engine online execution: container-isolated data: years of real market history verdict: honest
Backtesting, under oath

Your backtest is beautiful probably lying.

Every strategy looks like a money printer on the data it was born from. Peirama assumes yours is lying and spends the next thirty seconds trying to prove it.

A battery of tests, each built to catch a different way a backtest fools you. If the edge is real, it survives. If it's luck in a costume, you find out here, not with real money.

✓ You're on the list. We'll write when the doors open.
No spam. One email when early access opens.
peirama · run/bollinger_mr.json
Real verdicts on textbook strategies. Years of one-minute market data, fees included. Strategies do pass. They just have to earn it.
The problem

Every backtest tool shows you an equity curve.
Almost none ask if it's real.

Three lies produce most "profitable" strategies, and standard backtesting software checks for none of them.

Overfitting

Try enough parameter combinations and one will look brilliant by luck alone. The more you iterate, the more impressive your best result, and the less it means.

Look-ahead bias

One misplaced index and your strategy quietly trades on information from the future. The equity curve looks incredible. Live trading does not.

Luck mistaken for edge

Markets trend. A coin-flip strategy can print +40R over a good stretch. Without significance testing, you can't tell skill from noise, so most people never try.

The method

It runs the gauntlet. No exceptions.

Every strategy, AI-generated or hand-written, goes through the same sequence. Each stage can reject. Only what survives the whole thing earns a passing verdict.

01 / GENERATE

Describe it

Say what the strategy does in plain English. The engine writes the code against a strict, auditable interface.

02 / SANDBOX

Seal it

Static analysis rejects unsafe or non-conforming code, and execution happens isolated from the network, before a single bar is processed.

03 / GUARD

Blind it to the future

Market data is served through guarded arrays. Reading tomorrow's bar doesn't skew the result. It halts the run.

04 / BACKTEST

Trade it for real

Event-driven simulation with realistic fills, fees and slippage, over years of genuine market data.

05 / INTERROGATE

Try to break it

A battery of independent tests, each attacking the result from a different angle. This is where most strategies quietly die.

06 / VERDICT

Get the truth

A plain answer: what passed, what failed, and whether the edge is real, after counting every attempt you made to find it.

The trial counter is the part you'll hate, and the part you need. Peirama records every variant you test and raises the significance bar accordingly (deflated Sharpe, after Bailey and López de Prado). Run fifty tweaks and "profitable" has to beat the best of fifty coin-flips, not just zero. Your holdout data stays locked until you deliberately spend it. That is the line between evidence and curve-fitting.

The point
Peirama can't be sweet-talked, and it doesn't care how good the equity curve looks. That's the product.

For tradersA strategy that fails the gauntlet in three minutes would have failed in your account over three months. Peirama just charges you less for the lesson.

For learnersThe fastest education in quantitative finance is watching your brilliant idea get interrogated, and understanding exactly which test it failed, and why.

Early access

Find out if it's real before the market does.

Join the waitlist to test your strategies before everyone else, and lock in founding-user pricing.

✓ You're on the list. We'll write when the doors open.