Binance — and every other exchange — sells grid bots ranked by "Best PnL" and shows you a 7-day drawdown. That's marketing: you see the winners (survivorship bias) and the real risk is hidden.

So I backtested them honestly: 4 years of real price data, across crash / bull / sideways markets, net of fees.

What I found

A grid bot is volatility harvesting, not alpha. It profits when price oscillates in a range, but:

• In a crash, price leaves the range and you ride the inventory all the way down — real drawdowns of 30–58%, not the 1–2% of "7 days".
• In a bull, the grid sells early and misses almost the entire move (+7% vs +127%).
• The "+0.59% in 20 days" they show you is a cherry-picked ranging window.

Even giving the grid its best shot (modeling trailing up + a range sized to each coin's volatility), the large caps (BTC, BNB) are only the "least bad". Memecoins are feast-or-famine: +25% in a good range, −56% in a bad year. None beats buy-and-hold.

How it works (the honest method)

• Real hourly candles from each exchange's public API (free, no key).
• A conservative grid simulator: it models buys/sells at every grid level, including the inventory you pile up when price leaves the range — the loss the marketing hides.
• Net of fees, across fixed regime windows (crash / bull / sideways / recent).
• Python + FastAPI backend; the simulation core is ~150 lines.

Try it — free, no signup

Pick an exchange (Binance, Bybit, OKX, KuCoin, Bitget, Gate.io, MEXC), paste any grid bot's pair, and it shows you the real drawdown + Grid% vs buy-and-hold per regime — and links you back to the bot so you compare "what they sell you" vs "the honest number".

👉 soberquant.com/grid-validator

I don't sell the bots. The whole project is honest, survivorship-free validation of trading strategies — I publish what fails, not just what works.

Paste your own bot and tell me what you get.