Moore, Open Questions and ...is good. There are a couple features of good that are analyzable:
First, goodness is gradable. Hence the comparative and superlative forms, better and best, and extreme forms like great and excellent, which imply good but not vice-versa.
Second, goodness might have a scalar structure with a maximal endpoint. Plausibly, this is denoted by words like perfect (i.e., 'that which can't be better'). Of course that doesn't tell us for any particular thing whether it can be perfect, just that goodness in principle can admit of such endpoints.
Third, goodness is apparently not relativized to anyone in its ordinary uses. So when one person says 'this is good,' and another says 'this is not good,' they can contradict each other, be reported as disagreeing, etc. This is hard to explain if good means good for x and in most such disagreements the value of x differs across the claimants.
Fourth, goodness can nonetheless be overtly relativized, as in good for him (with something that is good for him perhaps not being good for me).
Fifth, goodness apparently does not track personal preferences. So there is no contradiction in claiming that something is good, even though one isn't pleased by it, doesn't like it, etc. Claiming that something is good often implies that one approves of it, etc. but apparently this is because we approve of things that are good, not because things are good in virtue of our approving of them.
Sixth, goodness, whether it can be reduced to any natural property or not, apparently must supervene on such properties. Thus, it is contradictory to take two situations totally identical in their descriptive or natural qualities, and claim that one is good while the other is not. Goodness cannot be a free-floating quality that the exact same descriptive situation can have or fail to have: rather, things must be good in virtue of those qualities.