First, let me just say I think that it's days too early to talk about Sandy coming to Connecticut. If this storm were to impact the state, it would not happen until early next week...7 to 10 days from now! Our forecasting accuracy decreases dramatically that far out. That being said, I understand why Connecticut is on edge about this. Last year's Irene and October nor'easter freaked us out. No one wants to go another week without power.
The storm is stationary, spinning in the Caribbean 400 miles SW of Jamaica with winds of 40 mph. Sandy will move north into the Bahamas. This is one thing the models can agree on. Further than that, the models are all over the place. Just look at all the different solutions from the GFS ensemble members below!
The European Model, among the most accurate with Irene, predicts landfall on Cape Cod next Wednesday evening. This would be devastating for the northeast with gusts to 60-80 mph, 4-6" of rain and storm surge. There is a sizable blocking pattern to the northeast which could act to block Sandy from moving out to sea. (Don't stop reading here. Because right now this situation is highly unlikely)
The GFS model (a trusted American model) has the storm curve safely out to sea with a weaker blocking pattern.
Researchers at University of Wisconsin at Milwaukee have plotted the ten historical hurricanes with characteristics closest to Sandy. Climatology is a major factor in hurricanes, so this is a reasonable tool. None came close to Connecticut.
So which model is right? Is it a direct hit, a miss, or something in between? The answer is not "which model is the best". What we need is... time. I know, I know. That's a boring answer. This storm is only in the beginning stages of development and we still need to watch how it responds to land interaction with Cuba and the Bahamas. This is also a slow moving storm. That's important to note. The slower a tropical system moves, the poorer our ability to accurately forecast it. Everything would need to line up or phase perfectly for the European solution to verify.
It's not impossible, but at this moment it's unlikely. Even tiny errors in the computer models will multiply over time... and 7-10 days is a long time.
So let me do the worrying for now. I'll keep you posted!