Gamasutra is full of great articles about making video games. You can learn from them how to make one. This is not one of these.
In this article I want to show you how one does NOT make a video game of his dreams.
1. Do not even START the project unless you have everything ready.
Seriously, don't. Unless you have whole story planned out, down to a single sentence, every bit of art and audio drawn/modeled/recorded and every single map and level drawn, what's the use of sitting up with your engine's editor (however powerful or crude it is) and actually making the game? If you aren't prepared, you won't succeed after all.
It'd be like making a skyscraper without any blueprints at all.
2. At the very first sign of the trouble, cancel the thing.
You're (or person responsible for that part on your team) obviously not qualified enough to do this, because if you were, you wouldn't encounter this roadblock in the first place. So why even try to fix it? It'd take weeks or even months. Which could be probably spent better making cheap knockoff of a popular mobile title. And it would bring more money too!
3. Refresh your team regularly.
In many game development projects, especially long ones, devs and artists burning out is a serious issue. So why not proactively prevent burnout and replace entire team every two months? After all, code is code, art is art and story's text is story's text.
4. Replace your game engine often.
To make the best game possible, you need the best tech, right? So you need regularly check what other engines than the one you're have that yours does not. And if it's particularly nice (for example better foliage or level design toolset), switch over, even if the game is almost finished.
5. Write every single bit of code (that isn't your engine) yourself.
Or possibly even the engine. By using someone else's code you sacrifice possible performance gains or even run possiblity of encountering bugs specific to these extensions. So why not write these by yourself instead? After all, even if you reinvent the proverbial wheel, your code will be more fit to your game, more performant and less buggy as Ubisoft's games have proven so many times.
6. Never, ever use other people's (than your team, past or current - see #3) models, textures, music or other assets.
Sure, you may think it'd make the dev process faster, but who wants to risk their game being called an asset flip?
7. Make a sandbox with seamless online co-op or a MMO as your first project.
Seriously, who play single-player games anymore? No one, I'll tell you. You really should start with a bang, so people will remember you. Even if the result will not be too good, you'll be remembered by the gaming community, and that's what gamedev is all about, right? To leave mark on this world. To not being forgotten. Making sandbox will also teach you many new things, though remember about #2, please.
There are just few of the ways you can NOT make a video game. I'm sure there are more (please leave your picks in the comments along with the info whether or not I can use it in future articles like this).
But to actually make the game, you need to avoid doing any of these. I'm trying to (kinda failing with #5 and #6 though, but only due to the lack of funds) avoid it with my game Computer Virus Simulator which is a 3D platformer where you are a computer virus and which devlog can be found here (though I am more active on my Twitter, where you can get all the newest info tagged with #ComputerVirusSimulator hashtag).