You've probably seen it: a strategy with a beautiful backtest, a smooth equity curve and a high win rate, that starts losing money almost as soon as it goes live. It happens so often that many traders stop trusting backtests altogether. The backtest isn't useless. It just answers a narrower question than most people think: would these exact rules have made money on this exact past data? Here's why the answer to that question so often fails to carry over.
1. Curve-fitting
Every rule and setting you tune makes the strategy fit the past a little better. Tune enough of them and the strategy stops describing a real market behaviour and starts memorising noise in your data. It looks brilliant on history and has no edge going forward. This is the single most common reason backtests fail. Read how to spot it.
2. Look-ahead bias
A strategy can accidentally use information that wasn't available at the time, such as the close of a bar it's supposedly trading inside, or an indicator calculated with future data. Even one small leak can turn a losing idea into a winning backtest. A good testing engine only hands a strategy each bar after that bar has closed.
3. Unrealistic costs and fills
Commissions, exchange fees, the bid-ask spread and slippage add up quickly, especially for strategies that trade often or target small moves. A backtest that assumes perfect fills at the exact price can hide the fact that costs eat the whole edge.
4. Too few trades
Twenty trades can look amazing by luck alone. With a small sample, a few big winners can make any strategy look skilled. You need enough trades, across enough different market conditions, before the results mean much.
5. The market changed
Markets move through regimes: trending and choppy, calm and volatile, low and high interest rates. A strategy built on one regime can stop working when the market shifts. No strategy lasts forever, which is why testing has to keep going.
6. Testing only on the past
Even an honest backtest with realistic costs is still a test on data you've already seen. The only data a strategy truly hasn't seen is the future. That's what forward testing, also called paper trading, is for: running the exact rules on live data as it arrives, without real money at risk.
- Keep strategies simple, with as few tuned settings as possible.
- Test on data the strategy was never tuned on (out-of-sample).
- Include realistic costs and slippage.
- Judge on risk-adjusted results and enough trades, not win rate.
- Forward test on paper, live, before you risk real capital.
PaperShieldAi is built around that last step. Every strategy first has to pass a historical replay with no look-ahead, then trade live on paper against your best strategy over a full proving period before it can be crowned.
Test your own ideas without risking capital.
Join the PaperShieldAi waitlist. Refer a friend: when they subscribe, you both get 30% off your monthly plan for life.
Join the waitlist