System Scorn: The Excesses of 3rd Edition Dungeons & Dragons
or the deep reasons why 3rd edition broke like it did
Thank you to my lovely test audience for their feedback on this post: jedavis at The Wandering Gamist, Nathan Hardacre, Contentronix and others.
Introduction
One of the most hobby defining tabletop systems was Dungeons & Dragons 3rd Edition (D&D 3e), released in 2000. The Revised Edition (aka D&D 3.5e) was a clean-up of the earlier 3rd Edition books, mostly class balance changes (that, ironically, made some classes ridiculously strong) and the amount of skills being reduced.
It was the first edition after major video games using the Advanced Dungeons & Dragons 2e (AD&D 2e) ruleset were released, such as Baldur’s Gate (1998) and Icewind Dale (2000), and it had Neverwinter Nights (2002) using it as a ruleset.
The success of these games was a huge surprise to the developers, publishers and I expect Wizards of the Coast.1 Baldur’s Gate alone sold a million units within the two years before 3e’s release, when the publishers expected a maximum of 200,000 across its lifetime.
There was also a Dungeons & Dragons (2000) movie with Jeremy Irons as the principal villain. While it was absolutely panned and considered a box office flop, it still got around 7 million people to watch a fantasy film before The Fellowship of the Ring (2001) and Harry Potter and the Philosopher's Stone (2001) brought the genre into the mainstream.
Therefore, it should not be much of a surprise that the success of D&D 3/3.5e was absolutely massive. Roughly a third of current tabletop players were introduced to the hobby through the system, and that is including the mass influx of new 5e players.
The issues with D&D 3/3.5e have been discussed at length, but they have tended to focus on surface level troubles. Spellcasters are broken, clerics and druids are the spawn of Satan and the fighter is alone in the corner rubbing his longsword dreaming of AD&D.
In reality, these are symptoms rather than the ruleset’s actual problems.
The true problems were four (relatively linked) issues:
Splatbook Compounding - rules getting worse through splatbooks
System Intricacy - people make small changes that make big changes
Dungeon Master Burnout - difficulty of actually setting up a session
Forum Culture - the analogue world meets the digital
Observant individuals will quickly notice that at least three quarters of these are still unfixed within more modern rulesets. Pathfinder 1st Edition (PF1e aka 3.75e) suffered from these, as 5e does now.
1. Splatbook Compounding
(and how more options made the game worse)
The three core books of 3.5e (Player’s Handbook aka PHB, Dungeon Master’s Guide aka DMG, Monster Manual aka MM) were a relatively full and functional ruleset in themselves.
In the five years after the core release, Wizards of the Coast released over 50 rulebooks with classes, feats, items, spells and races. There are supplemental rules, additional rules and alternative rules spread across these books. The genre fiction starts to dry up after a certain point, but this results in beautiful abominations like Eberron.
This development increases the spread and availability of options so that players are able to break core assumptions of the base game, which results in unbelievably monstrous builds.
Ironically, the quality of the mechanics goes up as more books are written, the understanding of the system is deepened and the writers get more skilled.2
This did not really help.
The splatbook spread allowed the malignant flower of core rulebook problems to grow. It made the vast gulley between the best and worst optimised characters turn into a galactic trek.3
The solution given to these is always to ban specific spells, classes, races, or books, but after a certain point you may as well ban the ruleset and play a different game as many people chose to do.
I. Race & Class Options
In the core books, there were 7 available races. The bonuses available from these races were very limited. There were no Racial bonuses to Intelligence, Charisma or Wisdom, which means at least one less spell per spell level for casters. The skills and powers of these races were scattershot and tried to represent the genre rather than offering a specific package.
By 2008, there were over 140 races. They offered nearly every combination of stat bonuses possible. Some of these races offered bonuses far beyond those of the core races, occasionally without a required Level Adjustment.
In the core books, there were 11 core classes and 16 prestige classes. These classes were often very limited in their powers, and ultimately the multiclass mixing and matching was limited by what was available.
By 2008, there were around 150 core classes and nearly 800 prestige classes. Every possible combination of powers was available, with more than enough space to dip into a class every single level. Many of these classes were frontloaded with goodies, encouraging the pick-and-mix attitude towards levelling.
This growth of options and the laissez-faire attitude towards character coherency lead to unbelievably optimised builds that still worked within the bounds of the game.
II. Modifier / Bonus Types
In earlier versions of D&D, it was not clear if and how bonuses should stack.
Does a Ring of Protection stack with a Shield spell? How about an Armour spell? Armour and Shield are explicitly non-cumulative, does that mean everything else is?
The solution in 3.5e was to bring in modifier / bonus types.4 There are 18 types listed on the 3.5e SRD.5 Only the highest of each of these types applied, the lower modifiers were ignored until the higher level effect expired.
Therefore, you could have both an Deflection Bonus and Dodge Bonus from two separate effects, but could not stack two Deflection Bonuses at the same time.
In addition, there were untyped bonuses that explicitly stacked without limit but these were, by design, few and far between.
For the three starting books, this was a very clever solution. The amount of bonuses was limited by what was available. You may be able to get an Insight bonus to your attack throw, but then you would not be able to have a Racial bonus to add onto that as well. Enhancement bonuses were a dime a dozen due to their use on magic items and that meant spells that gave them were only useful in rare situations.
Then came the splatbooks. Players began hunting for and stacking different bonuses from every source they could find.
The common Diplomacy skill exploiter (aka Diplomancer) stacked synergy bonuses from skills, Competence bonuses from a Circlet of Persuasion, the Marshal class Motivate Charisma aura, the Racial bonus from half-elf and so on and so forth to have a +15 Diplomacy skill bonus at 2nd level.
III. Spell Availability
A big critique of 3.5e was that spellcasters were extremely dominant in comparison to non-spellcasters (aka martials). This is absolutely true, and was true from the core book release. In fact, some of the most broken spells are from the PHB itself.
This balance issue is still solvable, or reducible in how much impact they have.6 It is possible to reduce the number of castings per day or increase required rest time or just do errata to balance the PHB spells.
What is not solvable is the way having a huge amount of spells leads to this power steadily increasing without the same happening to martial characters. While the race and class options could increase martial power, they were still capped by the maximum power available to their level.
Every single spell added increased the power of spellcasters by a tiny multiplier.
Mostly this was because there was no reason to recreate spells from earlier books. If an earlier book had Iceblast7 then you need to make Vitriolic Splash8 in the new book. And, suddenly, you have three potential spells at 3rd level that cover off major resistance types.
Sorcerers and Wizards (amongst other Arcane casters) both gained from the increase of spells, but Sorcerers were limited by their very small repertoire and Wizards limited by what scrolls the DM doled out to them.9
The main beneficiaries were Clerics and Druids1011 (and some other Divine casters) who could, as per rules, cherry pick spells to prepare every day from every potential spell available to them from the books they had access to.12
There were 25 1st-level Cleric spells in the PHB. By 2008, there were ~125. Assuming that each of those spells multiplies the power of the Cleric by 1%13 then a Cleric with access to all of the splatbooks would be 250% more powerful than a Cleric with none.
2. System Intricacy
(and the tendency for people to ignore it)
System intricacy is how interlinked systems are, and how much a minor change can tesselate through the rest of the system.14
D&D 3.5e was actually a beautifully developed system, with knowledge of the issues of 3rd edition allowing the refinement of its development.
But it was highly intricate, and small changes could have massive outsized effects later on. Groups had a tendency to ignore elements that they decided were overly complex, even though those elements were key to the system.
A lot of the complexity in the system was in character building and development. 3.5e had a relatively standard race and class based system, where players chose a race and then chose and levelled up through their core class.
Players could choose more than one core class which was called multiclassing and / or could choose specialised secondary classes called prestige classes.
I have three examples of this being the case, but I reference some more in the Forum Culture section.
I. Milestone Levelling
3/3.5e assumed that groups were using XP tracking based off defeating monsters for the purposes of levelling. However, as the edition rolled on there was a general trend for groups to use milestone levelling instead.
Milestone levelling is a system where the DM decides how fast characters level based off of story advancement. The players complete a story arc or similar, and they level up.
This system started to be heavily used during 3/3.5e, due to the wild complexity of calculating how much experience a party should receive from a combat. The Challenge Rating (CR) system was a comparative system where monster difficulty and the party’s overall party were taken into account.1516
The balancing factor for multiclassing was reducing that character’s XP gained by 20% for each class that was more than one level away from the highest levelled class.17 This could, if a player was foolish, absolutely cripple a character, but at least it was an attempt at a balancing mechanism to stop these kitchen sink style characters.
Groups often accidentally dispensed with this rule by moving to milestone levelling. This lead to characters being able to multiclass dip without any true penalties.
Multiclassed characters tended to be stronger than single-classed characters when both were built properly.18 This lead to situations where a 20th level Fighter was weaker than a 20th level mish-mash of classes. (Ironically, both were weaker than a 20th level single-classed spellcasters, as I mentioned in the Splatbook Compounding section.)
Milestone levelling also caused issues with the magic item crafting rules, where characters were supposed to spend experience to craft items. With experience having no real impact on character power, players started crafting whatever they liked.
II. Prestige Class Special Requirements
Prestige classes were 3rd edition’s answer to AD&D 2e’s kits. They were a method to specialise and personalise a character, and had good rewards for players that put in the effort to meet their requirements.
The 3.5e edition DMG had the first 16 prestige classes in it, and has the following text:
“Prestige classes are purely optional and always under the purview of the DM. We encourage you, as the DM, to tightly limit the prestige classes available in your campaign. The example prestige classes are certainly not all encompassing or definitive. They might not even be appropriate for your campaign. The best prestige classes for your campaign are the ones you tailor make yourself.”
Despite this, prestige classes were rarely treated as optional, and the concept of a DM making their own prestige classes is ridiculous since there are only the most basic guidelines to modifying core classes.
Prestige classes nearly always had three or more of the following requirements:
Base Attack Bonus: Essentially how strong of a fighter the character was
Skill Ranks: How good at a skill the character was
Feats: Special abilities the character has picked up on their levelling journey, which are very limited
Spell Level: The ability to cast Arcane and / or Divine spells of X level, sometimes with specific spell requirements
Special: Roleplaying or gameplay achievements that are needed to enter into a specific prestige class
From these, special requirements were almost always ignored.
Special requirements were the road block to stopping players from taking as many prestige classes as they wanted to. Even if the prestige class was allowed, the DM or the setting could prevent prestige classes depending on those characters not completing those requirements.
There were two big reasons that the special requirement was ignored.
A good number of special requirements would have involved solo sessions with the character, or would not organically come up in the progress of the campaign.19
Secondly, the power of a character build was often tied extensively to being able to obtain the relevant prestige class. Not obtaining the relevant prestige class would cripple that character, and infuriate the player behind it. As this would be a major social faux pas, DMs tended to avoid doing it.
The problem from this was that ignoring the roleplay requirements allowed the worst bits of the prestige class system to bloom. A player could find front-loaded prestige classes and use those classes to unlock other prestige classes, without the need to meet roleplay requirements.
As with all these issues, this was exacerbated by Forum Culture, which I will go into in the relevant section.
III. Carrying Capacity
One of the most hated parts of many tabletop systems is the concept of carrying capacity. People complain that it is onerous, often complicated and usually to little gain.
They are two-thirds right, and 3rd edition was no different.
Carry capacity was a factor of the character’s Strength multiplied by their size. Once the character reached their medium load20 then they were slowed down, had their maximum Dexterity bonus reduced and took an untyped21 penalty to a good amount of skills. Abilities that were affected by medium or heavy armour were also affected by a medium or heavy load.22
Carry capacity was a very clear balancing factor.
It was also maligned and ignored.
Considering a lot of the lot of 3rd edition’s combat revolved around grid based movement, having your speed reduced was a massive constraint. The maximum Dexterity bonus could reduce your survivability massively if you hit heavy load.
Despite this, players functionally played as pack mules, with an inventory more full of goodies than your standard piñata. DMs ended up trying to bring in other balancing factors such as only being able to retrieve a certain percentage of items, rather than simply playing the rules as written.
3. Forum Culture
(or what happens when the analog meets the digital)
In 200023, at the release of 3rd edition, there were around 300 million internet users or roughly 5% of the world population. By 2003, with the release of the Revised edition, there were 600 million internet users or roughly 10% of the world population.
By 2008, there were 1,800 million internet users (1.8 billion) or around 25% of the world population.
The growth of the internet was exponential, and somewhat unexpected. It went from Usenet boards utilised by geeky Americans to accessible forums to hosting every piece of information within the space of a decade.
But what really matters is how this affected D&D 3.5e.
Broadly, the system was not designed for the level of community and network effect that would arise from the internet uptake, and the greater community was not inoculated against these effects in the same way it would be for later systems.
There are a lot of ways that this represented itself but there were three main ones.
I. Theorycraft
Theorycraft is discussing and mathematically analysing the different options available within a game system. As a concept, this was nothing new, but the ability to instantly communicate with a large like-minded community was very fresh.24
Through this, forums inadvertently encouraged theorycraft. At the average roleplaying table in 2003, there was often only one person who was actually interested in pushing their character to its absolute limit. The ability to share a hyper-optimal build with individuals also interested in the breaking of the system gave a more positive response than they would otherwise have received.
3rd edition, as a system, was not made to take the pressure of tens of thousands of gamers analysing every element of the game to find the optimal choice. It didn’t help that a good amount of the theorycraft was done in the void25 when the assumptions made would likely break down at a table.
The system was made for a small group of people (and, sometimes, their larger community) to play it using the books they had to hand and allow the natural variation from that small subset to give slight perks to optimisation.
Instead, every potential abuse available due to the Splatbook Compounding issues was put into play, and the results from that are shown by people’s opinions of 3.5e.
II. Cultural Homogenisation
One of the outcomes from the access to a shared community (as opposed to the fractured individual city-communities before the internet) was that there was a huge push for cultural homogenisation, even when that cultural homogenisation was harmful to the system.
The reasons for this are the same as the reasons it happens to every online culture. Moderator oversight and the development of in-cliques meant that their cultural beliefs were pushed over and above others.
An example mentioned earlier is Milestone Levelling, which is now firmly embedded in 5e’s D&D culture. Milestone Levelling had a heavy narrative focus,26 and this lead to the trad culture27 that is very dominant today.
Another is the tendency to believe that wealth in 3rd edition was fungible and could be used to purchase magic items.28 This contributed to the problems mentioned in Splatbook Compounding.
Finally, there is the 15-minute adventuring day and the cultural disgust towards random encounters and wandering monsters, which Justin Alexander goes into detail about in his article on “The Death of the Wandering Monster”.
Ultimately, all these community culture decisions lead to a gradual degradation of what actually held the system together, even as the splatbook expansion lead to the Splatbook Compounding problems.
III. Digital Piracy
Digital piracy29 really started to come to the fore shortly after the release of 3rd edition. Limewire emerged in 2000, and the BitTorrent protocol was released in 2001. Both of these made the relatively arduous task of file sharing much easier.
Scanners were also becoming much cheaper and higher quality, allowing people with time and ability to build and upload full online books.
From that emerged the ability to access nearly all the splatbooks without cost and to cut out snippets without context. This massively contributed to Splatbook Compounding and theorycraft problems.
There is not a lot more to say on the piracy issue, except to note that this was a pretty rapid technological shift and it was completely unprecedented. Wizards of the Coast actually sold digital files early on, but they were scared off by the rapid sharing of those files on web services.
4. DM Burnout
(the difficulty of actually setting up the game)
By now, the problems with running 3.5e are likely obvious.
If you have set your group up, you will almost certainly have:
A list of forbidden options as long as your arm
Somehow still a huge gulf of power between your characters
Players arguing that your rules decisions are wrong due to interpretations they saw on the internet
These three things alone should be enough to understand why 3.5e was a time of massive DM burnout, but there are other crucial issues with the system that contributed.
I. Character Power Divide
As mentioned in Splatbook Compounding and Forum Culture, characters drifted apart in power due to a combination of player skill at building characters, whether they had access to the internet and how many splatbooks they had access to.
This resulted in characters with vastly different power levels.
In combat, the action economy restricted the effects of this somewhat at the earlier levels but, after 6th level, the game quickly spiralled out of the DM’s control. At this point, some characters have much more power and options than others and some players will feel their input is worthless.
II. Official Content Trivialisation
The optimisation of player characters meant that official content was quickly trivialised.
How is it possible to build a challenging, but not unbeatable, module or adventure when the designer doesn’t know if Bob the Fighter or Bob the Superman is going to go through the door?
It was not.
Combat was pathetically easy or ridiculously brutal.
Any DC was easily destroyed by optimisers or unable to be matched by less internet savvy players.
Then you have Diplomancers mindbreaking NPCs, Wizards perfectly lockpicking anything they came across and Clerics turning hordes of enemies into Swiss cheese.
And probably two thirds of your players got to watch as the final player demonstrated the results of their ability to use Yahoo to find the best builds.
III. Content Development Difficulty
The final nail in the coffin for DMs in 3rd edition was the difficulty of building content for the system, due to the complexity and intricacy needed.
The perfect example is making NPCs.
Enemies in 3rd edition were designed to be built similarly to player characters, having attributes, skills, feats and other complexities.
As an example, there were unified attributes shared across both monsters and players, with the attribute scale being used as an absolute metric.
Therefore, a standard 1st level player character could have a Strength score ranging from 3-20, whereas a standard ogre would have a Strength of 21.
This score would then affect the ogre’s skills. The ogre’s attributes and feats determined what feats it could take. Those feats would then feed back into the skills, the HP and everything else.
To build it properly and to make it difficult, you will have to reference a dozen books and use those books to optimise your ogre, and that is without including the class levels that enemies could take.
The issue was that this both made referencing enemy stat blocks at the table cumbersome and made building enemies take a long amount of time.
And therefore this is now an edition where the official content is useless and trying to build content to actually challenge your own players will take every spare bit of time in your week.
Final Word
There is a lot of good in 3rd edition, and there is a lot of bad beyond even that what I have mentioned so far.
What is most interesting is how the reactions to 3rd edition’s flaws differed. Some designers tried to fix the core system, some moved away from it, and some pretended like these flaws did not exist at all. There is a potential second part to this article demonstrating this depending on the response.
A lot of gamers are probably looking at the systems they play and wondering why these issues still exist.
In my opinion, these flaws are contributors to a very good business model:
Splatbook Compounding - encourages players and DMs to buy more books to get access to the multiplicative power effects
System Intricacy - trying to change the system yourself will break it, so you should buy official content
Forum Culture - tabletop roleplaying games are heavily dependent on the network effect and encouraging forum culture is a positive
DM Burnout - burnt out DMs can be sold quick solution packages to fix their problems or burnout instead of developing it themselves
I cannot comment as to whether this is intentional or not, as it is entirely possible that companies have been aping the success of 3rd edition for many years without truly understanding the fact that these choices are what makes the system start to fall under its own burden.
Nonetheless, it is important to remember that ultimately major tabletop companies are businesses.
Their goal is to encourage the customer to spend as much money as possible for as long as possible. Anything else is a failure on their part.
And for a number of companies, it seems they make a problem only to sell the customer the solution.
My question to tabletop gamers is:
Why give money to a company that does that to you?
As talked about in Matt Chat 255: Feargus Urquhart on Baldur's Gate, Shattered Steel, and Fallout.
There are rightful arguments that the options within the DMG are very strongly broken. However, these problems are easily fixed, unlike the things I am addressing here.
Monte Cook (one of 3rd edition’s main authors) wrote a blog post about the maligned “Ivory Tower Game Design” of 3.5e, where he claims that many of the balance issues were intentional. It is notable that this was written in 2005, after years of the internet absolutely destroying any semblance of balance in the game. Personally, I think this is a rationalisation of the issues rather than deliberate choices. Also, Monte Cook dramatically misunderstands what “Timmy cards” are in the blog.
Modifier and bonus are used interchangeably, e.g. Armor Bonus and Profane Modifier. I think that bonuses are always assumed to be positive but at least one (Racial Bonus) is not always going to be the case.
There are a few extra types accidentally introduced through poor editing, such as Complete Champion (2007) having a Divine modifier instead of a Sacred modifier.
See Pathfinder 1st edition, which had Wizards very close to 3.5e’s but with some balance changes. They became broken after enough splatbooks came out to give them a variety of spells but the majority of their strength there was Summon Monster abuse.
Not the delicious slushy, but functionally a frosty Fireball spell.
Which is an acidic Iceblast spell.
Or, as is mentioned later in the Forum Culture section, through the assumption of wealth fungibility allowing purchase of scrolls from Ye Olde Magicke Shoppe.
Druids were buffed going from 3e to 3.5e. I have been informed that this was due to druids being tested only as casters with animal companions and occasional shapeshifting, rather than the transforming monster that their builds usually were.
Druids also benefited from every sourcebook that dropped as they had access to a lot of the creatures inside it for transformations, therefore being able to pick and choose what ones were best.
In my opinion, this was a major contributor to the build group called CoDzilla (a clever Godzilla pun), which was famed for being able to cast better than a Wizard, fight better than a Fighter and be hated more than a Rogue.
Not a terrible assumption. A lot of spells were terrible, but standout spells multiplied the power of Divine casters by much more than a single percentage point.
In software development, this would be called modularity.
In earlier editions, monsters gave the same level of experience regardless of character level, but levelling XP doubled per level therefore giving an exponential curve. This meant that fighting weaker enemies would take a ridiculous amount of time to level. In 3rd edition, this curve was softened slightly which meant the CR system had to come into effect.
There is an encounter difficulty calculator on the SRD here. Imagine trying to do that either by hand or, even worse, at the table.
See here. There are a few ways to mitigate this penalty, including favoured classes. Note that it is not clear whether this penalty is multiplicative or additive, i.e. whether the penalty for 5 classes outside of range would be ~30% or 0% XP gained.
This is due to an interpretation of the class levelling tables that I could spend an entire article on. Classes tended to have good bonuses within the first few levels, so people believed that these would stack instead of classes counting as combining their levels. Whether this is the correct interpretation was a huge (and unsolved) debate.
Such as the “Beast Heart Adept” from Dungeonscape (2007), who had to, “Spend at least one week living alone in a dungeon complex.”
More on Modifier Types later.
I think there is an argument to be made that spellcasting should have had a chance of failure from a medium or heavy load, since wearing armour gives clear Arcane Spell Failure boundaries.
Shortly after the Millenium Bug did not occur.
Theorycraft supposedly has its origins in the Starcraft (1998) community, but Urban Dictionary does not list it as a term until 2005 so I feel that may be slightly apocryphal.
Assume the elf is spherical.
In terms of the GNS theory of game design.
According to the Six Cultures of Play.
Unfortunately supported by every magic item having something approaching a price listed against it.
You wouldn’t download an orc girlfriend.
Because I prefer that problem over OSR's "Reprinting B/X over and over again." I'm also a big proponent of mechanical innovation in game rules,--RPGs are woefully underdeveloped compared to even video games
Did you know about Vampire Survivors? It's basically a sort of reverse bullet hell shooter. But the most Important part is that it's designed like it's a slot machine(because it's designed by someone who worked on those) with all the lights and sounds, the game is basically replacing the need for money with reflex and time.
It is massively popular and has basically spawned it's own genre.
Point is, the problems you've said is easily just as much of an appeal as it'd be something that looks undesirable.
Excellent article overall. I was introduced to DnD via 3.5 and as a result I have a deep and unshakable nostalgia for both it and Pathfinder 1e (which my group eventually moved to soon after the release of 4e). I tend to go through cycles of "ugh, this busted-ass system, how did I ever like it" and "you know, actually maybe that system wasn't too bad, maybe I should shine it up and give it another whirl" and your post really sums up a lot of the ups and downs. Playing and running 5e has given me a lot of appreciation for various design details, successes, and mistakes from all sorts of previous editions and makes me realize where my problems with modern 5e design came from. So many of the problems in modern DnD come from a core of a distaste for restriction and a demand for options, both of which are fantastic in measured doses and catastrophic when cranked up to 11. I think 3.5 has way too many options and a whole lot of restriction whereas 5e has not so many (but still a bit too many) options and not nearly enough restriction. Discarding XP as a whole, ignoring carrying capacity, and treating the magic item section of a DMG as a buying catalogue has certainly changed what players expect from a game called "DnD", and we could argue whether that's a good or bad thing till we're blue in the proverbial face, but just knowing about it and puzzling over it in my free time will teach me a lot about the hobby I love, I think. I appreciate you putting in the work to voice all of these thoughts and theories and ideas; stuff like this tends to stick with me while I'm constantly on the hunt for that Ideal System that doesn't exist (because the Ideal System is literally the friends we made along the way).