Your above analogy is so vague, I'm not even sure what you mean by it.
Evolution can and does make predictions, and I'll show you one example how.
Below is a phylogenetic tree. It depicts branching events of various vertebrate groups from their common ancestors, with time plotted on the Y-axis, so you can see when the groups are supposed to have diverged.
The tree itself is an hypothesis, based on a comparison of morphology/genetics. It predicts that certain groups diverged from a common ancestor at a particular time. In the above example, the tree predicts that mammals diverged from the reptile stock sometime during the Permian. And regardless of how you might feel about it, juvenissun, this
is a prediction -- and a testable one at that. All we have to do is go out into the field in search of mammal-like reptiles from rocks of Permian age and see if our prediction holds up (and lo and behold, it does).
The same is the case of
Tiktaalik, as explained in the link I gave you above.