“Quadrant theory,” as I believe it is called, has been a useful tool for assessing which cards stay and which cards can go.
Basically, you break down the game into quadrants (e.g., early game, late game, post-boardwipe, waiting-for-wincon, stalemate, etc.) and consider how a given card will perform in various stages of the game.
A card that can be useful in more game quadrants is obviously preferable to a card that will sit in your hand if you draw it at the wrong time. For example, a counterspell can serve as removal when an opponent drops a threat, or it can protect your board while you go for the win.
This approach favors modal spells at the expense of powerful niche spells, so it should be employed judiciously. But it’s a good way of considering to what extent a card can carry its weight.