I only skimmed through some of the replies to the OP, so forgive me if I missed someone mentioning this. However, when speaking of party composition, realize that in Secret of the Silver Blades and Pools of Darkness (and maybe some of the other higher level campaigns, though I never played them), as soon as a Ranger can cast level 1 Mage spells (level 8 or 9, iirc), he or she can be dual classed to mage. Once the Mage level exceeds the level in Ranger, all Mage spells can be cast in full armor. Tank-mage FTW.

Sure, it takes time, and the mageling will be very weak (yet still retain the Ranger HP pool they had gained earlier), but they just become silly powerful later on. They are essentially the Eldritch Knight prestige class from D&D 3.X, but with even fewer casting restrictions.

Edit: Technically, a Ranger could be dual classed to Mage at any point after level 1, however it should be only be done after the Ranger can cast mage spells on his/her own if one intends to take advantage of the mage-tank 'loophole'.

--------

If one uses the ranger/mage option, it makes forming a party pretty trivial as they can perform both casting and front-line duty.

Personally, I would go with a party composition of this: Ranger/Mage, Ranger/Mage, Paladin, Fighter/Thief (preferably dual classed), Cleric, and a 6th member of whatever you want. I think I went with an elven Fighter/Mage/Thief, just to have a demi-human represented somewhere (since thief levels aren't capped); not the optimal min/max choice, I admit.

Yeah, a Human dual classed Ranger 14/Mage is the overall most powerful character you can have in the Pool series. A tank with two attacks per round who can cast spells while wearing metal armour; with Mirror Image and Fire Shield he's more powerful than a single classed Paladin, Ranger or Mage.

For the Fighter/Thief I prefer to use a Dwarf. A dual classed human is more powerful in the long run, but if you don't have another Thief it means going without one for a long time, although in Pool of Radiance you can probably use a hired one for the few instance where you need one.

I think what I did with my Fighter/Thief is that in Curse I just used one of my imported Fighters to dual to Thief, then made Rangers and/or Paladins to replace the others. A dwarven Fighter/Thief is certainly a valid option, and probably good for some race diversity (the Fighter level cap for dwarf is relatively high, as I recall).

In Radiance, I think I used a multi-class as my thief. An elven Mage/Thief, perhaps? It was easy enough to swap out in Curse after my fighter had leveled the thief dual class up a bit.

How would your party compositions change if we add a restriction: No two characters can have the same class or class combination. (In other words, having a Fighter and a Fighter/Mage is allowed, but two Fighters or two Fighter/Mages is not.) (For this purpose, count dual-classes separate from multi-classes.)

Personally, I would prefer my characters to not be clones of each other.

For Pool of Radiance it would be easy, because using multiclass demi-humans doesn't hamstring you as much (since it is harder to reach the race caps). So, you could just mix and match to your heart's content. Just be sure to have at least 2 tanky classes, at least one cleric, a couple mages, and a thief.

But for the later games... hmmm...

Ranger/Mage, Paladin, Fighter/Thief, Cleric, Mage ... some other dual class (Fighter/Mage, Fighter/Cleric) depending on play-style. Something like that, perhaps... EDIT: heck, just go with a regular Ranger for the 6th position, if you want to avoid the hassle of dual-classing another character.

Same here.
I embrace diversity and multiculturalism - in CRPGs. I always try to have all races, genders and classes represented, but it's not always practical, which I learnt in Curse of the Azure Bonds (my first GB game), when I ended up with a lvl 5 Half-Elf Cleric.
In the Pool series having any other demihuman than a multiclass thief is unpractical, so I usually end up with this party:
Human Paladin
Human Ranger (may be dualed to Mage)
Human Fighter (may be dualed to Mage)
Human Cleric
Human Mage
Dwarf Fighter/Thief

Small addenda:

Knights of the Sword can't advance above 18th level, so you should wait until your Sword Knight gets enough XP to be an 18th level Rose Knight and then switch him over. This is probably because the devs figured out there was no advantage to going for Rose Knight.

Party holy wars: it's really all about your preferred style of play. I like being able to toss high-level fireballs at the enemy, so I like to load up the single-class wizards. Some of you may enjoy building invincible tanks and sending them into combat, so you make multiple Knights. It's just a game.

Female strength: this was fixed in 2nd edition, but the Gold Box games are 1st. Demihuman level caps are, again, fixed in 2nd ed; apparently the original reason was that Gygax hated Tolkien and the nonhuman races were added to help sell the game in the Tolkien craze. As for why they did that--it was 1975, as people have said. The past is a foreign country, and the further you go back the more foreign it gets.

Diversity:
If you're just going to play Pool, you can go pretty far with this one. You could probably get away with Dwarf Fighter, Gnome Fighter/Thief, Halfling Fighter/Thief, Elf Fighter/Mage, Half-Elf Cleric/Fighter/Mage, Human Cleric--suboptimal, but still winnable.

Actually, demihuman level limits were not removed until 3rd edition. Dungeon Hack, which is 2nd edition, still has them (though relaxed somewhat for that particular game). The developers of the Infinity Engine chose not to implement that rule.

That's right. I meant the *low-level* limits were fixed, but demihumans didn't have full advancement until 3rd.

It's kind of funny for me as a guy who grew up playing 1st to find all the people who got into it playing 3rd going, 'Low female strength? Wha? And half-elven clerics can't get past five? What's with these weird dual-class rules?"

I had to look that up, I thought you were kidding about it. But nope, it was real!

Yes, but I really don't recommend the option for a first-time player.

Dual-classing is, mechanically, a very bad rule in tabletop, for obvious reasons -- it trades power now for power late Loaning or investing power is bad for the collective experience -- a single player's character should not be substantially more powerful or less powerful than the whole party -- and specifically putting power into a savings account is bad for the player, since there's no guarantee that the game won't end until it starts paying off.

Dual-classing originated when a mage character got stuck in a no-magic world, gained plenty of experience, and the player suggested that he should gain a level. Since he was incapable of using his primary abilities (spells) and thus was unable to practice them, and also because he didn't really have any use for spells while stuck in said world, the referee ruled that the character should gain a level of fighter instead.

These contrived circumstances illustrate why it's also a very bad rule from a roleplaying perspective -- it doesn't actually correspond to any realistic situation which might happen during the adventure -- and the roleplaying opportunities in Gold Box games are few and precious.

And from a mechanical perspective, in CRPGs, to use dual-classing effectively, you really should know when (and how) to persevere and when to cash out. If you're a struggling player with no previous knowledge of the game, you don't need the extra difficulty in the beginning. If you're very good and/or have previous knowledge, you don't need the extra power spike in the end other than for power's sake (and you won't need advice for that decision). So I don't recommend it for new players.

But yes, you're right, dual-classing will eventually result in very powerful characters.

If done in Secret of the Silver Blades, it actually goes very quickly. Experience comes fast and quick; there would only be a few encounters when the newly dual-classed character would truly be gimped. Even so, I wouldn't do all intended dual class characters at once, I would stagger them, doing them one at a time. I think it is a judgment call, depending on if the player is struggling or not, which, as you say, is aided greatly by experience.

As for what you say about it being a bad rule, I don't disagree with you at all. It is one of the things WOTC changed in 3rd Ed. that I really think was a good move. The way adding new classes to a character, and eventually allowing for sensible prestige classes, was designed in the later edition made it a bit less artificial and more organic. Some of your arguments against it still stand in the later version, but I think it is a lot less pronounced.

Possibly that is part of it. However, much of it was also because Gygax didn't want humans to play second fiddle to demihumans in any major way. Hence the level caps on demihumans and the addition of allowing humans their own form of multi-classing. Gygax was not a fan of demihumans (at least not as played characters), and it has been said that he only ever included them because Lord of the Rings (books he didn't like at all) was so popular.

Which is why I think a 'best party composition' list should include them. A caveat can be made that it is certainly tougher to accomplish, but it does end up being the superior party, from a min/max standpoint.

With this said, I find that 2e multiclassing rules work better than 3e multiclassing rules for spellcasters. With 2e multiclassing, a multiclassed spellcaster can actually cast spells decently, being only at most one spell level behind at lower levels and already having strong enough spells to be useful at higher levels. In particular, you trade a bit of power for a more diverse ability selection (which is how multiclassing should work IMO). A Cleric/Mage has more total levels of spells than a single class spellcaster, but less powerful spells in the upper levels.

In 3e, however, you split levels (not experience) so a balanced multiclass character (something you rarely see in optimized builds) has the same number of levels as a single-class character, As a result multiclass character's spells end up being weak, having only half the spell levels of a single class character. A Cleric/Mage ends up with weaker spells than a single-class spellcaster and does not get the increased spell diversity to make up for it. This problem was so bad that they made the Mystic Theurge prestige class to patch this problem up a bit. (Also, the 3e Spell Resistance mechanic means that being behind in casting levels makes offensive spells useless rather than just weaker.)

Incidentally, Bishops in most of the Wizardry series suffer problems similar to 3e multiclass spellcasters. They get new spell levels at about the same rate, but they are divided between Cleric and Mage spells. And in Wizardry 6 and 7, with the new skill system, Bishops still have to split their spell picks evenly between Cleric and Mage spells. (Wizardry 8 actually made Bishops actually good.)

By the way, Elminage Gothic Bishops suffer from the same issue, except that the game expects you to reach levels where Bishop spellcasting caps out and the only difference is Mastery skills (which are more reasonably balanced here).

Or, altenatively, just take the quick route and dual class from Fighter to the class you really want at level 2. Makes your casters more durable and doesn't significantly delay getting spells.

I agree with you about how spellcasting classes are a lot less balanced, but that is where prestige classes such as [url=http://www.dandwiki.com/wiki/SRD:Eldritch_Knight]Eldritch Knight[/url] or [url=http://www.dandwiki.com/wiki/SRD:Arcane_Trickster]Arcane Trickster[/url] come in. They both allow for advancement in spell levels, but also in skills applying to their other base class. For instance, a 2F/5W/10EK casts at level 14W, so only a loss of 3 spell levels for a very tanky character. That is all I'll say about it, as I don't want to sidetrack this thread more than I have already.

Edit: Forgot for a moment that Eldritch Knight capped at 10 levels. Corrected.

Also a good example.

A viable option.