% of all traders everywhere fail.” Original paper and methodology.
Brazilian equity-futures evidence provides another warning against “persistence proves progress.” Chague, De-Losso and Giovannetti studied people who began day trading in 2013–2015. Among those who persisted for more than 300 trading days, 97% lost money in their studied outcome. That is a selected persistent subgroup in one market, not a universal rate for every trader or prop program. It does show why accumulating hundreds of trading days is not, by itself, a useful definition of improvement. Original study.
Finally, a study of retail forex trading found that traders increased size, size variability and activity after gains even where recent performance did not predict future success. The effect was stronger among novices. This is a plausible false-learning mechanism: a winning week changes behavior before it establishes skill. It is not a claim that every individual trader responds this way. Original research abstract.
The defensible takeaway
Citation-ready synthesis: Experience is not a performance metric. To test whether a trader improved, compare a stable, after-cost process on later trades that were not used to choose the process—and keep traders who quit in the denominator of any population claim.
This is TSB's synthesis of the cited research and a measurement rule, not a new TSB survey result. “Experienced trader” alone is not a proxy for rule fit, payout eligibility or a promised return.
Why survivor-only charts can mislead
Imagine a cohort of new traders. Some quit quickly after losses; some continue; a few find a repeatable method. A later chart of active accounts contains more of the latter two groups and none of the departed accounts. The active group's average may rise even if many continuing individuals have not improved. That is survivorship bias.
The converse mistake is also possible: saying no one can learn because the cohort average is negative. A negative average does not rule out a skilled minority. The Taiwan persistence result makes that distinction visible. An honest report therefore needs both views: entry cohort (all who began, including exits) and within-trader change (the same person's later, comparable decisions). If a firm publishes only its current successful accounts, the reader cannot infer the entrants' chance of success.
How to test whether your trading improved
- Freeze the question first. Choose one market, account, setup definition and strategy version. State what change you made—such as no trades outside the planned session or a different exit rule—before looking for its best-performing slice.
- Reconcile the record. Use complete closed trades, fills and charges. Keep deposits, withdrawals, funding payments, commissions and account transfers separate from trading result. Missing trades are not zero-result trades.
- Compare like with like. Examine net expectancy, realized versus planned risk, rule breaks, worst session and drawdown. If sizing changed, normalize results by planned risk and inspect account-level dollars. A larger P/L from larger exposure is not necessarily improved decision-making.
- Reserve later evidence. Do not select a “best” session or setup on the same data used to prove it. Review a subsequent untouched period under the frozen definition. If the strategy changes again, start a new version rather than merging the histories.
- Report uncertainty and constraints. A short later period can reverse by chance. Correlated trades, changing volatility and partial fills make a raw trade count look more independent than it is. Say “promising but unconfirmed” when the evidence does not justify a stronger conclusion.
There is no universal 60-, 100- or 200-trade magic threshold that proves an edge. Required evidence depends on the effect size, variance, trade dependence, costs, number of tested variants and the decision's stakes. The existing sample-size guide owns that calculation; this page owns the different question of learning over time.
What TSB can and cannot add
TSB can organize a trader's own imported history into setups, sessions, execution and costs, then let its AI Coach question the selected evidence and point back to the supporting trades. That is a way to investigate why this trader's process changed. It does not turn a personal journal into a random sample of traders, fill absent fees, reconstruct an unrecorded equity path or prove that a new habit caused better returns.
If you are evaluating a prop program, test the exact phase and rule path against the same frozen trade history. A better after-cost setup can still be a poor fit for a particular trailing-drawdown or payout rule. Conversely, passing one simulation does not prove a durable trading skill. Start with your own trading history and state which parts of the answer remain unverified.
Bottom line
The research supports a careful answer: some traders are skilled, many do not become profitable merely by continuing, and survivor-only comparisons exaggerate apparent progress. Improvement is a within-trader claim that needs a defined change, complete costs and later evidence. “I've traded for years” is biography; “this frozen decision rule held up on new, comparable trades” is a testable statement.
Source and scope note
This guide synthesizes published studies, not a TSB-wide performance analysis. Taiwan stock day trading (1992–2006), Brazilian equity-futures entry cohorts (2013–2015) and one retail-forex sample have different markets, units, periods and outcomes. Their percentages are not pooled or presented as current global rates. The practical test is an editorial framework; it is not proof that journaling, AI analysis or any specific firm causes profitability. See the broader trading statistics guide for population definitions and the editorial methodology.
