- The imbalance between melee fighting styles
- The inadequacy of the thief
- The breakdown of the reaction roll mechanic
- The long-term supremacy of first-generation PCs, and resulting party cohesion problems
How are ability scores generated?
ACKS' default character generation process is 3d6 in order, choose a class that you qualify for, and then you can trade down non-primereq stats to boost your prime req, at a rate of 2 points from one stat for 1 point of a prime req, down to a minimum of 9 in stats traded down. There's also an optional rule, "Generating Multiple Characters", where each player rolls five sets of 3d6 in order and chooses one to play and two backups. We've always played with this optional rule and it does what it says on the tin - "gives players a variety of characters to choose from." It's also fun, if a little time-consuming. As a side effect, though, it raises mean prime requisite a lot.
Consider: you're starting a new campaign. Four players players show up and are rolling characters. You are probably going to want a mage PC and a fighter PC, classes for which high prime requisites matter a lot. If you each roll one set of stats, the best Int at the table will be (in expectation) right around 14 - enough to get +5% and one bonus spell known. The best Str at the table will also be around 14, for +5% fighter XP and +1 to hit and damage.
For a single-prime-req class, there are five other stats, distributed independently, each of which has a 23% chance of being an 11 or 12 (which can be traded down without reducing a bonus). So we should expect 1.2 points of obvious / no-real-cost prime req boosting via trading down. So the best PC fighter in an "average" party will have a Str of 15 and a +1 bonus in one other stat that he might trade down to bring it up to a 16 for +2 to hit and damage and 10% XP (and, incidentally, among his other four stats, probably 2-3 +0s and 1-2 -1s).
In the case where the optional Generating Multiple Characters rule is in effect, each player rolls five sets of stats, for 20 sets total. In expectation, the maximum strength rolled is something around 16.3, before any tradeoffs are made. There is a 9% chance that the highest Str statset has an 18, and 23% chance that the highest statset has a 17. If tradeoffs are in effect, there's around a 33% chance that the highest Str set can be made into an 18 without having to compromise any other bonuses. If you're willing to compromise other bonuses, you have about at 86% chance of a 15 or better, which you can probably turn into an 18.
OK, fine, rolling more stat sets gets you higher scores. Nothing surprising there. The trouble is that really high peak stats mess some parts of the game up.
Melee Fighting Styles
Previously, I did some analysis of the comparative power of the three melee fighting styles: sword-and-shield, two-handed weapon, and two-weapon fighting in ACKS. The results supported my players' general impression that sword-and-board is superior under most circumstances across the entire level range. Two-weapon fighting has better cleaving through masses of weak opponents at high levels, and two-handed weapons have a better chance to one-shot and be able to cleave through 2HD foes at low levels, but the superlinear utility of increased AC outstrips these benefits most of the time.
That analysis was conducted under the assumption of 16 Strength for the fighters. With 18 Str, shields are even more dominant by that set of metrics. Increasing strength is an effective increase in fighter level (since the main benefits of leveling under consideration are increased to-hit and damage), which tends to favor shields because the marginal utility of a point of damage decreases as you get more of them. At 16 Str, going from 1d6+3 damage to 1d10+3 damage increases your hobgoblins-per-round from 0.34 to 0.44; 10 percentage points and about 33%. At 18 Str, going from 1d6+4 to 1d10+4 only increases your hobgoblins-per-round from 0.44 to 0.5, 6 percentage points or 13%. These numbers are before cleaving is taken into account, but look pretty similar - the gain in killing power with cleave from the d10 weapon is still around half as good with 18 Str as with 16 Str.
While running these analyses, I discovered a magical place at the other end of the Strength scale, where sword-and-board and two-hander are almost perfectly balanced. That place is Str 10 against orcs, with a +1 Fighter Damage Bonus. Two-weapon fighting still doesn't hold up.
Reducing maximum prime reqs solves a limited subset of the Fighting Style problem, at low levels. At high levels, fighter damage bonus and magic weapons fill the role of 18 Str in minimizing the benefit of two-handed weapons compared to shields (which also get better with level due to magic).
Why do thieves suck? Who ever thought this class design was a good idea?
My understanding is that, under OD&D rules, thieves were not that much worse in melee offense than fighters. At that point in the development of the game, Str did not give you a bonus to hit and damage - it served as the prime requisite for fighters, for an XP bonus, and did little else. With no fighter damage bonus and no Str bonus to damage, the thief was inferior to the fighter in melee offense only by dint of a slightly weaker THAC0 progression, which was partly made up for by backstab's to-hit bonus (and both could use magic swords for slowly-scaling damage bonuses, which put them ahead of clerics in offense). The relationship between the OD&D thief and the OD&D fighter was much closer to the relationship between ACKS' assassin and ACKS' fighter.
But with fighter damage bonus and Str to hit and damage and PC fighters able to trade down other stats for higher Str, ACKS' thieves are left in a pretty marginal position for melee. The higher the available Strength scores, the bigger that gap becomes. As with fighting styles, fighter damage bonus is also partly to blame here.
The reaction roll system is pretty easy to break in ACKS. A 1st-level bard who puts his tradeoff points into Cha can take Diplomacy as his 1st level general proficiency, Mystic Aura as his first class proficiency, and probably pull +6 or +7 on most reaction rolls, at which point almost any intelligent creature encountered in the dungeon will be at worst indifferent. Problems of this form are well-known on the ACKS forums, and have led to a wide variety of proposals, including switching to 2d10 for the roll, codifying per-monster reaction roll modifiers so that monsters that are "supposed to" be hostile are more likely to be, enormous tables of situational modifiers, and so forth.
But you know, 2d6 would be fine if 1) 18 Cha were rarer, and 2) the proficiency bonuses were maybe +1 instead of +2 and didn't stack. Fundamentally this is the same problem as ability score generation with tradeoffs - we're taking a gaussian-esque distribution and adding a constant to it, resulting in a shifted distribution where extreme results are much more common than desired. The solution is to reduce the impact of the constants being added (either by increasing the impact of randomness through bigger dice, or by reducing the constants themselves).
These methods of stat generation provide a big edge for first-generation characters over later-generation characters, and also for rolled characters over henchmen who might be promoted to PC status.
What happens when a new player joins a game where the party generated stats using the "multiple characters" method? He rolls five sets. Among those five sets, the highest Strength will in expectation be only 14.5. The same is true of Intelligence. After trading down 11s, he may have a 16. Regardless of which prime req you choose, in expectation yours will be lower than the highest one in the existing party, and you're stuck playing second fiddle (for classes whose prime req matters, ie everyone but clerics). This analysis does make the assumption of efficient allocation of high stats by the Old Party (ie, you didn't roll all your high stats on one set, and not one player rolled all the high stats), but in practice that seems to happen. The exponential structure of the XP curve is designed to help replacement or new characters catch up with the rest of the party, but you're stuck with the lower ability scores for the entire life of the character. The situation is very similar with replacement PCs; if you rolled five sets, used the best one for your first PC, and got killed, your replacement sets are going to be weaker, and you're going to be behind on more than XP.
While rolled replacement characters are penalized by the "generating multiple characters" option, using henchmen as replacement characters is penalized by the trading-down rule. To get a henchman with an 18 Str, you need (in expectation) to survey about 216 henchmen. For campaigns in smaller markets, that is more potential henchmen than you will see over the course of the campaign. The prevalence of 18s in characters generated in a large group with trading down is just dramatically higher than the baseline. You are going to have a difficult time finding henchmen with comparable stats to replace them when they die, and most of the time using a henchman as a replacement for a first-gen PC will be a step down in terms of stats.
My players perceived this at a gut level before I did. This feeling further contributed to their love of shields and plate, because they recognized that a first-gen 18 Str fighter was unlikely to be replaceable. This supremacy of first-gen characters may also contribute to their intense love of Restore Life and Limb, which at least one of my players has commented negatively on ("People never stay dead, it's like friggin' superheroes.").
Roll three sets instead of five per player, maybe (maybe even just two sets per player). For a four-player party, three sets each gives you 12 sets, which puts expected max for any given stat just shy of 16. Then also get rid of the trading down rule and you should stop seeing too many 18s and the big first-gen advantage (15-16 Str is very doable for a henchman without trade-down). Maybe given new players joining an existing party more sets of stat rolls?
You're still going to see 16 Str fighters, where shield is superior, unfortunately. Maybe provide extra damage and to-hit bonuses to TWF / 2H with level? Maybe reduce fighter damage bonus to +1 at 1st, +2 at 5th, +3 around 9th. Maybe give thieves comparable damage bonuses, maybe get rid of both Thief and Assassin and build a Fighting 1.5 / HD 0 / Thief 2.5-ish class somewhere in the middle.
Cut the reaction roll proficiencies in half; Diplomacy as a general for +2 is just ridiculous. Do we really need five different proficiencies (Diplomacy, Intimidation, Seduction, Mystic Aura, and Bribery) for improving reaction rolls anyway? Get rid of all of them and make a monolithic Diplomacy prof that gives you +1 with no situational caveats and can only be taken once, like Leadership, and you're 90% of the way to fixing reaction roll exploits (... well, maybe we can keep Seduction; we are here for entertainment after all).
One thing that seems obviously tempting but that would be precisely wrong: Fighter Defense Bonus. If you give fighters a bonus to AC, that means you don't need the shield as much and 2H / TWF are more viable, right? No! The more AC you have, the more valuable each extra point of AC is. If anything, the right way to make shields weaker is to go full Viking Age and get rid of plate. Limiting magic armor and shields to +2 would help too.
See also, however, a follow-up post and correction