Humans are not very good at remembering event frequency statistics. We tend to only remember the unusual, and forget about the common case. There are plenty of other cognitive errors we make, but this seems to me one of the most troublesome.
An example: Every day I take the subway from Forest Hills in Queens to Manhattan. Either the E or F train will arrive first, and I need the E train. I was noticing, though, that it always seemed to be the F train arriving first, making me wait for an extra few minutes.
Why would that be? Could it be a situation like Feynman's girlfriend puzzle, where the E train closely follows the F? I decided to keep some statistics to find out.
Every day, I'd record which train came first. After a few days, I noticed the E had been coming first lately. Surely a fluke.
I kept at it, and before long the trend was clear. The E comes first almost twice as often! I counted 18 E firsts, and only 10 F firsts.
Obviously, I was completely wrong about even the most basic statistics for my morning commute, and that's about something I experience every single workday morning. Think how wrong I can be about everything else!
I take this as a warning about not trusting experiences when making decisions. Get data instead. Even just some data will be better than opinion or memories.