I don't know if you are familiar with the theorycrafting websites that exist around games like World of Warcraft, and which can tell you exactly with detailed mathematics what choices to make in character creation to get the absolutely most powerful character possible. Or the sites for Magic the Gathering or Hearthstone describing the "net decks", that are the most powerful in the currently tournament-legal set of cards. To a lesser degree that also exists for Dungeons & Dragons, see for example this Guide of Guides, linking to guides for every D&D class to optimize a character.

Some of the characters I described yesterday as me playing in my various D&D campaigns deliberately deviate from the optimum. For example I have both a barbarian and a paladin built around maximum dexterity instead of strength. It is pretty obvious that the barbarian loses out on his "rage damage bonus for attacks based on strength" if I don't use strength attacks, and big strength-based weapons deal more damage than small dexterity-based ones. Of course I don't just take something sub-optimal with no advantage at all: The armor class of characters with high dexterity is better in the absence of armor, and so is their initiative. For the barbarian the dex-build actually gives the best possible armor class, and then it opens up a barbarian/rogue multi-class build. For the paladin I took dexterity in order to use two-weapon fighting; having previously observed that the best use for paladin spell slots is often using them for a divine smite after rolling a critical hit, I am trying whether by making more attacks and thus having an increased chance to crit I can make a good build.

But the main reason that I build sub-optimal characters in D&D, or "fun decks" in Magic Duels instead of "net decks", is that optimization inevitably results in having fewer options. There are far more sub-optimal builds for a given class of D&D, or a given deck type in Magic, than there are optimal ones. If we all play just optimal characters in D&D, then you will meet the same build over and over. At one point you are just sick and tired of half-orc barbarians with greataxes, and a halfling barbarian with a scimitar sounds more fun, even if he deals less damage per round. And if you play Magic only with optimal decks, you don't use 80% of your cards, and miss out on a lot of variety. So what if my zombie deck isn't tournament viable? As long as I am having fun playing that deck against an AI deck, I don't care.

Certainly in Dungeons & Dragons, "winning" is not the purpose of the game. And in a game like Magic the Gathering, especially in PvE, "having fun" can also be more important than winning. Having a completely useless character or deck is not fun, but there is a wide variety of slightly sub-optimal builds that can be more entertaining than the ultimate optimum. In D&D that is not just true for yourself, but also for your fellow players. The halfling barbarian and the tabaxi (cat-person) paladin make for more interesting travel companions than their more traditional and optimized versions. And even in combat, optimizing average damage per turn is somewhat boring to the other players, while a character that does slightly less damage but can occasionally shine with maneuvers the optimal build doesn't have can be preferable.

Fortunately at least for D&D this is uncontroversial, and well supported by the game developers. For online card games like Magic and Hearthstone, developers seem to want to push people towards the most competitive game modes, because players need to spend more money to build a competitive deck than they need to build a fun deck. Especially the latest online version of Magic the Gathering, Magic Arena, is going down a path where it only appeals to highly competitive players, and doesn't even offer the opportunity to play a fun deck against an AI opponent. I don't think that this is a sustainable business model for the long term. I'm still playing Magic Duels instead, in spite of there not being any new cards added to it.