First: what frequency should you hit? There are many, many different factors at play in determining the natural frequency of an object I know from experience. These are (not limited to): Thickness, density, elasticity modulus (you'll need two of those, e.g. Young's Modulus and Poisson Ratio), and of course shape. I'm not aware of any papers publishing a relation between these variables and frequency.
Then, you need to take damping into account; this can again be influenced by many things. For example, the roundness of the glass (even a slightly elliptical shape will produce a 'beating' sound like tubular bells/chimes) is important. Compare with trying to 'whistle' with a soda bottle; as soon as you dent it even a little, the tone disappears immediately. Then, there is the shape, where generally rounder will make for less damping, and also for a stronger glass.
Now, to break a glass, you need to add as much energy as possible, with as little as possible dissipating due to damping, so that the internal 'vibration' energy becomes larger than the binding energy of the weakest bit of glass. Here, we run into a bit of trouble; except for thickness, you can generally say that properties that allow for greater resonance, also make the glass stronger. For example, a glass will break on discontinuities in the glass, but these discontinuities will also continuously dissipate energy long before the glass breaks; energy concentrations will form on 'fancy' shapes in your glass, but these will again allow for less resonance.
So really, the only thing that we can say for sure, and which a 5-year old would also answer, is that thinner is better. For all the other variables, you will need to come up with an approximation. You could do FEM analysis, but I would rather just use experimentation. This will mean breaking a LOT of glasses, and asking a manufacturer for glass samples to test all the other properties of your glass; there are probably papers out there that have this information based on chemical composition and manufacturing process. Then, I guess, you could come up with some factor that captures the shape of the glass in your empirical relations (a bit like the drag coefficient $c_d$ in $\frac{1}{2}c_d\rho v^2$). And then, if you have WAY too much time on your hand, you could try and predict this factor for other shapes. If you do, please publish your results for the world to enjoy :)