In contemporary financial quantitative analysis systems, accuracy usually means lower risks and higher profits, and latency usually implies more uncertainty. A more accurate implied volatility calculation does not promise better performance in the real market. It is necessary to consider the cost of accurate implied volatility computation and sophisticated models’ inference time consumption, especially the high-frequency trading and risk management tasks, which were unfortunately omitted by previous research. In this research, we formulate the trade-off between accuracy and latency as a hyperparameter optimization problem and apply the evolutionary algorithm to improve the performance of the simulated trading system. The proposed optimization implements better balances between time consumption and accuracy, with portfolio assessment criteria as the optimization target. The conclusion and proposed method are applicable in future high-performance financial engineering industrial application development.