Here are two new books on evolution. I wrote chapters for both :o
Predicting Evolution and Visualizing High-Dimensional Fitness Landscapes
Effects of Epistasis and Pleiotropy on Fitness Landscapes
You can download the chapters via those links at least for a while. After that they can always be downloaded from the arXiv.
Both chapters are about fitness landscapes. The first one explains that fitness landscapes is the third "parameter" that is needed to make predictions in evolution. The first two are the population size and the mutation rate. If all the known effects of mutations are known (which implies the fitness landscape), then it is possible to statistically say which direction an evolving population will go and where it will end up. Stochasticity is still important, but one can still make sensible predictions.
Admittedly, both population size and mutation rates can change over time, and so can (and does) the fitness landscape. But we have to start somewhere if we want evolutionary theory to make predictions about the future. Simplify the problem to one that can be solved, rather than including every factor that contributes to the real-world problem.
In the second chapter I argue that ruggedness of fitness landscapes (think of many hills and valleys, rather than a single peak to ascend) is created solely by epistatic interactions between mutations or genes or traits. No other mechanism creates ruggedness, and epistasis always creates ruggedness.
For more about fitness landscapes and epistasis, see
Smooth and rugged fitness landscapes
Can we predict evolution?
Epistasis in evolution
Crossing valleys in fitness landscapes