When we have a pool of different possible trading strategies, different trading signal combinations, and different trading parameters, typically, we use the backester to build a portfolio of strategies to deploy to live markets in a way that minimizes risk for the entire portfolio. This step relies on having a good backtester, the lack of which causes live trading strategy portfolios to perform poorly and take more risk than historical simulations would have you believe.

Again, since a good backtester is at the core of this step, without it, live trading strategies and portfolios do not perform as expected. When backtesters vary in their deviations from live trading results for different trading strategies and different trading parameters, this problem can be even worse in the sense that not only do strategies that appear to be profitable in simulations not perform well in live markets, but we might also be missing out on strategies that appear to not be as profitable in simulations but actually might perform quite well if deployed to live markets because the backtester is pessimistic to those specific trading strategies or parameters.