What to measure
Learning paths through this topic
How much of the game is fighting, talking, walking and menus?
The mix of activities tells you what your game really is, which is often not what you planned. This entry shows how to add up time per activity from change events, including the activity the player was in when the session stopped.
How long does a monster fight really take when many players run away from it?
Average fight length looks fine until you notice that the longest fights are exactly the ones players abandon. This entry shows how to report fight length honestly and when to distrust it.
Which areas can players cross without ever being hurt, and are they too easy?
An area where almost nobody takes a hit is either a rest stop you planned or a stretch that has gone dull. This entry shows how to count it fairly, so people who just ran through do not make a hard area look easy.
Do players hoard potions and scrolls until the end and never use them?
A full bag of unused items at the end means the game never made them feel worth spending. This entry shows how to measure that fairly, and why the players who stopped playing early must not be in the count.
How much gold does a typical player carry at each level?
Gold numbers get skewed by a few players who save everything and by players whose game writes down their wallet far more often. This entry shows how to find the wallet of a typical player so you can tell whether prices fit.
How can I tell from walking paths that a player is lost in the world?
A lost player does not stand still, they walk in circles. This entry shows how to turn a short walk into one number that separates lost from busy exploring, and what to do with players who are simply standing.
The server is "only slow sometimes", so how much of the game time does that cost?
Saying that a small share of server steps ran late sounds harmless, but late steps are long, so they cost far more time than their count suggests. This entry shows the figure that matches what players feel.
How often does the host pull players back, and for whom is it worst?
Counting snap-backs in total says little when some players only stayed a minute. This entry shows how to turn them into a fair rate and how to avoid blaming the wrong people on the basis of a very short stay.
Playtime numbers are wrong because the game was left open overnight
If you measure how long people play from the earliest to the latest sign of life, one forgotten tab turns a ten minute visit into eight hours and wrecks every average. This entry shows how to cut the stream of events into real sessions and real play.
Spot the player who earns a hundred times more gold than everyone else
A duplication trick or broken reward shows up as a few players with absurd earnings, but the usual way of measuring "unusual" is blinded by those very players. This entry shows a measure that is not fooled by the outliers it is hunting.
Which combinations of heroes win most, and does a team without a healer really lose?
Teams are made of the same heroes picked in a different order, so counting line-ups naively splits one combination into many. This entry shows how to name a combination so it is counted once, and how to test a single missing role.
Two teams look equally strong on paper, but one match is a slaughter
Comparing the average skill of two teams can say everything is fair while one side has a far stronger star player. This entry shows two numbers to log for every match so unfair pairings show up before players complain.