Honest answers about backtesting.

Describe the strategy in plain English — for example "buy BTC when RSI drops below 30, take profit 3%, stop loss 1.5%". Trading Bot House's AI compiles that sentence into exact rules and runs it through an anti-bias backtest engine on up to 3 years of real exchange data, returning the equity curve, every trade, and honesty warnings in seconds. No code, no spreadsheets, free to try.

Almost always one of four reasons: look-ahead bias (the test used information that wasn't available at trade time), overfitting (parameters tuned until past data looks good), survivorship bias (losers missing from the dataset), or ignored costs (fees, slippage, funding). We audited our own backtests and found our quoted 45% success rate was really 6.8% once these were removed — the full forensic is at https://tradingbothouse.com/learn/we-debunked-our-own-backtests.

Look-ahead bias is when a backtest uses information that was not yet available at the moment of the simulated trade — like exiting at a price peak that could only be identified in hindsight, or entering on the same bar a signal formed. It is the single most common reason published backtest results are inflated. An honest engine executes entries on the next bar after a signal and never references future data.

Test the neighbors: if take-profit 4% is profitable but 3% and 5% both lose, you found noise, not an edge — a real edge is a plateau across nearby parameters. Also split data by time: an edge found on 2023–2024 must survive on untouched 2025–2026 data. Trading Bot House runs both checks automatically on every backtest (a robustness sweep and a train/test split) and prints the warnings above the profit number.

We tested 7 of the most-hyped retail strategies — golden cross, death cross, RSI bounce, MACD crossover, Bollinger fade, stochastic, RSI short — on 3 years of real BTC data at sane risk sizing. Not one beat simply holding the asset; three lost money outright. The full results with every trade shown are at https://tradingbothouse.com/reality-check.

Realistic fees and slippage on every trade, next-bar entries (no same-bar fills), a train/test split with out-of-sample results, a parameter robustness check, regime breakdown by year, and a benchmark against buy-and-hold. If a backtest report shows profit before it shows its warnings, be suspicious.

No. A backtest can only prove a strategy would have worked in the past; markets change regimes and edges decay. What honest backtesting does is filter out strategies that never worked at all — which is most of them. Forward paper trading on live data is the next gate before risking money, and Trading Bot House includes it.

Yes. On Trading Bot House a strategy that passes the honest backtest can be paper-traded forward on live exchange data, then deployed to Bybit, Binance, OKX, Bitget or Kraken via API keys, or to MetaTrader 5 for forex and metals via an expert advisor. Paper first is strongly recommended.