Alright, time to crush this attach rate argument here...
The obvious problem with comparing GT3's attach rates (now) with Halo 3's attach rate, is the massive amount of PS2's sold after GT3 was no longer being bought. To solve this, I'll show you specific dates in which GT3 had already sold most of its copies (with not too many PS2's sold):
Week ending August 19th, 2001: Japan's GT3 sales at about 1.7mil with 5.5mil PS3's sold, or a 31% attach rate. That's close enough to the current userbase of the PS3, so if GT5 is a hit / still popular there, we can expect at least 1.5mil from that region.
End of March 2003 (see page 31): Over 11 million copies of GT3 were shipped by the end of March 03. Since under 2 million were shipped to Japan, we know that at least 9 million were shipped outside Japan, so... That's over 9 million GT3 copies shipped to ~35 million PS2 consoles sold (VGC), likely close to a 25% attach rate in the west.
Attach rate really doesn't tell us much, since there are too many outside factors (like PS2's extremely fast rate of consoles sold), I don't think it's worth putting that much weight on.
As for other factors:
While people have mentioned Halo's popularity has constantly been increasing, the same can be said for Gran Turismo. Although the second title per console always performs worse, the first of the console (GT1, GT3, soon GT5) performs well. The exception here is European sales, which increased with every release besides GT2. Expect GT5's Europe to be nearly on par with Halo in America.
The increased competition (such as Forza 3) may mean less GT sales, but the same can be said for Halo. CoD is surely the bigger franchise out of the two, yet people think Halo will still do fine. I think the same will probably happen for GT5.
Bundles. I have no idea how many American copies of GT3 were sold due to bundles, so this could hurt the past sales' arguments for GT5 by quite a bit.
As for my prediction? I see GT5 selling at least 1.5mil in Japan, 3 million in America, and 5.5 million in Europe, which would clear 10. I just don't know how far it can go after that, though.
