Running your first combat

Your first combat as a Dungeon Master will go badly in one of three ways. It will be too long, too easy, or too hard. The good news: all three failure modes have the same root cause, and you can debug them before your party ever sits down. This lesson is the debugger.

1. The 6-second turn (and why it takes 12 minutes)

Every round of combat in D&D represents six seconds of in-fiction time. Every combatant — every player character, every monster — gets one turn per round. On their turn they get a move, an action, possibly a bonus action, and they can take one reaction between turns.

That is the entire turn-structure rule. Read it twice. Run it once at the table, and you'll discover the actual rule: a six-second turn takes between 90 seconds and four minutes of real time, depending on what the player has to think about. With four players and three monsters, your "quick combat" is running on a 12-to-20 minute clock per round. A five-round fight is an hour and a half.

The widget below lets you feel the math. Toggle which actions a character is taking. Crank up the decision time. Watch the time-dilation factor (real-life seconds per fictional second) blow up to ×30 once people start juggling bonus actions and reactions:

Two takeaways here, and they will save your first session:

1. Pre-cap your encounter length, not your encounter difficulty. If you have a 3-hour slot and want time for roleplay, your combat budget is closer to four rounds than ten. Design for the rounds you have, not the HP pool you wrote.
2. The bonus action and reaction toggles cost more time than you think. Every time a player has both available, they have to weigh both, every turn. A monk with Flurry of Blows and a rogue with Cunning Action will each add a minute per round because the option space is bigger. That's not a problem — it's why those classes feel cool. Just budget for it.

Veteran GMs handle this with what looks like indifference: when a player says "uh, what can I do?" they answer in three seconds and move on. They aren't smarter. They've just internalized the menu of legal actions to the point that they can prompt without thinking. You can shortcut to that feeling by printing the action menu on a card and putting it next to every player's character sheet. We've put one at the bottom of this lesson.

2. The hit chance you didn't know you were designing

Every attack roll in 5e is a d20 plus a modifier against a target's Armor Class. The math is elementary, but most first-time GMs have never sat down and looked at the actual probability curves. The result is that they design encounters where the players miss two-thirds of their attacks and call it "challenging."

Here is the math you need to internalize:

• A natural 1 always misses. A natural 20 always hits and is a critical hit.
• The roll plus the attack bonus must meet or exceed the target AC. (Equal hits.)
• So the probability of hitting is P(hit) = (21 − (AC − bonus)) / 20, clamped between 5% and 95%.
• Crit is on a natural 20 only by default — a flat 5% — and crit damage doubles the dice but not the modifier.
• Advantage rolls two d20s and takes the higher. Disadvantage takes the lower. Both bend the curve in dramatic ways near the middle of the AC range.

The widget below is a live calculator. Drag the sliders. Pay particular attention to the shape of the histogram, not the average. Most people design for the average and then are surprised when the table experience doesn't match it.

The pattern that bites every first-time GM: they read that an Owlbear has AC 13 and the fighter has a +5 to hit, conclude "65% hit rate, the fighter will dominate," and then watch in confusion as the fighter rolls four 7s in a row. Two things are happening. First, 65% is the long-run average; over four attempts the chance of hitting at least twice is 85%, but the chance of hitting only once is still 11%. Combat is a small sample of a noisy distribution, and noisy distributions feel personal.

Second, the histogram has a long tail. With a 1d8+3 longsword on a normal hit, the most likely damage outcome is somewhere between 6 and 8 — but the second-most-likely outcome is zero, because the miss bar is huge. The thing the dice are doing is correct. The thing your gut tells you about averages is misleading.

The advantage trick

Advantage is the most powerful mechanical lever in 5e. Switch the explorer above from "normal" to "advantage" and watch the hit chance jump. The math is easier than it looks: if your normal hit chance is p, your advantage hit chance is 1 − (1 − p)². So a 65% chance becomes 88%. A 50% chance becomes 75%. A 30% chance becomes 51%.

The most useful intuition: advantage is worth roughly +5 to your attack roll near the middle of the AC range, and less at the extremes. If the fighter is hitting on a 5+, advantage barely helps. If they're hitting on a 15+, advantage is enormous. That's why environmental tactics — knocking enemies prone, creating cover, hiding before attacking — are the cheapest damage upgrades in the game.

3. The action economy is the whole game

If you only remember one phrase from this entire lesson, make it action economy. It is the single biggest lever you have over how a fight feels, and almost every "this combat felt wrong" diagnosis traces back to it.

The principle is brutally simple. Every combatant gets one turn per round. So a round contains as many turns as there are living combatants. If your party has four PCs and you put them against one big bad guy, the boss is taking 20% of the actions in the room. He could be three times as strong as the strongest PC and he'll still get ground down, because being outnumbered 4-to-1 is a worse position than being half as strong. This is why "lone boss" fights die in three rounds without legendary actions.

The reverse is also true and much more dangerous: a single player character cornered against four enemies is in a death scene. The dice cannot save them. Even at four-to-one against weak enemies, the math says the lone PC takes four times as much damage as they deal.

Drag the sliders below. Try the four presets. Notice how the projected action share collapses once one side starts losing combatants:

Three design principles fall out of this:

1. Bosses need either minions or extra actions. A boss without action support is a punching bag. The published 5e fix is "legendary actions" — extra turns the boss takes after other combatants — and "lair actions," which are environmental hazards that act on initiative count 20. Every published-tier boss has them. Most homebrew bosses don't, which is why most homebrew bosses are anticlimactic. If you're running a CR 5+ creature against a leveled party, give it 1-2 legendary actions and a lair action even if the stat block doesn't list them.
2. Hordes win by existing. Eight goblins are scarier than one ogre with the same XP value because eight initiative slots is eight rolls of the dice and eight bites at the bonus action / opportunity attack apple. The party's best defense is area-of-effect spells (fireball, shatter, even thunderwave at lower levels). The party's worst nightmare is a horde encounter where the wizard rolled bad initiative.
3. The first kill matters more than any other. When you remove a combatant from one side, you compound the action-share gap on every subsequent round. This is why focus-fire is the dominant tactic in 5e and why "spreading out" damage feels intuitively tactical but is mathematically wrong. The faster you knock things out, the bigger your action-economy advantage gets.

4. Encounter budget, and the multiplier nobody reads

D&D 5e ships with an encounter-building system in the Dungeon Master's Guide. About 80% of GMs use the first half of it (the XP thresholds per character) and ignore the second half (the encounter multiplier). This is why their encounters are always either trivial or accidentally deadly.

The system has two pieces:

• Per-character XP thresholds. Each character has four difficulty thresholds — easy, medium, hard, deadly — that grow with level. A 3rd-level character has thresholds of 75 / 150 / 225 / 400 XP. A 5th-level character: 250 / 500 / 750 / 1100. You sum across the party to get the party's thresholds.
• The monster-count multiplier. One monster counts at ×1 of its XP. Two count at ×1.5. Three to six count at ×2. Seven to ten count at ×2.5. Eleven to fourteen count at ×3. Fifteen-plus count at ×4. This is the action-economy adjustment expressed as math: more bodies on the field means more dangerous, faster than linearly.

That multiplier is where every "this should have been easy" TPK comes from. A party of four 3rd-level characters has a "hard" threshold of 900 XP. Four CR 1 monsters (200 XP each) is 800 base XP. Underfoot, right? Except: four monsters means a ×2 multiplier, so that fight is actually 1,600 adjusted XP. That's not hard. That's well past deadly.

Run the numbers yourself. Try the three presets — the tavern brawl, the lone boss, the skeleton horde — and watch the difficulty marker move:

A few patterns you'll notice once you've played with this for a few minutes:

• Lone bosses are almost always under-budget. A single CR 5 monster (1,800 XP) for a party of four 5th-levels (Hard threshold = 3,000) is just barely Medium. To get into Hard / Deadly territory with a lone boss, you have to climb to CR 7 or 8, at which point the boss can one-shot a PC with a normal hit. This is the "either trivial or instant-death" problem that gives 5e bosses a bad reputation.
• Hordes get expensive fast. Going from 6 to 7 monsters bumps the multiplier from ×2 to ×2.5 — a 25% jump in adjusted difficulty for adding one body. Going from 10 to 11 is another 20% jump.
• Mixed encounters cheat the system. Two CR 3 monsters (1,400 base) plus four CR 1/4 minions (200 base) is 1,600 base × 2 = 3,200 adjusted. The math reads "deadly," but the minions die in one hit and the multiplier instantly drops. Mixed encounters are self-balancing as the round-by-round body count changes.

What about the 2024 rules?

The 2024 D&D update changed encounter building. The new system drops the multiplier in favor of flat per-character XP budgets at three difficulty tiers (low / moderate / high). It's simpler, but it's also new — most published material, most third-party adventures, and the whole accumulated culture of "the budget says hard, why was it deadly" intuition is on 2014 rules. Learn the 2014 system first; the 2024 simplification will feel obvious once you understand what problem it was designed to fix.

5. The five rulings you'll have to make on Saturday

The other thing nobody warns first-time GMs about: combat is not a sequence of attack rolls. It is a sequence of rulings punctuated by attack rolls. A new player will, on average, propose something the rules don't directly cover every two to three rounds. Your job is to say yes or no in under five seconds, and the call you make will set a precedent for the rest of the campaign.

Below are four scenarios that will come up in your first three sessions, guaranteed. Pick how you'd rule. Then read what the table will look like a month from now in each branch:

The pattern across all four scenarios: your rulings teach your players what kind of table this is. If creative actions get rewarded, you'll see more of them. If rules-as-written is the standard, players will internalize the rules quickly and stop trying weird stuff. Both are valid tables. What's not valid is wobbling between them session to session — that's the recipe for a game that feels arbitrary.

The single best habit you can build, well before "knowing all the rules," is the narrate-the-math habit. When a player says "I rolled 14, hit?" you say "14 plus your 5 — that's 19 against AC 15, hits." Two seconds. No ego. It catches every arithmetic error before it becomes a problem and it makes the rules transparent to everyone at the table. It is the cheapest possible quality-control investment.

6. The Monte Carlo: stress-test before the table sees it

We've now covered the math, the rulings, and the design principles. There's one tool left, and it's the one that will actually save your first session: simulate the encounter before you run it.

The reason combat surprises new GMs is that they can't intuit the variance. They look at a stat block, do mental math on the average damage, and conclude "this should take three rounds." Then the dice produce a six-round slog or a one-round murder, and they have no language for why. The variance lives in the dice. The way to see the variance is to roll the dice a thousand times in your browser before your party rolls them once at the table.

The widget below is a Monte Carlo simulator. It takes your party's stat block (HP, AC, attack bonus, damage dice) and an enemy block, then simulates 1000 fights — real attack rolls, real damage rolls, random target selection — and shows you the actual distribution of outcomes:

A few things to try:

• Run the "Lvl 1 vs 4 goblins" preset. Look at the TPK rate. The "tutorial" encounter from the 5e Starter Set is genuinely dangerous on the dice. About 1 in 15 first combats end with at least one PC down. That's not a flaw in the design — it's load-bearing — but new GMs who think it's a freebie are in for a bad night.
• Take the Lvl 5 vs CR 3 wight preset and double the wight. Set count to 2 and re-run. The math is no longer in the party's favor. Now triple it. This is the cliff most first-time GMs walk off without realizing it.
• Take the goblins preset and give them −1 to hit. Watch the TPK rate collapse while the average rounds barely budges. To-hit modifiers are wildly more impactful than HP changes when you're tuning encounter difficulty.

Two important caveats. The simulator is intentionally simple: random target selection, no positioning, no spells, no terrain, no opportunity attacks, no morale checks. Real combat is dirtier — the wizard fireballs the horde on round one, the cleric heals the dropped fighter, the ranger picks off the stragglers. In practice the player party will outperform the simulator in the high-skill direction (because tactics) and the variance will be wider in both directions (because crits, focus fire, and bad rolls). Treat the numbers as a sanity check, not a prophecy.

Second caveat: the only number that matters is the distribution, not the average. If your encounter has a 60% chance of clean win, a 30% chance of "one PC down," and a 10% chance of TPK, that 10% will happen to you at some point in the campaign. Ask yourself whether you'd be okay running it the night it does. If the answer is no, redesign it. The dice do not care about your average.

7. At your table tonight

We've covered a lot. Here's the entire lesson compressed into a one-page cheat sheet you can keep next to your screen during the session:

The actions menu, if you want to print one for each player:

Where to next

You now know more about running 5e combat than 80% of the people who have ever sat behind the screen. The remaining 20% know it because they've run combat fifty times. The only way to close that gap is to run combat. The math will become muscle memory, the rulings will get faster, and the action-economy intuition will start humming in the background while you focus on the part that actually matters: the story you and your players are telling together.

Three things to do next, in order:

1. Run a one-shot. If you don't have one ready, our free Crypts of Ember Hollow is a 3-4 hour adventure designed for exactly this. CC-BY-4.0, no account, just a PDF.
2. Search the SRD as you go. Mid-session questions are normal. Looking up "grappling rules" by flipping through the PHB is not. Our free SRD search covers nine open systems, including the full 5e SRD, with full-text search.
3. If you want a co-pilot at the table — searches your campaign notes, listens to the session, suggests rulings live — that's what LorePanic is. Free tier uses your own ChatGPT Plus subscription, no extra cost.