This is a personal viewpoint. It is not investment advice, not a recommendation, and not a report of results.
A backtest tells me how a rule would have behaved under the assumptions I coded. Fills, costs, the period, the universe, the hours I decided to include. Change any of those and the story can change. That is useful. It is a way to see my own assumptions. It is not a promise about the next year.
What a historical test cannot tell me is whether I will follow the rule when it hurts, whether the market I tested is the market I will trade, or whether a result I tuned until it looked kind will survive data I did not use to design it. A test that never leaves the sample I learned on is not confirmation. It is a memory of the design process.
I want the ugly cases in the test: gaps, costs, a stretch where the idea does nothing, a stretch where losses cluster. I also want a check on data I did not use to invent the rule. I have not published those results. Own Research is where I work on analytics and simulation tools. Treasure Capital (opens in a new tab) is where I am building the trading systems and the algorithmic research. Neither sentence is a performance record.
A backtest can reject a bad idea. It cannot, by itself, crown a good one.