Same bot. Same formula. DSR went from 0.00 to 0.999 — just by feeding it the right data with the right trial count.
The distinction matters: backtested Sharpe with parameter optimization will almost always fail DSR. That doesn't mean the edge isn't real.