Two ways to evaluate a strategy. A backtest runs it against historical data — how it would have performed. A forward-test (or paper/live run) applies it to new data as it arrives — how it actually performs going forward. The gap between the two is one of the most important and most ignored concepts in systematic trading.
Why backtests flatter
A backtest is dangerously easy to make look great, because you can — consciously or not — tune the strategy to the exact history you're testing on. That's overfitting: the strategy has essentially memorized the past rather than learned a durable edge. It also tends to ignore real slippage and fills, further flattering the numbers. A gorgeous backtest proves almost nothing on its own.
Why forward results are the only real ones
A forward-test confronts the strategy with data it has never seen and couldn't have been tuned to. If the edge survives there — on new market conditions, with realistic execution — it's more likely to be real. If a stellar backtest falls apart in forward-testing, you just caught an overfit before it cost you live money. The forward result is the truth; the backtest is a hypothesis.
Anyone can build a backtest that would have won. Only a forward-test tells you whether it will.
The disciplined loop
The honest process: form the logic, backtest to reject the obviously broken, then forward-test on unseen data with realistic costs before trusting it, and keep validating as conditions change. This test-forward-validate discipline — with no performance promises, only ongoing verification — is exactly how a serious system should be developed and refined, and it's the mindset behind how NoVo is evaluated over time.