1

This reminds me of an issue the original Final Fantasy had.
* For each spell level, there are 4 white magic spells and 4 black magic spells. (There's one class that can use both types, as well as some that only get one type.)
* A character can only learn 3 spells per spell level *ever*. Once a spell is learned, it can't be removed, and you can never get that character the spell you didn't learn before.
* Some of the spells are broken or useless. In particular, LOCK, TMPR, and SABR don't work at all, LOK2 *increases* enemy evasion, and XFER doesn't work when you cast it on enemies (it works properly when enemies use it against you).

So, in other words, you have to make permanent choices about which spells you learn, and some of the spells don't work at all, meaning you can be stuck with useless spells and be unable to get useful spells because of that.

The remakes of FF1 fixed these issues:
* You're still limited to 3 spells per spell level, but you now have the option of discarding spells. You don't get the money back, and you'll need to find the store and pay again if you change your mind, but this at least allows you to take a different spell instead.
* The broken spells I mentioned above now work properly. (In particular, TMPR, called Steel or Temper depending on translation, is now one of the most useful spells in the game.)

(FF2 and FF3 *did* allow unlearning spells. In FF2, you lose all of your spell experience for that spell (so it goes back to level 1 if you re-learn it. In FF3, there's no penalty for doing so; the spell goes back into your inventory, and it can be taught to someone else without having to buy or find the spell again. FF4 and later changed the system of learning spells to the point where this discussion no longer makes sense.)

Usually 2).

If there is a character creation where you pick some stuff, I look up precisely what everything does. If you just start a blank character and build it up with level ups, I usually spend a couple of hours immediately after starting the game just reading up on every skill, attribute etc. and planning my build/path. That being said, I never spoil anything from the game to myself. I only look up how everything works mechanically so I can make the most informed decisions.

Depends on the game, and how hard it is.

For games with no customization I usually play the pregens, like int the early Might&Magic games.

For hard games I try to optimize my characters.

For easy games I choose the first character(s) I roll.

For party based games, I try to have as diverse a squad as possible.
The ironic thing is that nowadays having races with different strengths and weaknesses is "racist", so in a modern game where diversity is only in appearance, I'd probably play a squad of white human males instead of a diversity squad.

In my CARPG I totally prefer to make my characters more LGBT... QPNZ with the WASD movement... and a mouse. I always play with a mouse.

I use the in-game information, common sense, and my previous experience with RPGs in order to judge what would most probably be a decent enough character, but also factor in what looks like the most fun to me. So tendentially #1, but I wouldn't call it "blind" or "random". Later, when leveling up, I exert caution, and if the game allows it and I'm not quite sure yet about how everything works, I often save up points, until I get a better idea of where they would be most beneficial to me, so that I stay flexible and can adapt as soon as I run into a trickier challenge. Occasionally, I will also look for specific info in guides or walkthroughs when questions arise, but I seldom do that before starting to play the game.

In wRPGs (which I don't often play due to not liking them too often), I'll often go the "hit random until it seems to appeal to me" route.

Otherwise, I prefer to know what I'm getting into. It upsets me when there isn't good easily available first party documentation about what stats/etc mean. It doesn't have to be FULLY transparent, but enough to make informed decisions. A lot of j-action-RPGs ("Souls" for instance, and others) seem to be gleeful about making sure the player knows absolutely nothing, and it's gross. That necessitates community resources (and makes community waste time having to test all these things, and even years later discover hidden mechanics...)

Sure, Final Fantasy Tactics has its super in depth Battle Mechanics Guide that you'd never learn from the game without deep experimentation (https://gamefaqs.gamespot.com/ps/197339-final-fantasy-tactics/faqs/3876), but the game's in-game help is REALLY GOOD about informing you about the majority of what you need to know to make decisions. (Of course excluding bugs that are just that -- bugs.)

So what I'm saying:
The game should prevent your #1. Include your #2 but at much less time. And make your #3 not necessary.

I'm personally really not fond of #3 at all. Build Guides are anathema to what I enjoy, and games that encourage and need them suck. Grim Dawn's (an ARPG) a great example. Pretty much any build, even haphazard assignment of points, can beat the first difficulty (normal). Being sensible at all about your selections (i.e., selecting anything even remotely synergistic rather than actively selecting skills that do not work together because they're all completely different damage types) can get you through the second (veteran). If your build can do veteran, you can do elite (the third difficulty) so long as you max your resists. Ultimate has some filters to it, but usually if you beat elite, you can probably get through Ultimate if you realize you need to shift some of your offense into defense (if you didn't have enough defense already). The handful of optional megabosses often need more specialization... And I feel that's a weakness of the game that only certain builds can handle them rather than a broader base played well. The game is good at letting you know all of the above if you read in-game help and hover help text. They also provide a good manual with tips that goes beyond this, including even the formulas for attack/defense. AND YET there's still plenty of room for optimizing... to clear faster, to do optional areas (Crucible arena mode, or Shattered Realm infinite dungeons) deeper, to find the actual perfect gear for a build.

EDIT: Oh yeah, and the game should have at least one big good respec, if not multiple easy ones. One shouldn't have to restart the game if they find their character isn't doing what they want to after they've learned mechanics. But most games that are opaque are also those without respeccing. Usually if the game is annoying someone, the developers did something wrong. Not the player.

For a challenge I choose the hardest settings a game offers.

Characters however, are always minmaxed for optimum performance.

I've only played Morrowind, but I came to dislike how it skilled. IIRC, your skills improved every time you succeeded at a task. That means weak skills were very hard to improve, while strong skills were always improving. I think that's unrealistic, and I saw a mod that reversed that. Skills improved for every FAIL, meaning low skills advanced quickly, as they should, and high skills advanced slowly, as they should.

Sadly, i didn't grab the mod when I saw it, and have no idea what it was called.

There's an old game called Dungeon Master which advanced skills as you used them, so your gameplay decided what character was good at what.

Also, while I think Might and Magic 9 is a Game That Shall Not Be Named, it did have an interesting method of development. When you start, you only get to choose from two classes: might or magic. That's it. After advancing many levels of either, you get a new choice. Mights could become warriors or rogues, and magics could become priests or wizards. Later, each class gets one final choice. Again, two choices depending on your current class, but the four total bases had eight total choices. So, as you go through the game and see how combat works, you can make a perhaps more informed choice of how to fine tune each character.

It was different from just choosing druid or thief right at the start.

Some games handled this better:
* In Dungeon Master, a skill failing would still get you experience, but not as much as succeeding. Furthermore, you get more XP later in the dungeon, and more powerful spells yield more XP on success. (One strategy I found was to use potions to increase your Wisdom, which gives you a better chance of succeeding at spells that are too difficult for you to cast reliably, then casting said difficult skills to get a lot of Priest/Wizard XP; note that, at high enough Priest/Wizard levels, you will always succeed at spells regardless of stats.)
* In Wizardry 8, failing a skill gives you *more* skill increase chances than if the skill were successful. Higher level spells give you more skill increase chances.
* Final Fantasy 2 and the SaGa series don't really have a notion of success/failure for a skill, so the chance of increase is not affected by success or failure. One thing these games do well is that, the stronger the enemies you're fighting, the higher the chance of stat/skill gains.

That particular game also favored more balanced characters; a class you haven't leveled up in will level quickly, and every class level will give you stat gains. So, Wizards could get more mana by gaining Priest levels (and vice versa), and everyone could get big health and stamina boosts by getting a few Fighter levels.

(By the way, did you know that stamina growth in Dungeon Master is exponential?)

For skills, yes, it's great, though it still has its failings, pointed out by others already. But definitely not for attributes and stats. For it to truly be a great system, attributes should also increase with use, as in every X increases of associated skills, and stats should likely be tied to them, doing away with levels entirely.

SaGa games (except original SaGa 3) do away with levels entirely, but it turns out that the games tend to either have fixed stats (only skills and HP/WP/JP increase) or no skills (so stats and HP/WP/JP are the only thing that increase).

Often times:

> Read about the stats and skills superficially, and design a character
> Play for a few hours and get dissatisfied
> Carefully read about stats and skills, and create a new character
> Play for a few hours and get dissatisfied
> Consult character creation guides, and create a new character
> Play for a few hours and get dissatisfied
> Play something else

I think it took something like 10 tries before I eventually started with a Neverwinter Nights playthrough that I saw through to the end :P

Of course, then there's games that don't have these sorts of guides available. I currently have a game of Progress Quest running, and there really is no information about what the differences (if any) between the different races and classes, or what effect the stats have. (I have determined that STR affects encumbrance, so higher STR characters can carry more items before having to go back to town to sell stuff.)

The game also has the issue that XP awards are paltry; they remain at single digits even when levels take over 20,000 XP each. The only reason I've even reached level 22 is because the game plays on its own, so I don't actually need to do anything to gain those levels, and can just let the game run. (It's also the *only* way to play the game, as it turns out.)

Yes, sometimes I'm curious about the mechanics of the game, even if the game is so easy it's not important to know them, or even if the game is a 0-player game. I've actually been thinking of looking at the source code to se how the game actually works.

first read throu the whole char generation process to see what alternatives I can get
then it depends on the game as they are way too different to use the same steps
sometimes I search for viable builds just to be sure i wont start with a party/champ that will make game painful or impossible to finish

and if it is an action game like diablo i just lvl up a few with different classes to see which one i would like to play with