I think the analogy holds. Hypotheses with too many “moving parts” can predict anything and so tell you nothing (they overfit the data). Hypotheses with too few moving parts aren’t really hypotheses at all, just passwords like “phlogiston” that fail to explain anything (they underfit the data).
Analogously a mechanism with too many parts takes a lot of effort to get right, and it’s weaknesses are hidden by its complexity. But if someone tried to sell you a car with no moving parts, you might be suspicious that it didn’t work at all.
As Einstein said, things should be as simple as possible, but no simpler.
I think the analogy holds. Hypotheses with too many “moving parts” can predict anything and so tell you nothing (they overfit the data). Hypotheses with too few moving parts aren’t really hypotheses at all, just passwords like “phlogiston” that fail to explain anything (they underfit the data).
Analogously a mechanism with too many parts takes a lot of effort to get right, and it’s weaknesses are hidden by its complexity. But if someone tried to sell you a car with no moving parts, you might be suspicious that it didn’t work at all.
As Einstein said, things should be as simple as possible, but no simpler.