You need to be specific when saying things like that and it's often a point of confusion or contention.
- Does string theory give physical predictions of any kind?
Yes, it predicts a slew of physical processes and phenomena.
- Are any of them actually measurable now by our technology?
Yes, the main example being general relativity. String theory develops this within a quantum field theory framework. However, given this isn't a new set of phenonema, it is not seen as 'evidence for string theory' since you can get the same model without all the 'baggage' string theory comes with like extra dimensions etc.
- But why do people say string theory has no tested predictions?
Because they poorly word their description. String theory has tested predictions, what they are referring to is a prediction which no other theory does and which is found to be true by experiment. For instance, relativity superplanted Newtonian theory when we observed light bending around the sun in a way which nothing but relativity could explain. String theory has no such experiment for itself.
Part of the issue is that while string theory covers so much of physics, it covers theories we already have. It models gravity using general relativity, but if you want to model a gravitational system, why bother with string theory when the formalism of GR is simpler? It covers supersymmetry, but why not just use stand alone supersymmetry rather than all the complex string theory methods? It covers cosmological models, but why not just use a version of general relativity for that?
If you want to do a specific model, we already have so many well developed theories for various things, apart from quantum gravity it's pretty hard to find an area of physics that isn't already at least partly modelled by something. But the plus side of string theory is that it's ALL under the same heading, all part of the same formalism. Things like GR or susy or quantum field theories seem to be effective models of string theory (at least that's the general hope), you take a certain limit of the theory and you get quantum mechanics, you take another and you get GR. At present the problem lies in nailing down the details of this. GR is pretty much done, it's getting the QFT sector to play ball.
- But what about other physical predictions?
Unfortunately that's part of the problem with string theory and relates to the 'baggage' I mentioned. You want a model of GR and supersymmetry put together (otherwise known as 'super gravity') then you're going to need to consider 10 or 11 dimensions, reduce 6 or 7 of them to a minute size and then do your calculations. But what shape do the 6 or 7 dimensions take? That's the kind of thing my work is on and it's been the major problem to string theories for about a decade now.... Not to mention, as Mehh says, such things can't be measured directly and at present it's not known if a round-about way of measuring such predictions would ever be found.