A million to be succesfull, but for bigger games im looking for at least 2 million.

Profitability.

Merely being profitable is never a good indicator of whether a game was successful or not.

you're getting the term "successful" confused with "worth playing"

It doesn't have to be a game that is worth playing to be a successful game, it just has to meet the original purpose of making the game, which is to cover all the costs of making it, and hopefully provide enough extra to fund making another game after that.

1 mil

Unfortunately, I was not confusing those two concepts. In fact I neither mentioned nor implied quality as a consideration. I was talking about the maximisation of the return on investment - a more important, and more complex, concept than profit and a more relevant concept when talking about how successful something is.

For example, a game could make a nice profit but not be successful simply because the resources used to make that game could've been applied to make something else that obtained a higher profit. You will perhaps (or perhaps not) note that this is actually the original purpose of making the game - that is to obtain the highest return on investment to maximise the value of shareholders.

You cant put a number figure on it - if Halo 3 had sold 1 million copies it would have been deemed a huge failure.

For new IP, I would say 1 million
For old IP, More than the previous game

1 million imo.

As Picko points out, the OP's definition of success means no PS2 game has every been "successful." Heck, darn near none have. To make matters worse, that definition makes it easier to be successful on a console that barely sold than on one that sold gangbusters...

For me, the line is drawn wherever the title begins to see a high return on investment. So a DS game selling 100,000 copies usually equals great success; A Wii game selling the same amount may be a success, but probably isn't; an HD game with those sales is a flop.