Underrated Postgres: Extended Statistics for Better Execution Plans

One of the features of Postgres which is very-little known and even less used (at least I rarely see it in production) is extended statistics to cover and encode the dependencies, relations, potential combinations of multi-column values. The query planner typically creates execution plans based on single-column statistics. But meaning and relationship encoded across columns is very common in real world applications. I personally would've loved to know about it earlier. Usually, I'd add another index in the hope that it improves access patterns and yields better execution plans. But the cost of all the indexes quickly reaches or even extends the origin table. Extended statistics are here to solve the issue. Not always but with high tendency to most often. Not saying indexes are useless but they aren't the best way to solve bad execution plans! The Problem with Single-Column Statistics PostgreSQL's query planner relies heavily on statistics to estimate the cost of different execution pla