Test case, really: I'm assured that it's fine to put critical comments on this forum provided they're not rude or unconstructive, and I've no reason to think that isn't true. Actually this isn't even criticism exactly, but hopefully it's constructive!

I wrote a rather long post on the G3 board a few days ago about the design choices that go into different AI/tactical mods (IA / Tactics / Big Picture / my own Sword Coast Stratagems / etc) which seems to have led to a modest amount of constructive discussion. Since I think there's quite a lot of people who read BWL and not G3 (and vice versa, no doubt) and since a lot of people interested in tactical mods read this forum, I thought I'd copy the comments here.

Some thoughts about what makes tactical mods different from one another. Basic theme: it's not exactly about "good" vs. "bad" mods, more about the demonstrated fact that different people want different things out of their tactical mod.

I'm going to use, as examples: Tactics, Improved Anvil (IA), and my own Sword Coast Stratagems (SCS) - the TUTU version and the upcoming BG2 version. I want to make very clear that I'm not doing this as an excuse to sell my mod! - obviously I prefer it, or I wouldn't have written it, but equally obviously there are plenty of people out there who it doesn't suit and who'll have more fun with another AI mod.

That said, here are ten questions that you might ask to work out if a tactical mod is for you:

1) Is it customisable?

At one extreme, you've got Tactics (or Improved Battles, come to that). It's modular by its nature and so it's very easy to choose exactly the components you want.

Not all mods are innately modular like that. SCS, and IA (and Big Picture, come to that) muck with various interrelated aspects of the game, and the bits they muck with are chosen precisely because they interrelate. (Sikret introduces lots of new scary golems and then changes the spell system so as to include anti-golem spells; SCS makes Protection from Normal Missiles stop magic missiles too and scripts wizards on that assumption.)

At that point it's a bit of a judgement call what to do. SCS errs on the side of extreme modularity: you can still pick whichever arrangements you want, even if that leads to a slightly odd mix. IA errs the other way: it's all-or-nothing, pick the whole shebang or don't pick anything at all. (BP isn't far off that.) The advantage of my strategy is that it lets players pick and choose just what they want; the advantage of Sikret's is a more stable install and less risk of choosing some really mad combination of options.

2) Is it compatibility-friendly?

Again a judgement call. There's a liberal strategy (SCS's): maximise technical compatibility, let the user worry about conceptual capacity, work to minimise possible clashes. And there's a conservative strategy (IA's): don't risk a user having a messed-up experience and not appreciating their mod by mixing it up with another mod.

In a sense, G3 modding philosophy (if there's such a thing) is in the middle here: technical compatibility is maximised but most people have in mind a relatively (cmorgan-style, perhaps?) tame install. IA (or NEJ) are at one extreme; mega-modding is at the other.

(Tactics isn't wildly compatibility-friendly, but that's an accident: it's an old mod, technical compatibility was harder in those days.)

3) Is it faithful?

This is kind of hard to pin down, but I mean something like "is it in the general spirit of the original BG2 / the Forgotten Realms"? A mod that does nothing at all except improve AI and targetting definitely passes; a mod that adds lightsabers to the game definitely fails. But in practice things are usually murkier than that. (Ascension is very faithful, for instance, even though it adds "cheesy" abilities to bosses, because it doesn't really go beyond the general flavour of ToB in doing so.)

There's no particular virtue to being faithful. Bioware isn't God. But some people (me, for instance) like the flavour of the existing game and want to keep it. Others are bored as hell after their 1000th replay and want variety.
I'd say that Tactics is not especially faithful, nor is IA; I hope that SCS is.

I say again: this is a design choice, not a criticism. Sikret says of IA that it's a "long-term project to revolutionise BG2" (I'm quoting from memory). SCS absolutely doesn't aim to "revolutionise" anything. Take your pick.

(There's also a subgroup of players who are after a slightly different notion of "faithful": namely, "faithful to 2nd edition PnP D&D. That's not what I mean here though.)

4) Is it holistic?

That is, does it uniformly modify the game environment or does it just tweak individual encounters? Does every mage get enhanced or are there still lots of stupid ones?

In principle, holism is always good. More is more (plus, it helps realism). In practice, it comes at a price, because BG2 has hundreds or thousands of creatures and it takes forever to write material for all of them. IA modifies encounters more-or-less creature-by-creature, and inevitably that's a huge project and lots of creatures aren't done yet. SCS modifies pretty much everything, at the price of having relatively few distinct scripts: every single SCS mage uses the same script, for instance. Hopefully it's relatively hard to spot because the scripts are very long and cover a lot of possible situations; even so. I went this way because I find it jarring when some enemies are bright and others are stupid; players who want to maximise the variety of the challenges that are improved might not want to go with holism.

(Tactics does something in the middle, incidentally: it modifies whole groups of creatures - mages, beholders, illithids - but leaves many others alone, and it's not always completely careful about ensuring [a]all[/a] creatures of a given type are covered.)

5) Is it exploit-blocking?

By "exploits" I mean things like off-screen cloudkill, extensive use of invisibility, repeat thieving, etc, etc. You can take one of two lines on this: you can alter the rules so as to make them impossible, or you can do the best you can within the rules and accept that this won't always work. In SCSII, for instance, creatures try to run out of Cloudkill, but they're only medium effective; in IA, often their regeneration rate is so high that clouds don't matter. In SCSII, very few creatures can do much about traps, so only the player's self-restraint really stops them from trapping their way all the way to Ch.10; in IA, even relatively mundane creatures like Roenall's wizard are trap-proof.

The case for relying on players' self-restraint? It's hard to block exploits without overkill and a feeling of unrealism. The case against? It can spoil players' immersion in the game if they have to restrain themselves rather than being able to do anything they like within the rules.

SCS and Tactics basically don't block exploits. IA puts a high value on blocking them.

6) Is it transparent?

That is, can you normally tell from context, things you're told, and general BG2 experience what sorts of powers and immunities a creature has, or is the only way to work it out through trial and error - and, often, through many reloads. You can reasonably predict that liches are immune to nonmagical weapons and that fire may not work that well against fire giants, but if a human mage is immune to 5th level spells then it's going to be almost impossible for you to guess that.

Whether you like transparency has a lot to do with how you feel about reloads and metagaming. Some people (like me and Berelinde) prefer to reload hardly at all on the grounds that it feels less immersive and realistic. Others find that only through non-transparent special attacks / defences is the game still adequately challenging.

SCS tries very hard to be transparent. Tactics is fairly transparent, with some exceptions. Ascension is semi-transparent: lots of creatures have random immunities but they're usually not so high that you can't hurt them. IA is not transparent at all, although arguably it gets more so as you get the general hang of the mod.

7) How hard is it?

A rough measure of this is: how often do you want to have to reload? Hard to measure in practice, though, because people have different styles. And "hard" doesn't mean "clever", of course: you can make the game harder in more or less crude ways.

IA is generally agreed to be very hard indeed: all but the best players reload extensively. SCS, Tactics and Ascension are roughly as hard as each other, and much easier than IA.

8) Does it have double standards?

That is: can the enemies do things you can't do? This is a bit vague, of course (as I've noted elsewhere). Firkraag can do something you can't do: breathe fire. A beholder can cast ten uninterruptable spells per round. etc. But generally speaking, if enemies cast uninterruptable spells, have undispellable ultrahigh resistance, or put eighth level spells into their spell sequencers, the mod has double standards.

(I'm leaving out pre-battle buffing; that's a special case.)

Everyone agrees that double standards are in principle a bad thing. But there's also a widespread view that it's the only way to compensate for players' vastly higher intelligence compared to the computer, so (again...) saying that a mod has double standards isn't really meant as a criticism.

Tactics has a lot of double standards, normally at the script level: it force-casts a lot of spells, notably. Ascension and IA try quite hard at the script level to make their creatures play by the rules, but they extensively modify the creature files in ways that aren't usually available to the player (immunity to traps, ultrafast casting, uninterruptable casting, and the like.) SCS has a pretty sustained go at avoiding double standards, though it isn't quite perfect (see below).

9) What is its targetting like?

How carefully does it decide to use its spells?

This is hard work from a scripting perspective, and hard to work out except by playing a lot and/or by carefully dissecting the code. But to (say) target a charm person spell on the nearest PC who's not already out of action in some way and who's not charm-proofed, magic resistant, improved-invisible, or protected by spell turning, takes about 300 lines of code. (Very few entire scripts in the original game are that long). So most mod-writers take a lot of shortcuts. That sometimes works and sometimes doesn't; it can be relatively easy to catch them out if your strategy isn't quite what the mod-writer anticipated).

It is possible to hold out for really careful targetting all the way through (SCS tries very hard to do this) but it isn't cost-free. Leaving aside the cost in time and effort for the modder, it has two big problems:

i) it leads to crazy-long scripts. The longest SCSII script is 19,000 lines long, which is too slow for older computers and occasionally causes short lags even on modern ones.
ii) it leads to psychic opponents. Enemies detecting your Minor Globe of Invulnerability is one thing, enemies seeing that you're wearing a Helm of Charm Protection is arguably another; but it's hard to avoid this sort of thing without making enemies into the sort of morons who'll try fireballing you ten times even though you took no damage any previous time. (There is a well-known way around this, but it roughly doubles the length of the script).

So there are principled and practical advantages of sub-perfect targetting. On the other hand, it's kinda cool when enemies walk to the edge of cloudkills, avoid melee with creatures protected from their weapons, re-cast their own protection from magic weapon spell as soon as their first one is dispelled, hold off from casting further offensive spells on someone about to disappear into an Imprisonment effect, hold off their magic missiles until there are badly-wounded enemies to finish off, etc - all of which are only possible with serious targetting.

10) Does it have an arms race?

Does it provide, as well as enhanced dangers, enhanced rewards? (e.g.: powerful magic items; more experience).

Some people want improved tactics and only improved tactics. Others think that with great risk comes great reward. (Of course, that means that the next risk is still more challenging - etc, etc.

SCS borders on the puritanical - no new rewards at all. Tactics gives out lots and lots of rewards. IA is (or at any rate was) built around the idea of collecting and making ultra-powerful items; on the other hand, through various different means it tries to keep the arms race a bit under control.