9 Out of 10 Lose: The Real Odds in Retail Trading, and the Only Edge That Matters

Three different markets. One pattern.

India's derivatives market: SEBI's August 2026 study found 87.7% of individual F&O traders lost money in FY26, with aggregate losses of about ₹91,685 crore. Options drove 92% of losses. Traders who lost two years running and kept going lost again in the third year about 90% of the time.

Prop firm challenges: FTMO's own data puts failure at about 92%. Of those who pass, only around 20% ever receive a payout — real odds of getting paid around 1.6–2%. And 70–79% of failures come from one cause: breaching a drawdown limit on an ordinary bad morning.

AI trading apps: Across 8 apps on Hyperliquid, 1,365 users traded $21.0M in 30 days — and in 4 of the 8 apps, users lost money overall, with fewer than half the users in profit in most apps. The advertised returns behind those apps? 17 of 18 claims checked had no independently verifiable record.

Three funnels, three products, three marketing stories — and the same outcome distribution: roughly 9 out of 10 lose, and the house (the broker, the prop firm, the app) gets paid either way.

Why the pattern repeats

It is not that 90% of people are stupid. It is that all three funnels sell the same thing: a shortcut past validation. A signal instead of a process. A bot instead of a backtest. A challenge fee instead of a checklist. And the market charges full price for unvalidated ideas — in F&O losses, in failed challenge fees, in app subscriptions that quietly lose a little every month.

The only edge that changes the odds

It is not a better indicator, a smarter bot, or a bigger account. It is a boring, repeatable standard applied before money moves:

• Recompute every claim from raw data (never from a claimed column)
• Beat the dumb baselines (always-bullish, no-change)
• Hold up fold by fold across time (not just on average)
• Count the trials and correct for them
• Model costs honestly
• Forward-test before paying anything — a challenge fee, a subscription, or a live account

That standard does not guarantee profit. Nothing does. What it does is move you out of the 90% who pay to discover what validation would have told them for free, and into the small group whose ideas have at least survived contact with reality before their capital did.

The odds in retail trading are brutal, and they are published, and they are consistent. The question is not whether you can beat them with a better prediction. It is whether you will validate before you pay — because everyone who skipped that step is already counted in the 90%.

BuildIT Index exists so you can run that standard mechanically: hand it a prediction file, get a recomputed accuracy, baseline scores, walk-forward folds, leakage gate, multiple-testing correction, calibration, and a verdict with an exit code.

The odds in retail trading are brutal, and they are published, and they are consistent. The question is whether you will validate before you pay.