The short answer is that alanines don't make a very stable core.

Alanine is indeed very common in natural proteins, and does favor α-helical structure. And, by default, an alanine-rich α-helix scores really well in Foldit.

However, since alanine has such a small sidechain, an alanine-rich helix has a "featureless" shape. You can imagine the sides of the helix would be relatively smooth (whereas a helix with larger sidechains would present an irregular, studded surface). If you were to design several of these helices into a bundle with an alanine-rich core, you would get a protein that forms helices in solution—and would probably even form a bundle to bury the alanines. But, the smooth sides of those alanine-rich helices would allow them to pack against each other in many different ways, effectively forming many different bundles. So instead of all the protein copies in a sample forming the same structure, you would get a wide mix of helical bundles with slightly different packing, constantly changing shape as the helices slide against one another. We say this protein exhibits molten globule behavior.

Instead, we prefer designs with bulky, interdigitated sidechains in the core. With only one way to efficiently pack those sidechains together, all the protein copies are expected to adopt a single, well-defined structure.