Too hard, then too easy: difficulty curve game design with time-to-kill targets and three more tricks

A game that’s too hard in one place and too easy in the next usually has its enemy numbers set by guessing. Someone typed in a health value that felt right in one test, and then the player got a new sword. Difficulty is mostly two numbers per enemy: how long it takes you to kill it, and how long it takes it to kill you. Plan those, then tune by measuring. Slay the Spire’s team tracked “how much damage players using a given card took, on average, from a particular enemy”, and used it to retune fights (Game Developer).

Each demo below plays the same fight twice: hand-guessed numbers on the left, tuned numbers on the right. Under each fight is a small live chart of the numbers behind it, so you can see the difference as well as feel it. Point at a bar or dot for the exact value. Where there’s a Play it yourself button, you can take over: tap or press Space to swing (or block, in demo 3).

1. Plan a power curve

Write down how strong the player should be at each point of the game: damage per second (DPS) and health, in a small table. Then read enemy stats from that table, not from the player’s actual level. If enemies grow with the player, a new weapon changes nothing. Oblivion scales enemies to the player’s level, and GamesRadar calls it a “disappointing” part of the game that the remaster left in (GamesRadar). In our demo, you pick up a sword that doubles your damage. On the left, the slimes level up with you to 30 HP and still take 1.5 seconds each. On the right, their health comes from the plan, so they die in 0.75 seconds.

2. Aim for a time to kill

Give each type of enemy a target time to kill at the player’s expected damage. A good start: fodder 0.5 to 1 second, regulars 2 to 4, elites 8 to 15, bosses 60 to 180. Then health = expected DPS × target time. Designer Alex Trowers works out “how long it would take the player to kill the monster” the same way in his tuning spreadsheets, and adds that “it’s no replacement for playing the damn game” (Alex Trowers). In our demo, three lanes fight at 10 damage per second. The guessed health values give 3.5 s for fodder, 1.25 s for the regular and 22 s for the elite. The formula gives 0.75 s, 3 s and 8.5 s, each inside its band on the chart.

3. Aim for a time to die

Now the other direction. Decide how many hits from each enemy should kill the player: about 6 to 8 from early regulars, fewer later on. Then damage = expected player health ÷ hits to die. Trowers’ spreadsheet has the same line: “timeToKillPlayer = playerHealth / monsterDamage” (Alex Trowers). In our demo, an early orc and a late brute attack you. On the left, the orc kills you in 3 hits, with no room to learn, and the brute needs 20, so it’s no threat at all. On the right, it’s 7 and 5. Tap when you see ”!” to block.

4. Check the whole curve for spikes

A single fight can be on target and the game can still have a wall in it. For each section, add up enemy health ÷ expected DPS (the seconds of fighting) and plot it. It should rise in waves, with a rest after each peak. Flag any section more than about 1.5× harder than the one before, and any long flat stretch. Slay the Spire’s team found one fight, The Awakened One, too hard for some decks from their data, retuned it, and then saw “more players beating the boss” (Game Developer). In our demo, a run through 8 sections plays fast-forward. On the left, section 4 is 2.6× harder than section 3 and kills you, and sections 5 to 8 are flat. On the right, it rises in waves with rests after the peaks.

5. Double it or cut it in half

When a number is wrong, a 5% change is too small to feel in a playtest. Change it by 2× or ½, play, and then close in from both sides. Soren Johnson calls this one of “Sid’s Rules”, after Sid Meier: instead of lowering a weak unit’s cost by 5%, double its strength. The original Civilization’s map was cut in half to fix slow play (Designer Notes). In our demo, an orc takes 8 seconds to kill and should take 3. On the left, five 5% nudges still leave it at 6.19 seconds. On the right, halving twice and then meeting in the middle lands on 3.0 seconds in 3 changes. Press Play it yourself and try to reach the band with each method.

Targets are a starting point

The numbers in this post are a place to start, not a rule from any one game. Use them like this:

• Measure, then play. Run a bot or a scripted fight at the expected power against every enemy type and section. Real time to kill and time to die should land within about 25% of the target. Then play it, because the spreadsheet can’t tell you if a fight is fun.
• Tune for your genre. A fast action game wants shorter times to kill than a slow RPG. Move the whole table together, not one enemy at a time.
• Keep the rests. A curve that only goes up is tiring. A quiet section after a hard one is part of the curve.
• Planned difficulty first, adaptive difficulty second. If you also want the game to adjust to each player, start from a good fixed curve. See adaptive difficulty and rubber-banding.

If hits feel spongy even when the times are right, look at the numbers the player sees: combat numbers and the damage formula.

Add it to your game

1. Copy the skill at the end of this page into your project.
2. Ask your agent: “My game is too hard in some places and too easy in others. Tune the difficulty curve.”
3. Play it, then tune in plain words: “The elites in the second area still take too long to kill.”

Art in the demos: Tiny Dungeon and Kenney Mini by Kenney (CC0).