i think donelines are more important than deadlines -- milestones w/ a list of things that have to be done for that milestone, but no set date, because it's usually impossible to predict how long something will take to finish.
It's not impossible, it just takes some practice to get good at it.
When setting deadlines, you want your incremental milestones to be about six weeks apart, and to have lots of different things lumped together in each. That way, if one task takes longer than you planned, then it's likely that one of the others will take less time and your milestone workload will balance out overall. Remember that you want to include slack time and debugging time in your considerations as well. The larger the team and the more complicated the project, the more slack time and debugging time per person you want to allow for. My general rule of thumb when you're new to this sort of estimating is to take your best guess about how long a task will take to complete, multiply that guess by two, and then add one. (And that final number will probably still be a bit low, but it'll at least be within spitting distance)
But yeah, the first time you do this for a project your numbers are probably going to be way out. But you'll get better at it as you do it more often, and learn your own and your team-mates' work habits.
And do remember that there's no problem with overestimating how long it'll take to complete something; if you say that it's going to take you a week, and it ends up taking you two days, then you're ahead of schedule and can start on the next milestone's stuff early, or can work on polishing or take a well deserved break or do whatever else strikes your fancy. But it can be really useful to know
early when you're falling behind from where you need to be in order to finish your game on the date you wanted, so that you can decide whether to move your finish date or to change the project requirements in order to try to catch up. It's far better than just arriving at that finish date and wondering how you missed it by so much.
But again, I don't recommend scheduling milestones for projects that are going to take less than about six months; they're not worth the time investment for short-ish projects like those. But they're totally essential for long projects, and projects with lots of team members.