Playtesting
Learning paths through this topic
Some players solve my puzzle in fewer moves than I thought possible
Solutions shorter than your own best are either a gift or a leak: a clever path you never saw, a shortcut that breaks the puzzle, or a mistake in your own number. This entry shows how to tell the three apart from the logs alone and what to do with each.
Which task did my five playtesters fail at, and where did they get stuck?
Watching a few people play is the cheapest test there is, if you note down the same few things every time. This entry turns your notes into a clear list of the tasks that failed, or that nearly did, with the numbers to back it up.
Did the tester stop playing because they were scared or because they were bored?
Two testers close the game at the same scene. One was shaken, the other was yawning, and they need opposite fixes. This entry reads the last minutes before a quit to tell them apart, and keeps test sessions that simply ran out of time from muddying the picture.
Do testers reach the endings you meant, or just the ones random clicking reaches?
Some endings are rare only because few paths lead there, others because players avoid them. To tell apart bad luck from taste you need to know how often a blind clicker would arrive, and compare your testers against that.
One tester complained six times: which problem should I fix first?
Raw feedback is loud where one person is loud. Counting how often something is said puts the chattiest tester in charge of your to-do list. This entry shows how to rank by people affected while keeping rare but severe problems from disappearing.
Do testers read faster and faster as the story goes on?
A story that loses a reader does not announce it. This entry turns the speed at which text is clicked through into one clear signal for a whole session: slowing down, steady, or hurrying, and shows how one long break can fake a trend.
Which choices make testers reload their save and pick again?
When someone reloads right after a choice, they either hated the outcome or hoped for a better one. This entry finds the choices that are second-guessed and shows which option people moved to, so you see where the story punishes or tempts.
I reordered my levels, and now the old and new numbers cannot be compared
After you move, merge or cut levels, "level 12" means something else than last week, and the comparison between versions quietly measures the wrong levels. This entry shows what to log instead of a position, how to carry old results over and what cannot be carried at all.
I replay a reported bad run from its seed, but it ends differently
A saved start value and the player's inputs should rebuild a run exactly, and when they do not, something in your game depends on more than you think. This entry shows how to find the first moment two runs part ways cheaply, without storing every state.
Can I replay exactly what a tester did, and know the replay is the same story?
"It broke at the third door" is useless without the way there. A short log of choices can bring a bug back on demand, but only if it was recorded against the same version of the story. This entry checks a log before you trust it.
Players replay a cleared level: do they love it or chase a better score?
A replay after a win can mean the level is a joy, or that the score left them hungry, or just a stray tap on the button under the winning screen. This entry shows how to split these apart so replay counts say something about the level.
Which scenes make testers press the "I need a break" button?
A game with disturbing scenes should let people step out, and the presses tell you where a content warning or a softer version is needed. This entry counts people per scene, asks nothing about why, and treats a single press as worth reading.