It depends on how you configure your intrusion detection system and the particular polymorphic code.
For example, if you filter all non verified applications that execute commands to open sockets or other critical system operation then it does not matter if the code is polymorphic since you are not focusing on the details of the code.
As another alternative answer to you question it depends on the polymorphism executed on the code. For example, there will always be code needed to decode the rest of the morphed code. These sections can be targetted. In addition some polymorphic code may lie in the same offsets in the code.
In the end, it depends on the particular threat since, a designer of malware always has the advantage of choosing the polymorphic techniques that he wants, and once it is found that you are filtering based on a specific technique that targets a particular assumption of the attack that cannot be hidden, then attacker may choose to try a variant of the attack.
One should remember that no "lock" is absolutely unbreakable and there always is a chance that a new sophisticated attack goes undetected ("under the radar ").
Sir , An attacker is like a Ghost , we cannot predict which random segment of code is used in Polymorphic Engine to generate the code. So it still unreachable to decode the code and predict or detect the attack.....