Get a Real Backtest.
No Bullsh*t Results.

Plain English in. Real candles, an honest verdict and a TradingView strategy out.

Market sentiment today

Loading

Loading house tests

How it works

Four steps, no black box

1 · Describe it

Plain English. Parsed into a strict, hashed rule set.

2 · Replay it

Your exact rules, real spot candles, next-open fills.

3 · Judge it once

Tuning on training data only. Verdict on unseen candles.

4 · Read the receipts

Win rate, drawdown, robustness and a TradingView strategy.

Train / unseen split

Search here
Judge here

The search only sees the left side. The verdict is measured once on the right.

Why most backtests lie

Six flattering shortcuts we refuse

Fills at the signal candle's close

Fills at the next candle's open

Stops filled through gaps

Gap opens fill at the worse, real price

Fees on a fixed notional

Costs on the actual traded notional

Indicators warmed up inside the window

Warm-up candles fetched before it

Tuned and reported on the same data

Tuned on train, judged on unseen

A different window every run

UTC-snapped windows, printed on every row

Plans

Your first report is a Pro report

New accounts get one complete Pro report, every panel unlocked.

Free

$0

  • 5 backtests in total
  • Overview with the real verdict
  • Up to 4 timeframes per run
  • Advanced Analysis + export locked

Pro

Pricing announced at launch

  • Full Advanced Analysis: robustness, regimes, Monte Carlo
  • TradingView Pine v6 export of the executed rules
  • Unlimited improvement searches and forecasts
  • Up to 14 timeframes, one identical window
Compare plans

FAQ

Straight answers

Where does the market data come from?

Public spot market candles from major exchange REST endpoints. Every report names the exchange, the symbol, the candle count and the exact window it used, and incomplete candles are excluded.

Do you invent rules I did not write?

No. Backtest mode tests exactly what you describe. If you give no stop or target, the report says none was given and shows how the position was sized instead — nothing is filled in for you.

How can the improvement search be trusted?

It only searches the training window. The winner is frozen and then measured once on candles the search never saw, and both numbers are shown side by side.

Will the same description give the same result twice?

Yes. Parsing is cached by a normalised prompt hash and windows snap to completed UTC days, so a repeated run on the same day reproduces the same rules, numbers and text.

What do I actually get to export?

A TradingView Pine v6 strategy generated from the executed rule set, with the same entries, exits and costs used in the report.

Is this financial advice?

No. Backtestify is a research tool. Results are historical simulations, they are hypothetical, and past performance never guarantees future results.