I love this topic and have some stuff to say, but I'm at work. Posting now to make it easier to find this topic tonight.

Cheers!

Actually, Final Fantasy turn order was completely random, with the algorithm favoring the enemies. Specifically, the game starts by ordering the combatants, enemies first, and then swapping random pairs of them. (The remakes made it so that agility affects turn order, but some enemies (like Wolves) were made more annoying by the change, as they now always act before you unless you get your party's agility up *really* high.)

It isn't until Final Fantasy 2 where turn order is affected by stats; in this particular case, evade % is the stat used (which means characters wearing heavy armor will generally act last).

There are some games, notably Wizardry 1-5 and SaGa 2, in which the initiative algorithm favors the party. (In SaGa 2's case, this is clearly deliberate, but (if you're not playing the DS remake) can lead to initiative overflow, usually in Robots).

Personally, I think a limit of 18 is way too low, as it doesn't leave much room for the stat to grow over the course of the game. In fact, unless the game has a really low level cap, if stats grow at least one point per level, all stats will cap out way too early. (Also, if using a level system, I would prefer to avoid random stat growth; then again, I would probably prefer to use a different system, like something resembling how many SaGa games handle stat growth.)

Also, I think the difference between 35000 and 34900 *should* be small, as it's less than a 1% difference in stats.

Well, I like "low stats" - that way game designer can easily make/program real difference between their levels so, for example, they could be described not only just by number but also by text description. In "high stats" system all numbers lose their single importance and can lead to smaller excitement of the player when he/she gains a level.

Interestingly, my idea is to move away from having exciting level ups and no growth the rest of the time and instead use a system where characters constantly improve their stats without big sudden boosts. Of course, there will still be the excitement of learning a new ability, but that won't be the only time you get to see growth.

The sort of system I'd be looking at will simply not work with numbers that low; it needs at least moderate values to make sense. (They don't need to be as high as the values I used as an example for speed, but they still need to be able to reach into at least the upper double digits.)

I like "low stats" because when playing I like to plan a lot and take my time to make next move and "low stats" are easier to think about and "not so complicated". :D

You said you have stuff to say about this topic, but it seems you must have forgotten, so I wil bump this topic with this reply to your post in order to remind you.

Thanks for the reminder!

Turn-based tactical combat (position matters, and movement is part of the combat turn), large-ish party (4 minimum).

Some stat (speed if it's a dedicated stat for this purpose, dex or something else if you want a derived value) determines how many rounds out of a hundred you participate in between a value of, say 80 and 95. I have to sit down with a pencil and do some figuring to get the range exactly right. But the point would be that nobody in a combat party is so slow that they miss out on more than 2 or 3 rounds in 10. And nobody is so perfect that they never miss a round. Sort of like rolling a one on a 20-sided die is an automatic critical failure. Enemies might fall outside this range - but there would be compensating factors. Imagine a rock golem who only participates in 35 out of 100 rounds, but his damage resistance is incredibly high, and his attacks do 3x the range of more common creatures' attacks.

OK, so now you have this value per 100 that determines how many rounds each character will participate in. Rolling dice creates a round-by-round set of who gets a turn that round. Then, turns order is determined by a different stat - or derived from speed divided by dex or something. Again, need some time with a pencil.

Because we're calculating rounds out of 100, and battles rarely reach that length, there is a chance that even a slow character will not get skipped in a short battle, or that even the fastest character misses the second round (or even the first, if they get a critical failure on their initiative roll or something like that).

So, now you have your rounds and turn order mapped out. These are then changed a bit by some additional rolls - for initiative, clumsiness, etc. And the results may be changed throughout the battle by spells and injuries and terrain and such.

I think a key to making a system like this work would be the addition of automatic reactions. So, when attacked, not only does a character get a block chance, but also gets a chance to counter. A successful dodge roll has a chance to move the character to the next spot over on the grid (adjusted for terrain, that spot being occupied, etc.) .

The upcoming turn order would be represented on-screen by a succession of party and enemy avatars - I've seen this in a few games. So, yes, you would get to plan ahead some. The turn order would only display some given number of turns ahead - say 8. And even those 8 are subject to shuffling based on what happens in battle. A bard could successfully sing a rallying song, and half his party could suddenly get an additional initiative roll, etc.

You're welcome. It's interesting to read what other people have to say.

Personally, I don't really like the soort of battle system you described. Specifically:
* I prefer the battle system to be non-tactical. Tactical battle systems have the downside of making battles, especially battles against easy enemies, take longer. In a non-tactical battle system, a fast auto-battle option could allow the battle to be over in a few seconds; you can't really resolve tactical battles that quickly. There's also the fact that I find the numerical aspects of RPG combat more interesting than the spacial aspect.
* To be honest, as I've said before, I don't like having a stat determine the number of turns a character gets; turns are just too valuable a resource in turn-based games to handle that way. Your rock golem example could be handled without that; make it strong, but also make it so that it has to spend a turn charging up before it can attack. (I note that this makes the Defend command, if present, particularly useful when fighting enemies of this sort.) I would just give everyone one action every single round, with extra actions only happening in exceptional circumstances (bosses getting extra attacks, high level spells like Time Stor, etc.).
* "speed divided by dex" raises a few eyebrows. For one thing, I find it confusing to have more than one out of Agility, Dexterity, and Speed as stats in the same game. (I note that Wasteland has all three of these as stats; go figure.) Also, whenever there's a division, I worry about what happens if the denominator is low; a low denominator can lead to unexpectedly large results, and a denominator of zero can have strange results (which might include the game crashing with a division by zero error).
* I don't like automatic reactions, at least not as a default ability. I don't like the fact that someone gets an extra turn just because of something the opponent decided to do (this applies to both counter-attacks as well as things like attacks of opportunity). Also, such mechanics might easily favor physical attacks over magic, or if spells aren't countered, might sometimes favor the reverse.
* You mention the game showing the next 8 turns. That, however, doesn't work well if you're fighting, say, 396 berserkers; in that case, even if only 3% of them get to act before your character's next turn, your character won't appear in the turn order listing.

I think this is just taste or preference. I enjoy tactical rpg battles more.
I don't think this is too much of a problem. For one, it is mitigated by my range between 80 and 95. Even a low speed character is viable. But a high speed character does get an advantage. I think with some fiddling, this could be properly balanced.
I do like them. Mainly because attacking someone should be kind of risky. In Darkest Dungeon, several characters have an action that activates "riposte," which is a reaction attack. It comes with a reduced damage and some other mitigating factors. And it works well. I think this is a matter of working out the right balance, rather than the idea being just good or bad.
Yeah. This system would not be suitable for very large battles. It also wouldn't be suitable for very small battles, which is why I stipulated minimum party size of 4. And that doesn't bother me either, because battle systems in general are not very scalable. The size of battles tends to be fairly consistent within one game.

In the context of a tactical battle system, there is an inherent risk in using a melee attack: To use a melee attack, you (generally) have to go right next to the enemy. This means that, if you don't kill the enemy, the enemy will be able to attack you on their turn, putting your character at risk.

Dungeons and Dragons 3rd edition has a particularly interesting example: If you move right up next to the enemy and attack, the enemy will, on its turn, be able to make a full attack against you. Assuming a base attack bonus of at leastl +6 (for a Fighter, this means being at least level 6), or if dual or multi wielding (some monsters can make more than two attacks), this means having to deal with lots of attacks from the enemy this round.

In a non-tactical battle system, attacking means you don't get to defend or heal this round, so there is still this cost of not attacking. For example, if we look at Dragon Quest 2, where there's a reusable item that heals the user for free, if you are low on HP, you might not want to attack. On the other hand, there are some enemies you might not want to leave alive, either because they can heal (the final boss can heal itself fully, for example), or because the enemy can do something nasty (like casting a spell that can instantly kill your entire party).

Different sized encounters can lead to good variety. I can think of a few games where encounter size varies:
* In Bard's Tale 3, you sometimes encounter dozens of enemies, but often those enemies die easily to group spells. On the other hand, occasionally you might encounter more powerful enemies who don't die so easily, and there you might want to use single target attacks (like melee attacks or death spells) so that the enemy dies faster.
* SaGa 2 is interesting. Some areas tend to give you small encounters, where physical attacks (especially if you have a robot with high agility) can end the battle quickly; however, you might not want to end the battle quickly so that your humans and espers (mutants) get a chance to increase their stats. On the other hand, some areas have you fight hoards of enemies at once (and in the DS version, there are chain battles where you fight even *more* enemies, perhaps as much as 50, in one battle), and there you want to use magic to kill lots of enemies quickly; those battles are very good for getting money (not XP, which doesn't exist in this game). Then you have bosses; mostly single enemies, and the later bosses take half damage from physical attacks, somewhat balancing robots and spellcasters.

Also, I find that battle systems with positioning tend to flow better if you only have to control one character; Ultima 4 is a great example of this. In Ultima 4, the game flows best if you don't recruit any allies; if you do, battles take a lot longer. (It's worth noting, however, that you can't actually beat the game without a full party; you can, however, pass all the game's combat challenges solo.)

Kings bounty has a good turn order mechanic, and it is very significant. Given that damage dealt is propotional to units surviving, if you can kill all or many of another stack then you can effectively neuter that unit before it gets a go. As a result it's important to have some high "initiative" units in your army so that you can be sure you'll get a turn in before most enemies.

I feel like Heroes of Might & Magic V really nailed that down. It is a very dynamic and variable system, depending on many factors like base unit initiative, many items that boosted the initiative of units, positive or negative morale, spells affecting initiative (haste, slow and such) or using the wait mechanic.

The base mechanic behind it is that a unit with an initiative of 10 can take 1 action every combat turn, meaning a unit takes an action every 10/X turns where X is unit's initiative. So a unit with 15 initiative can take 3 actions within 2 combat turns while a unit with 5 initiative gets only 1 action every 2 turns. The game takes into account even decimal numbers (but those are not visible). This makes initiative one of the most valuable stats on units and those with very low initiative are generally viewed as just fodder as they don't play very often and their movement speed is usually not that great either. But they are in most cases much more durable than their faster counterparts and some of their main uses are either defending your shooters or blocking choke points.

I prefer ATB systems,

Why wait for an entire command string to be executed and then deterministicly shown (making strategy boring), when you can throw some active variance into the mix? Pure turn based is quite boring, unless executed in a different way or exceptionally well.

(Hint: Pokemon is turn based gone horribly wrong, for reference.)

What's wrong with Pokemon? It actually has a lot of depth when taken to its limits, like in competitive play or speedrunning. (Note, of course, that I am looking at speedruns that play the game "normally"; many first generation speedruns (especially the 151 category) end up using major glitches and don't actually see much of the battle system.)

One problem with ATB is that it causes the game to no longer be strictly turn-based, which can be a problem for those who can only play turn based games (like older people whose reflexes aren't what they used to be, or those with disabilities that impact their ability to play games).

Also, I hope your model of ATB is not FF6 or FF9; those two games have a serious issue due to time not stopping during animations, resulting in the queue filling up; the fact that I have to set the battle mode to Wait and enter a submenu just to enter the queue is a sign that the game is flawed. I should not have to do that sort of thing (just as I shouldn't need RNG manipulation on a casual playthrough of a game unless the game is intentionally designed with RNG manipulation in mind, and I know of no such games).