Probability and baselines before models

A probability is a count divided by a total. A baseline is the result you can get before the fancy model enters the room.

If 75 percent of the examples are positive, a model that always predicts positive is already 75 percent accurate. A model score only becomes interesting after you compare it with that baseline and the cost of being wrong.

The builder move is not "the model got a big number." It is "the model beat the obvious baseline by enough to matter."