How important is backtesting when choosing a trading bot?

One thing I keep seeing when researching trading bots is backtesting. I understand the basic idea of testing a strategy against historical market data, but I’m not sure how much importance I should actually give it when comparing different bots. A strategy can look great on historical data and still behave very differently when market conditions change, so I’m wondering what other factors should be considered alongside backtesting. I’m particularly interested in things like fees, supported exchanges, risk controls and whether the platform allows you to test a strategy without immediately using real funds. Is there a resource that compares trading bots while taking these kinds of details into account?