Most trading strategies don't survive real data. We check, and publish every result.
We test popular and academic strategies the way a quant fund would: real spreads, the market's true clock, pessimistic fills, and a holdout nobody has looked at. Twelve gates. The failures go on the site too, including our own.
Eleven gates passed. Then the holdout.
Our most instructive test so far: a strategy that looked ready for a funded account until the one piece of data it had never seen.
Nasdaq breakouts on high-fear days
Passed eleven of twelve gates, then lost money on the holdout nobody had seen.
Read the full testLatest tests
Each page has the rules we tested, the costs, a chart rebuilt from the trade list, the full scorecard and the files behind every number.
ICT Silver Bullet (10:00-11:00 New York)
A fair-value gap traded towards the nearest liquidity pool. A simple backtest shows a profit; realistic fills and costs turn it into a loss.
Momentum pullback to the 9 and 21 EMA
Buy the strongest stocks when they dip to a moving average and turn up. Negative at both fill bounds over 16 years.
Machine-learning intraday forecaster
70 features, gradient boosting, a promising IC. The signal was oil prices read five hours early.
Weak-day reversal (lower high, lower low)
Buy the break of a weak day's high. Small, positive, and not distinguishable from noise.
Opening Volatility Stretch on DAX, Bund and Nasdaq
The DAX and Bund results came from a frozen pre-open quote. Nasdaq is positive but unproven.
We mined 10,248 strategy rules on a GPU
Breakouts, sweeps, VWAP reversion and engulfing bars on five indices. The best result was weaker than what noise produces.
Twelve gates, three questions.
A backtest has to answer three questions in order. Is it real, or a bug? Is it more than luck? And for funded accounts: can it survive the firm's rules? One failed veto ends the test.
Is the edge more than luck?
The backtests that lie the most are not unlucky. They are bugs.
Every one of these produced a spectacular result in our own lab before a gate caught it.
A price from the future
A filter grouped by calendar date read a price 22.5 hours after the trade. Result: t 7.52 and twelve green years. Real result: t −1.52.
A minute where nobody traded
The "open" bar was a frozen quote one minute before Frankfurt opened, and the simulator filled at the trigger after the gap. +806R gross became t 0.38.
A clock five hours off
Oil prices stamped in New York time were read as London time, and a 70-feature model learned to read them early. After the fix it matched a shuffled control.
Test your own trades against prop-firm rules.
Upload an MT5, cTrader or CSV history. The simulator replays it under each firm's published rules (daily limits at the firm's own midnight, trailing drawdowns, consistency) and runs the same luck tests we use in the lab. Your file never leaves your browser.
What it reports
- Simulated pass and blow-up rates, built from blocks of your own trading days
- A luck test deflated for the number of strategies you tried
- The largest risk per trade at which one bad day still fits the daily limit
- The same history under FTMO, Topstep, FundingPips and The5ers rules
Questions
Is this investment advice?
No. We publish historical tests of trading rules on our data, and a free simulator for your own trade history. We do not recommend any trade, strategy, broker or firm, and we do not sell signals.
Why would we publish failures?
Because they are the useful result. Most strategies sold or shared online have never been through costs, the right clock and a holdout. A clear "no, and here is why" saves you the money and months a live account would take to say the same thing.
Has anything survived all twelve gates?
Not among the tests published so far. When one does, it gets the same page: the rules, the data, the scorecard and the receipts.
Who runs the tests?
A small independent lab. The coding and the test runs are done by AI research agents working to a written standard; every number on a page comes from a saved output file, and results that look too good are audited by a second agent before they are published. More about how we work.
How do you make money?
Today, we don't: the site and the simulator are free. If we add referral links in the future, they will be labelled on the page and will never change a test result or the order of anything. Details.
Can I suggest a strategy?
Yes. Send the rules, or a link to where they are described, to kaloyan052@gmail.com. We test what many traders ask about first.