Double or Quits?
How long do you keep trying something that doesn't seem to work? How many times can you put up with hearing that "it will be ready tomorrow, hopefully"? When do you 'throw away' what you've done when there's nothing to show for the time you've spent already?
It's a tough call. If you could see the future, then you'd know what to do. Maybe with the benefit of hindsight others will berate you for having taken what was 'obviously' the wrong decision.
Welcome to the gambler's dilemma.
Before you protest that software projects have nothing to do with gambling then consider the obvious similarities. In both cases there is overwhelming pressure not to walk away from your losses. And in both cases there is a small - but diminishing - chance to pay off all your losses by continuing to bet on the same outcome (the spin of a roulette wheel or the activities of a development team). This is why large, late and over budget projects are given more money or time after failing to deliver. If you were the business person who had commissioned a large system wouldn't you be trying to save face by paying 'a little bit more' to get some of it, rather than admitting that your baby was dead in the water?
But there is another way. With an XP project you don't have to play double or quits like that. You know that what you get every couple of weeks is working software with business value. Any time you want to cancel the project you're still left with a positive return for your investment. That sounds more appealing, doesn't it...?