Difficulty
Learning paths through this topic
Read Your Game's Difficulty Curve From Player Data
Write down how hard each level should be, then compare it with first-try wins per player. A sudden fall against the level before is what you fix first.
Level Completion Rate: How to Read It
If 41 of 100 players finish a level, the true share is likely 32% to 51%. Split up the 59 who left, because each group needs a different change.
Retry Rate: Do Players Want One More Go?
Retry rate is the share of players who start again after losing. High means one more go, low means they gave up. Read it next to who wins in the end.
Where Players Die: Read a Death Map
Count deaths on a grid, but rank spots by how many different players died and how many passed through. Raw death counts point at the wrong place.
Difficulty curve
Your AI works out how hard each level really is from what players did, counting each player once, and compares it with how hard you meant it to be. It finds the level that is suddenly much harder than the one before, shows who tried again and who gave up, and says when there are too few players to tell.
Death map
Your AI lays a grid over a level, counts where players died and draws the result as a simple map, sometimes called a heatmap. It tells a real trap from one stubborn player who died there thirty times or a busy spot everyone passes anyway, and each hot spot comes with the numbers behind it.
Which attacks hit players before they could have seen them coming?
Some deaths feel cheap because the warning was shorter than any human reaction. This entry shows how to find which attacks do that, including the ones that gave no warning at all, so you fix the right animation.
How many tries does a boss really take if many players give up first?
Averaging the tries of those who won makes a brutal fight look friendly, because the ones who gave up are missing. This entry shows how to count the quitters honestly and where the typical player really stands.
Which phase of my boss fight actually kills players?
Counting where players die makes the first phase look deadliest simply because everybody passes through it. This entry shows how to read the attempts so each phase is judged fairly, and how to treat the attempts that won.
Check that automatic difficulty adjustment treats weak and strong players alike
A game that quietly adapts to the player should leave everyone winning about equally often. This entry shows how to verify that from results by skill group and what a wide gap tells you about the adjustment rule.
Which levels only look easy because most winners used a booster?
A fine-looking win rate can hide a level that nobody can beat on their own. This entry shows how to split winners by help used, which two shares to compare, and why you should not conclude that the helpers are to blame.
How many tries does a level really cost, counting those who gave up?
Telling a designer that people need about two tries is wrong if the people who needed twenty left long ago. This entry shows how to follow the whole group attempt by attempt so the answer includes everyone who quit.