Where Players Die: Read a Death Map

Count deaths on a grid laid over your level, then rank each spot by how many different players died there and how many passed through. Raw death counts often point you at the wrong place, because one stubborn player can outweigh a real trap.

Why do raw death counts mislead?

Ashvein Mine is an invented mining game, and these numbers are made up for this example. In the level called Lower Shaft, 118 players died a total of 260 times. A grid of deaths drawn over the level is often called a heatmap.

The three worst cells, by raw deaths, were these.

Spot	Deaths	Different players	Passed through	Of those, died here
Rail bridge	96	3	5	60%
Rockfall bend	64	52	61	85%
Main tunnel	41	38	230	17%

Look at it by deaths and you would rebuild the rail bridge. Now read across.

The rail bridge had 96 deaths from 3 players, and one of them alone died 71 times there. Its 60% rests on 5 people, so it says nothing. That is someone practising a shortcut for fun, and almost nobody else ever goes there. It is not a trap. It is a hobby.

The main tunnel looks worrying, with 38 players dying. But 230 people walked through it, so about 1 in 6 died. The tunnel is simply where everyone goes. To see that, the game writes a small passed event with a position every few seconds. Without it, a busy road always looks like a dangerous one.

The rockfall bend is the real find. 52 of the 61 players who got there died there: 85%, very likely between 74% and 92%. It also caught 44% of everyone who died in this level.

What do you do with a hot spot?

Go there and play. A spot on the map is a question, not an answer. The bend might be a fair, planned danger. Your plan decides that, so write down where a player is meant to be in danger before you look.

The maker went down to the bend, and saw that rocks drop from above the top of the screen. The warning rumble had never been wired up. Players had no chance on their first visit, and only 9 of 61 got through.

The fix was to bring the rumble back, and add a shadow before each rock. He did not make the rocks weaker. A danger you can see coming is fair, and one you cannot is a coin toss.

Game developer Chris Pruett describes this same pattern in his write-up on player metrics. Plotting deaths on his level art showed players dying in pits in large numbers, and the rate did not fall as the game went on. The camera was moving the floor off screen during jumps, so nobody could see where they would land. Once he changed the camera, pit deaths fell sharply. The numbers showed where. Looking at the level showed why.

For players who quit without dying, see where players quit. If fewer than a handful of players died in a level, no spot can be called a trap yet. The death map recipe does this count and says so.

Paste this to your AI

Show me where players die in this level of my game: [level]. The event log is in [file or folder].

1. Read the level and describe the intended route as a Mermaid state diagram. Mark where the player is meant to be in danger and where they are meant to be safe.
2. Check that the game logs a position with every death, and a passed event every few seconds while moving. If not, add them. Log positions only, nothing about the player.
3. Use get_recipe with death-heatmap if Spielerando is connected. Show the map as text, then a table of the top spots: deaths, different players, players who passed through, the share of those who died, and the 95% range.
4. Point out any spot with many deaths and few players. It is not a trap.
5. Compare each hot spot with the route. Is it a danger we planned, or one we did not?

For each unplanned hot spot, tell me exactly what is at that position in the level. Say what you can show from the level, and do not guess at why people die if you cannot. With too few deaths to tell, say so.

Or skip this and connect the MCP: your AI then finds all of this by itself.Connect the MCP

How to check it worked

• Stand on the spot. Find it in the game from the position your AI gave you.
• Play it as a first-timer. Can you see the danger coming?
• Count again after the fix. Deaths at that spot should fall, and should not simply move to the next one.