Yeah, some of it is based on statistics, although in my experience, the relative distributions of rows don't massively change in my databases, so an ostensibly optimal query plan on day one isn't likely to be rubbish later. You can obviously imagine other people's mileage varying.