Strategy builder for MetaTrader 5
Build and validate your strategy.
A genetic engine searches thousands of rule combinations on your markets. An anti-luck shield then tells you which strategies chance cannot explain. Often the honest answer is none, and the program says so.
The anti-luck shield
A beautiful equity curve is not a strategy.
With thousands of attempts, chance produces beautiful curves too. Every builder can find them. NR StrategyLab runs each real search next to null searches on data that holds no information, and judges every portfolio on years the search never saw.
What a builder usually shows you
- The best curve out of everything it tried
- Statistics on the same years it was fitted to
- An out-of-sample period you can keep re-running until it looks good
- No way to know how much of it luck would have given
What NR StrategyLab adds
- The same search on data with no edge, so you see what chance finds with the same effort
- A percentile against 2,000 time-shifted copies of the portfolio's own trades
- A test on the years before the training, the hardest one to fake
- A sealed final exam and a counter of every attempt you made
Null searches
Each search runs again on results shifted in time. Whatever the null searches find is the bar luck sets. A real portfolio has to clear it.
Placebo percentile
The portfolio's trades are moved to random weeks 2,000 times. Its real result has to beat at least 95% of them in each test period.
Test before training
Training on 2016–21 and checking 2006–15 as well: strategies that only work going forward in their own years fail here.
Sealed final exam
The last months stay hidden from everything until you open them once, for the one portfolio you trust. Every exam is numbered.
Attempt counter
Judging the same markets and years again counts as another attempt. The program shows the stricter bar that many attempts would require.
Robustness
Double costs, entering one bar late, nudging every parameter, removing the best trades. Solid is not the same as good, so both are shown.
How it works
From your data to your MT5, in five steps.
Data
1-minute bars from your own MT5 with your broker's real spread, from Dukascopy, or imported. A quality check flags bad ticks and zero spreads.
Search
Choose markets, bar size from 1 minute to weekly, exits and up to 75 indicators. The genetic engine combines and evolves the rules.
Shield
Real and null searches, portfolios judged on unseen years against placebo. Only portfolios that pass every test become candidates.
Check
A sheet in plain words and money: what each strategy does, out of every 100 trades how many win, year by year, worst drawdown.
Deploy
Compare it with what you already trade, check the EA against the engine in the MT5 tester, and deploy it to MT5 without clicks. Demo first.
Features
Everything between an idea and a live account.
Judge
Bring the MT5 tester report of any EA, from any builder, and find out whether it is luck or edge.
Prop-firm aware risk
Challenge rules with daily and total loss limits. A Monte Carlo of your portfolio gives the risk per trade with the best chance to pass.
Own capital
Set the drawdown you accept and the program sizes the risk to grow the account within it, in 90 or 95 of every 100 simulations.
Portfolios
Does your account improve if you add it? The candidate on top of what you already trade, compared at equal risk.
Test bench
Each strategy runs in your MT5 tester with the NR_Pieza EA and is compared with the engine, trade by trade.
Deploy without clicks
One chart per strategy with the EA and its parameters, written to your MT5. Trading stays off until you switch it on.
Queue and alerts
Line up searches for the night. The queue survives a restart and you get one alert if something comes out.
Optional assistant
With your own Claude API key, an assistant runs the program for you and asks before changing anything.
Made for
- MT5 traders who build or buy EAs and want to know if a result is luck
- Traders preparing a prop-firm challenge
- Anyone tired of backtests that collapse on a live account
Not made for
- Signals or copy trading: it gives you none
- Managing other people's money
- Anyone looking for guaranteed returns: there are none
Closed beta
We are opening it to a small group first.
Between 20 and 50 traders, at a reduced price while the beta lasts. We want to learn how many of you find a candidate and how many keep using it.
Questions
Before you ask.
Will it make me money?
Nobody can promise that. It is analysis software: it searches for strategies and tells you which ones beat chance on past data. Beating chance in the past does not guarantee future results, and trading carries a high risk of losing money.
Does it actually find strategies?
Yes, it finds plenty. The point is the filter: most of what looks good does not survive the null searches and the unseen years. When nothing survives, you learn that before risking money.
Is it a signal service?
No. You choose the markets, the indicators, the risk and whether to deploy anything. The program never trades on its own: deployed EAs start with trading switched off.
Can I use it with my prop firm?
It simulates challenge rules such as daily and total loss limits to suggest a risk per trade. Check your firm's own rules on EAs before you trade.
Can I judge an EA I already have?
Yes. The Judge takes the MT5 tester report of any strategy, or a CSV of trades, and compares it with chance in the same way.
What do I need?
A Windows 10 or 11 PC (64-bit) and MetaTrader 5 with your broker. The data is downloaded from inside the program.