Metrics and adaptation

Mark Nelson [email protected]

Metrics and adaptation

Fall 2013 www.itu.dk

Game metrics

Anything that can be recorded about gameplay

Used to understand how the game works, and how players experience it

Range of goals and techniques Qualitative, quantitative, game user experience, monetization

Some types of game metrics

Which player are the metrics about?

Basic process

Choose variables to record Choose frequency of recording Store to a database

Run query/visualization later Aggregation, correlation, etc.

Simple example

Spatial, in-engine metrics

Record positions of interesting events

Visualize frequencies, relationships, etc. Usually by drawing on the map

One of the most commonly used kinds of metrics Easy to record, clear how to visualize

Heatmap: HL2 deaths

Heatmap: TR8 in-game help usage

More in-depth spatial analysis

Heatmaps are 2d density estimates from samples

Can do other kinds of density estimates

Or, combine with qualitative data

Binned heatmap

Using a GIS?

Geographic Information System (GIS)

Used for aggregating spatial data in geography, urban planning, etc.

Some recent experiments with using GIS to analyze ”game geography”

GIS overlays

ArcGIS overlay (Hitman: Blood Money)

In-engine/editor visualization

Analytical metrics

Analyzing the game to extract information without player data, or as a supplement to player data

Examples: Reachability analysis Game-theory analysis of a combat system

Recent area of research M.J. Nelson (2011), ”Game Metrics Without Players”

More on metrics

Annual conferences/workshops: Game User Experience

Engines increasingly come with their own metrics/viz infrastructure And extensions, e.g. in the Unity store

Adaptation

Can we use metric data automatically?

Offline or online

Dynamic difficulty adjustment

Adjust game based on player’s in-game behavior

But ”smarter” than classic rubber-banding style DDA

Information on what player does successfully/unsuccessfully, combined with a model of interestingness

Experience-driven PCG

Use metrics information to generate levels or other game content

Offline: a kind of partially automated game design Online: game customization

Like a more ambitious DDA

Togelius & Yannakakis (2011). ”Experience-driven PCG”

Final project

Project can be done individually or in a group Individually written reports in either case

Scope There are about 6 weeks left in the semester Depending on the project, leave ~2 weeks to write up a report

Kinds of projects

Implementation projects

Scientific-experiment projects

Engine investigation projects

Implementation projects

Examples Expand your software 3d renderer to do solid surfaces and lighting Implement a character AI/scripting API Write a Unity plugin to do something not currently supported

Scientific investigation projects

Do an analysis/comparison of algorithms or techniques

Possibilities Implement several standard algorithms from a textbook (AI,

graphics, etc.) and compare them Implement something proposed in a research paper and try to

replicate or extend its results

Engine investigation projects

Take an existing game engine and dive into a specific part of it (physics system, AI system, navigation system, …) to understand and characterize how it works.

Possible topics of investigation What were the design goals, if known? What does this support well and poorly? Benchmarks, etc.