Choose a topic and research it. A good hypothesis emerges from a broader topic. For example, in the social sciences a topic might be: wars in Europe since the nineteenth century. In the natural sciences, a topic could be: the effects of chocolate. These broader topics will give you a chance to study an area and develop a good hypothesis from the broader data set.
Pay attention. While researching your topic, pay attention to common trends and to unexpected outcomes. Recurring outcomes may suggest a pattern --- for instance, eating chocolate and gaining weight --- and unexpected outcomes may point towards an unusual but important effect of the phenomenon, or else a heretofore unknown property of a substance. For instance, while it is commonly believed that chocolate leads to weight gain, in some instances a moderate amount of dark chocolate can assist in weight loss.
Formulate your hypothesis. Your hypothesis should take the form of a statement, such as A may cause B because of C, D, and E. You can further add nuance to your hypothesis --- and complexity --- by adding conditioning variables. A conditioning variable is something that must occur in order for your hypothesis to hold. For example: If T and S, then A because of C, D, and E. For a real life example: If it rains outside for a long time, then the street will flood when the storm drains are clogged.
Test your hypothesis. Once you have formulated your hypothesis, ensure it is testable. A good hypothesis can be falsified --- that is, it can be shown to be false. A hypothesis such as "God created the world" is not a good one, since it cannot be shown to be true or false. Such a hypothesis may be a very real and important belief, but it doesn't make a good proposition for the social or natural sciences. A good hypothesis of the testable sort might be: a failure to water your plant will cause it to die.