PostgreSQL exploits a statistical approach to evaluate different execution plans. This means that PostgreSQL does not know how many tuples there are in a table, but has a good approximation that allows the planner to compute the cost of the execution plan.
Statistics are not only related to the quantity (how many tuples) but also to the quality of the underlying data – for example, how many distinct values, which values are more frequent in a column, and so on. Thanks to the combination of all of this data, PostgreSQL is able to make a good decision.
There are times, however, when the quality of the statistical data is not good enough for PostgreSQL to choose the best plan, a problem commonly known as "out-of-date statistics." In fact, statistics are not updated in real time; rather, PostgreSQL keeps track of what is ongoing in every table in every database and summarizes the number of new tuples, updated ones, and deleted ones, as...
