Two Problems that Provoked Bounded Accuracy

One of the key design features of D&D Next is something the designers call bounded accuracy. Bounded accuracy reins in the steady escalation of bonuses to checks and attacks that characters received in earlier editions. I love bounded accuracy.

To explain my affection, I want to consider two problems with (nearly) unbounded accuracy in the third and fourth edition.

Third and fourth edition both assumed a steep and steady increase of plusses to your skill numbers as your character advanced. This rewarded you with a sense of accomplishment as you saw your character improve, but the increases led to problems at higher levels.

In third edition, at each level, characters received an allotment of points to improve selected skills. If you reached high level, and concentrated your improvements on the same skills, you gained huge bonuses to those skills.

The huge bonuses created a dilemma for dungeon masters and authors trying to set DCs for high level adventures. You could set very high DCs that challenged players who specialized in a skill. These DCs were impossibly high for non-specialists, so if the party lacked a specialist in a particular skill, the task became flat out impossible. Alternately, you could set low enough DCs to give non-specialists a chance, but these DCs grant the specialists an automatic success.  (Again, by specialists, I just mean a character who concentrates skill improvements on the same skill, not a super-optimized character.)

Third edition assumes that the DM will justify the sky-high DCs required to challenge high-level specialists by describing obstacles of legendary proportions. At first level, the rogue must climb a rough dungeon wall; by 20th level, he must climb a glass-smooth wall covered in wet slime—in an earthquake. At first level, you must negotiate with the mayor; by twentieth level, he’s king. And you killed his dog.

In the skill section of the third edition Epic Level Handbook, the epic-level obstacles become absurd. Here we find the DC for balancing on clouds, sweet talking hostile creatures into sacrificing their lives for you, and so on. I understand that some folks enjoy playing characters as mythic, godlike creatures, but to me, that game doesn’t seem like D&D anymore. Given the rarity of epic play, I suspect I stand with the majority.

Fourth edition tried to resolve the problem of high-level DCs becoming either impossible for typical characters or automatic for specialists. The system grants every character a flat, half-level bonus to checks. Now skilled characters maintained a flat +5 bonus when compared to their peers. Everyone enjoyed steady increases, but no one fell too far behind. This approach fixed the math, but when you compare characters of different levels, it defies logic and breaks your suspension of disbelief.

By level 10, a wizard with an 8 strength, gains the same ability to smash down a wooden door as an first-level character with an 18 strength.

“Wow, Wiz, have you been working out?”

“Thanks for noticing. My strength will be 9 soon.”

Of course, Wiz never gets a chance to show off his new prowess, because those DC 16 wooden doors have all been replaced by level-appropriate, DC 20 barred doors.

In truth, the players never really advance because they stand on a treadmill.

You can see the treadmill on page 126 of the 4e Rules Compendium in the “Difficulty Class By Level” table. Using this table, your character no longer gets better at easy checks, she just faces higher DCs. That table makes Living Forgotten Realms adventures work across entire tiers.

Fourth edition is inconsistent about whether the rising DCs in the “Difficulty Class By Level” table represent increasingly legendary barriers in the game world. For example, the DCs for breaking doors rise as the doors become sturdier. But social skills tend to be pegged to the DC-by-level table. (The system just assumes you killed the king’s dog.) The Living Forgotten Realms adventures used for organized play mostly abandon any attempt to flavor the rising DCs as increasingly legendary challenges. The challenges never change, just the DCs.

By reining in the scale of skill bonuses as character’s advance, D&D Next solves both problems. The system does not reward players with the same magnitude of improvements as their character’s advances, but the small improvements are real improvements, not steps on a treadmill.

While bounded accuracy solves problems, characters still need to stand out from their peers. A specialist should stand out and enjoy a chance to shine. The current ability bonuses are too small to achieve this. You can read my opinion on ability bonuses and checks here.

This entry was posted in D&D fifth edition, D&D fourth edition, Role-playing game design and tagged , , , , . Bookmark the permalink.

6 Responses to Two Problems that Provoked Bounded Accuracy

  1. Pingback: Living Forgotten Realms Battle Interactive | DMDavid

  2. Charles Brown says:

    Have you thought about using d100 in stead of d20 for roll checks? The problem with d20 is that it restricts DC to the range 10+1 to 10+20, which as you approach level 20 everything become easy which force the DM to inflate the DC even higher for high level adventures. The problem with the current DC scheme is that there is no room for challenge.

    DC has to be composed of two parts, the part that characters can aspire to beat and the part that is always challenging. So make the DC range 50+1 to 50+20+30 and use d100. The part that characters can aspire to beat is the +20 portion (assuming max level is 20), the challenging part is the +30 portion. Nothing but randomness can beat the challenging portion. I think this solves your problem.

  3. Charles Brown says:

    Oops, I meant to say make the DC range 50+1 to 50+20+80 and use d100. Wow, I think +80 is too much challenge. How about use d20 + d10 instead on DC 10+5+1+1 to 10+5+20+10, in which the +20 is the part characters can aspire to beat and the +10 is the challenging part.

  4. Charles Brown says:


    Suppose Joe Average is a Rogue with Dexterity of 10 and 5 ranks in lock picking. Trying to pick a DC 15 lock, Joe needs a roll of at least 10 on a d20 (Success rate of 55%) to succeed. With 20 ranks, Joe needs to roll at least a 1 (Success rate of 100%).

    Now replace the d20 with d100.

    With 5 ranks in lock picking, Joe needs to roll at least a 50 on a d100 (Success rate of 51%) to pick the same lock, although the DC is now 50+5. With 20 ranks, Joe needs a roll of at least 35 (Success rate of 66%).

    Using the d100 to make roll checks, there is no need to raise the DC from 55 to challenge Joe when Joe achieved 20 ranks. If 20 ranks is the maximum that Joe can ever achieve, then Joe would always be challenged by the lock. Yes it’s easier for Joe to pick the lock with 20 ranks, but he has legitimately earned the extra ranks. It’s petty to raise the DC simply because Joe has earned 15 more ranks.

    The main point is that Joe’s maximum of 20 ranks can account for 100% of d20’s possible outcomes but only 20% of d100’s. Joe would always be challenged if d100 is used to make roll checks. There is no need for reining in Joe’s achievement or for putting Joe on a treadmill.

  5. Jason says:

    my humble opinion is that ALL versions of D&D since 1st have become broken at higher level to different degrees of brokeness. In my own games, i usually cap character advancement at 12th and don’t let NPCs of note go past 15th. Character advancement is slightly slower, but it helps resolve high level problems. And frankly the high level abilities become too over the top for my taste anyway. Epic level play is ridiculous by any account unless you have a sourcebook that actually SHOWS you how to play as a demi-god with the appropriate world shattering plots, not just higher number.

    • DM David says:

      Hi Jason,

      You’re absolutely correct about high level play being broken across all editions. On the ENWorld forums, Gary wrote, “The OD&D game was written to challenge PCs from first through about 16th level.” I think Gary is a bit optimistic, and the practical upper bound stood closer to 12.

      Fourth edition ranks as the first edition to establish working math for accuracy across all levels. The designers still botched the high-level damage math, the expected number of powers for high-level characters, and admitted that they should not have increased the size of area effects as levels increased.

      The fifth edition designers promised that their game would not see print until they took the time to make high-level play as solid as lower-level play. However, the public playtest closed before the public even got a chance to vet the game’s final math, so I think the need to start making money led the team to conclude the playtest before they could prove that they reached their goals. If you gave Mike Mearls truth serum, I doubt would express total confidence that 5E achieves rock-solid high-level play. On the other hand, bounded accuracy and the rule for spell concentration should help address some problems that appeared in earlier editions.

      Prior to fourth edition, high-level spells tended to break the play experience before the game’s math broke. Gary reported that characters above 14th level rarely saw play in his own games. I suspect this enabled him to plant game-breaking spells like Wish at the top levels of the spell lists. Gary liked systematic organization, so he tended to assign spells for every magical effect, even those that broke the game. In his game, he could stash game-breaking spells at the unplayed, top levels and only make them available to players through scrolls and other consumables.

      Of course, once the game saw print, players saw no reason to limit themselves to level 12, so game-breaking spells caused them problems that never burdened Gary’s circle of players. Ever since OD&D, the game’s designers have had to choose between (a) honoring tradition and keeping game-breaking, high-level spells in the game and (b) dropping the spells and creating a game that feels less like D&D.


Leave a Reply