Pwn2own doesn't work the way you think it does. The participants use prepared exploits. You can't infer anything about the relative security of different systems that both get exploited there.
Zero day hackers and the malicious Chinese hacker spies also use prepared exploits, so you can infer something about the relative easiness of finding exploitable holes.
Pwn2own may use prepared exploits, but researchers tend to go for the easy low hanging fruit, so there is a lot to infer from who falls on the first day etc.
Windows 7 is many orders more secure than OS X—just look at Pwn2Own, OS X is regularly the first to be eliminated.