Guardian is not a "delayed copper". It borrows a $ from this turn to put it into the next turn.
When you talk about the hand-gained Night cards you have to differentiate between what they do when you gain then and what they do when you play them after the first time.
In Kingdoms without attacks Guardian is an on-play delayed Copper and an on-gain extra Coin next turn. The former is pretty bad (and even 'prefer extremes to averages' arguments can rarely rationalize the existence of a delayed Copper in your deck; with a delayed Peddler like Caravan Guard that's a different ballgame) whereas the latter is pretty good.
A "delayed" copper, like any other delayed card would have that affect again and again. This is how we talk about Caravan being a delayed Lab or Wharf being a Moat now and a delayed Smithy.
On-buy you get a bonus coin. End of story. It has no interactions of the copper type (e.g. Gm) or even treasure types (e.g. Poorhouse). It does not enter your hand taking up space like a copper (e.g. Lib) or get played at the same time as a copper (e.g. Gm).
Quite literally Guardian takes up draw slot in your hand that all but always reduces your current buying power by at least $1 for much of the game. That makes it worse in many senses than a delayed copper or even a delayed coin (e.g. Caravan guard)
Guardian is not a "delayed copper". It borrows a $ from this turn to put it into the next turn.
These two things are not mutually exclusive.
For a lot of Dominion, you want more $3/5 odds and less $4/4 odds.
I still find this argument utterly unconvincing. Sure you usually prefer 3/5 to 4/4. But do you prefer 4/6 to 5/5? Do you prefer 7/9 to 8/8?
"A good hand and a bad hand are better than two average hands" is only true when your average hand is bad. And I think that in most cases, your average hand isn't bad enough to make this axiom valid.
EDIT: Put another way, I would rather have two good turns than a great turn and a mediocre turn. Mediocre turns can kill.
Funny it is almost as though I said, "for
a lot of Dominion" swinging towards the extremes is better. I.e. basically every first shuffle in the game. Sure we can all come up with plenty of scenarios were we would prefer the constancy of regular cash and they would not even be edge cases ... but then you just buy a second Guardian or a copper and forget about it. When it matters most, on average you prefer the $3/5 over the $4/4. By the time your average hand is $5, you typically are drawing deck or have compensated regardless.
This is not to say that Guardian is good, only that its one-off bonus next turn and its push away from the mean are typically helpful enough to make it better than nothing (or another copper) if you open $5/2 or have $7/2 buys early enough in the game.
People here always seem to get caught up in "is this card good" rather than looking at the
marginal utility of a card. Would you generally take it for free with a good card (e.g. mandatory Haggler gain), would you take it for "free" (e.g. $2 with nothing else to buy at $2), would you take it over silver, would you take it over a bog standard card at $4 (e.g. Smithy or Cutpurse), would you take it over the average $5 (e.g. Chapel type power)? Each of those is a worthy question that requires some thought.
Me I place Guardian-without-attacks somewhere between I would take it off a Silver/Haggler (a board that is really bad for it) and I would take it over Silver on an exceedingly rare number of boards (e.g. ones particularly bad for silvers, like Bandit camp) and that covers ~95% of boards.
Just knowing that a card is weak means little. Knowing why it is weak and what makes it stronger, that helps you actually analyze boards optimally.