Start with live evidence
A profitable backtest shows that a model worked on a historical dataset under its assumptions. A sufficiently long live record is stronger evidence because real orders face spreads, slippage, latency and changing liquidity.
Return without risk is incomplete
A strategy earning 20% with extreme drawdowns is fundamentally different from one earning the same amount with modest drawdown. Profitability should therefore be evaluated together with the path taken to achieve it.
Fees change the answer
Subscriptions, spreads, commissions, financing and performance fees can turn an attractive gross result into a much smaller net return. Compare what the user keeps, not only what the strategy reports.
Market regimes change
A model can perform well in trending markets and struggle in ranges, or vice versa. A longer history covering varied conditions is more informative than a short period with unusually favourable conditions.
Can profitability continue?
No historical record can answer that with certainty. The practical goal is to identify systems with credible evidence, understandable risk and enough transparency to keep monitoring whether the edge persists.