How to Build a Balanced Encounter in D&D 5e (The Math, With Examples That Match the Tables)
CR, XP budgets, and the multiplier pitfall: encounter balancing by the book, by intuition, and by the method most experienced DMs actually use.
What the book says, exactly
Dungeon Master's Guide gives a mechanical procedure: add up the XP of each monster, multiply the total by a danger multiplier based on how many enemies fight together, then compare it against four thresholds per player level (Easy, Medium, Hard, Deadly). A party of four level-5 characters facing four CR 1 monsters: 800 XP x 2.5 = 2,000, which lands in Hard territory. That's the system, and at low-mid levels it roughly works. It fails in the places where you actually feel the pain — and the math explains why.
Why the multiplier is a landmine
The party-size multiplier assumes the monsters' action economy matches the count. Two level-5 dragons: correct multiplier. Twenty skeletons (the multiplier's idea of an encounter): absurd. At CR 1/4 x 20, the book says 1,000 XP x 2.5 = “Deadly”, but everyone who's run a skeleton wave knows it's the easiest fight of the week. XP is a proxy for action economy, and minions break the proxy. The experienced-DM fix: build by actions, not XP — count enemy actions vs player actions; at even numbers, the fight is roughly Medium; add one round of enemy burst damage to go Hard.
A method that survives contact with real tables
Build in three parts. (1) Threat: one creature that makes the players care (CR at party level, maybe two at level-7+). (2) Pressure: a small stream of minions, HP low enough that the wizard can clear two per turn. (3) Terrain: the third enemy — choke points, elevated archers, hazards. The XP of part 1 sits at your target difficulty, parts 2-3 tune the feel, not the math.
How it actually feels, by level
Level 1-3: one extra minion can wipe a party (HP 18, a fireball does the rest); treat 4+ enemies as Deadly even if XP disagrees. Level 4-10: the system's math is good; trust the thresholds and reserve multiplier skepticism for the swarm case. Level 11+: high-level casters have save-or-suck and the party stops fearing resources — go big, the encounter math is mostly theater at this point. If a fight feels unfair, it's almost never the monster CR: it's who goes first (see initiative-tracker for the part of this that's pure bookkeeping).
The honest numbers of running it live
A well-built encounter at Medium difficulty is two rounds where players think they'll die, one round where they think they'll die, then a turn where they win. If your fights last 8 rounds, something is wrong — usually HP inflation or an uncounted multiplier. And every round costs about as much real time as one rules question; if the table keeps stopping to ask, you don't need new monsters, you need an organized combat flow: fixed initiative order, conditions on sheets, and the rules at hand. The rules mirror exists so that question costs twenty seconds, not five minutes.
A worked example
Party of four, level 6, standard party composition. Threat: one Young Green Dragon (CR 8, 3,900 XP) with a lair and one round of surprise — that alone lands at Hard (the budget for three of you is ~2,300, so it reads Deadly until you remember the 1.5x multiplier for a single solo enemy, which drops the math rating; everyone's table instinct says Hard, so trust the instinct). Pressure: two Giant Poisonous Snakes (CR 1/4) as minions harrying the backline. Terrain: a flooded ruin that halves movement. The dragon acts first in group initiative about half the time, the snakes act third, and the terrain keeps the wizard honest. Total XP says 'Deadly' and the table says 'brutal but fair' — that gap is exactly where the fun lives, and it's why we tune by actions and terrain rather than XP alone. If you want to run this experiment without paper and a spreadsheet, the initiative-tracker keeps the queue and conditions honest.