This isn't strictly true.
The development team will present a development roadmap, with dated milestones for the project (these could be related to asset creation, for example). The publisher then takes this information and says 'OK, at this point we should be good to show it off and slap a tentative, non-specific release date on it.' Probably build 6 months onto whatever the development team tell them too.
Then one milestone fails and gets pushed back by 2-3 months. This impacts on dependent strands elsewhere on the project, which in turn all get pushed back. When this happens in specialist multi-disciplinary environments where you can't simply redeploy staff, you are then left with cumulative delays stacking up, which pushes your next milestone back. Here's a simple example:
About 4 years agp I was PMing a data migration project between two pieces of clinical software, with ~500K patient records across 4 DBs, and we were merging like-like, i.e. it was 4 versions of the same application being merged into one.
This might sound easy, but it isn't, because you have to pick or create a target application & DB structure that as much as possible matches the configuration of the source DBs.
To fix this, we had an SQL guy who wrote what I can only describe as a piece of magic scripting that would manage the whole ETL process.
He missed his deadline by a week, and that caused a cascade effect into training, clinical management and day-to-day operations, as staff were using an interim shell application to record data.
When the script did run, a misplaced comma caused ~25K of kids vaccination records to not transfer, so our 'crunch' was getting the whole project team of 20 people. plus some temps in for a 4 day data-entry session to upload the missing records from spreadsheets culled from the legacy app.
This, in comparison with game development, was a piece of piss project. The more complex software creation gets, especially in a creative environment where your aim is to make something fun and that might at the project's outset have a pretty nebulous end-state, the harder it is to predict where delays and issues will occur (and this doesn't even begin to include human issues like sickness, family issues, resignations and so on).
It's also why you get things like Unity happening - media space tends to get booked out 6-9 months out if you're talking about holiday season, and the closer you get to campaigns going live, the harder it is to get out without incurring penalties for cancelling. Then you have shareholders etc etc...but that's a whole other part of the pie that we're not discussing here.
TL
R publishers work with information devs give them, and while they make dumb decisions, devs can be equally bad at managing expectations or overpromising on delivery dates.