If there are 5 concepts in a task, then the null likelihood is 1/5 = 0.20. Using just random utilities, we should be able to predict at a rate of about 20% which concept the respondent chose.
It's hard to know for sure if a person is gaming a CBC survey. If the respondent always picks the same brand every time...or the lowest price every time (you'd figure this out by analyzing the raw data file, either the .cho or the .csv file), it's pretty clear that such a respondent uses a simple rule to answer the CBC tasks. Now, whether that simple rule is an invented rule to get through your survey quickly, or whether this is a rule this person would actually employ in real life, we don't know.