Hm... this isn't a dig at you or anything, just a thought I'm having trouble answering for myself right now, but how low does one get to set the probability of a prediction, and still call it a prediction? At which point - seems it's not 50% - should you rather say "I'm predicting this won't happen?"

So, Voldemort has two approaches to the "prevent Harry from destroying the universe" plan. First is killing Harry, second is turning Harry away from destructive paths. He might reasonably be uncertain of how the first plan will work. Harry is Tom Riddle, and that means killing Harry might just put him in the Horcrux system, and now Good Tom and Bad Tom are having a psychic battle instead of a physical or a magical one. Given that Bad Tom has the upper hand in physical and magical battles, being reluctant to fight a psychic battle instead seems appropriately cautious. The second plan involves niceness, which Voldemort is not very practiced at. It does have the upside of having another sane individual to play chess with, and with Hermione keeping Harry from doing anything too crazy the stars are safe (indeed, safer, because now there's a second Riddle working to improve things). But, more importantly, because Voldemort hasn't tried this one much before, he doesn't know how and when to expect it to go wrong. So he makes rookie mistakes.