Backtesting is to determine how a particular trading strategy has performed historically.
As long as a trading strategy can be quantified with logic and reason and historic data is available the strategy can be backtested. Backtesting is important for fully automatic trading systems.
Backtesting can help discard investment strategies that show little promise. A strategy that has worked poorly in the past will probably continue to perform poorly.
Using backtesting solely to pick a strategy can be problematic. Backtesting is open to problems such as data fitting. When testing a number of strategies some will show really good results.
It does not give any guarantee that the results can be obtained going forward. Almost always the results will be worse to some degree going forward than the backtest has show was the case historically.
Backtesting is a form of historic paper trading. It can also be necessary to perform forward testing - trading using real data and real trading on a forward going basis to evaluate a trading strategy.