Confidence, that’s half the job. To be able to go into the last week of a project and be confident in your product is half of what you need to get the game done.
At this point in the process you should have everything programmed, all the art in the game, all the audio final, and should only have to polish. The only reason you should not be confident in your game is if you haven’t delivered on what you set out to get done, if the game isn’t what it’s supposed to be. By this point, confidence will say everything you could say about your own game.
Being able to answer the question “how do you feel about the game at this point?” with a straight face and a “great, we’re nearly finished we just have to tweak some things” then you’re golden.
That being said… that’s still only half of the job of finalizing the game. You still have lots of work to do. Publicity and polish/bug testing should be the priorities at this point. If you’re still working on getting a mechanic into the game, scrap it. Find a work around for why it’s not in the game and just move on, because it won’t get in and if it does then it will bring the game down. Time has to be spent on testing, bug fixing, and tweaking the variables to make the best experience out of what you already have. Nothing should change.
The best way to really describe it is ‘minimal risk assessment.’ If your next task has a possibly poor outcome that you can see ahead of you, don’t do that task. Find another way around finishing that part of the game or closing it off.
A good example I have of this was making a level in the Unreal Development Kit (UDK) where the player boards down a mountainside to get to the base. I made, I kid you not, an entire mountain-side for the player to go down. I realized afterwards that it was causing me more troubles with bugs than it was helping with design, so I had to scrap half the mountain-side so that I could polish the critical path down the mountain and keep it bug-free (relatively.)
Finalizing a video game is the game – sometimes you have to cut out half the mountain.