Does the back test include data that encompasses known market conditions? There will always be new market conditions. However, known market conditions should be considered and included in back testing. The quintessential example is ensuring an equities strategy performs in bull markets as well as bear markets.
Complex strategies inherently lend themselves to curve fitting. The more indicators used to enter a trade and the more criteria used to exit a trade will produce the perfect strategy for historical data. You are essentially creating a strategy with 20/20 hindsight.
Does the back testing produce consistent results. If a large percent of the equity is produced in a small percent of the time, this is symptomatic of a curve fitted system.
Its too good to be true. A system that produces perfect back testing results is likely curve fitted or exploiting the data without taking into consideration the cost to take a trade. An intraday system back tested with a small profit target will not achieve the same results in a live market. In back testing your strategy will always get a fill whereas in a live market there is no guarantee of getting filled.