What We Learn From AI Agents - Part 2

In Part 1, we looked at how AI Agents can help us understand spells, loadouts, strategies, and combat behaviour.

But every match also produces another important layer of information: how Agents use the game space itself.

Movement, positioning, encounter locations, and match pacing can reveal how different parts of the game interact during real gameplay situations.

Movement Through the Map

AI Agents generate useful information about how the map is being used.

We can observe which routes Agents take, how they move between different areas, and which locations they naturally prefer.

Over many matches, this can reveal:

• Frequently used routes
• High-traffic areas
• Rarely visited locations
• Common movement patterns
• Positions Agents repeatedly return to

Some routes may become popular because they are faster or safer.

Others may provide better visibility or stronger positioning.

At the same time, areas that are consistently ignored can also tell us something.

They may be difficult to reach, strategically less useful, or simply outside the natural flow of a match.

Where Fights Happen

Movement data also shows us where combat naturally develops.

If a large percentage of fights happen in the same few locations, those areas may have particularly strong positioning, visibility, access, or strategic value.

A location may become a natural combat hotspot because several important routes meet there.

Another area may give Agents enough space to use certain abilities more effectively.

Areas where combat almost never happens are just as interesting.

Understanding these patterns helps us see how the map is actually being used rather than only how we expected it to be used.

Engagement and Match Duration

AI Agent data can also show us how long individual fights and complete matches last.

This gives us useful information about pacing.

If engagements repeatedly end almost immediately, certain strategies or combinations may be producing very high burst potential.

If fights consistently last much longer, defensive play, positioning, movement, or survivability may be having a stronger influence.

Different Agent configurations can also produce very different overall match lengths.

Aggressive strategies may create faster matches with frequent encounters, while defensive configurations may create slower games focused more heavily on positioning and distance.

How Movement and Combat Connect

These signals are especially useful when we look at them together.

Movement influences positioning.

Positioning influences where combat happens.

Where combat happens can influence which abilities are most effective.

Those abilities can then affect how Agents move or where they choose to fight next.

For example, if a certain type of ability performs particularly well in open spaces, Agents using that strategy may begin favouring specific areas of the map.

A defensive strategy may produce completely different movement patterns.

This helps us understand the game as a connected system rather than a collection of isolated mechanics.

Finding Unexpected Interactions

Games contain many systems, and not every interaction between them is obvious in advance.

AI Agent matches can produce situations that may not appear during a limited number of manually designed tests.

For example:

• A spell may interact unexpectedly with movement
• A particular build may perform much better in one area of the map
• A weak strategy may become highly effective with the right positioning
• A movement route may create an unexpected tactical advantage
• A specific combination of mechanics may produce unusual match pacing

These situations can reveal behaviour worth investigating further.

From Data to Better Questions

AI Agent data does not automatically tell us what should be changed.

Instead, it helps us identify where interesting patterns are appearing.

It can show us that Agents repeatedly avoid one area.

It can show us that fights constantly happen in another.

It can reveal unusually short matches, unexpected movement patterns, or combinations that behave differently depending on positioning.

That allows us to ask better questions:

• Why is this area being ignored?
• Why does combat repeatedly happen here?
• Why does this strategy create much shorter matches?
• Why does this ability perform differently in this part of the map?

Those questions can then be explored further through traditional testing, development, and player feedback.

Building a Broader Picture of the Game

No single Agent tells us how the game should work.

The value comes from looking at behaviour across many different matches and configurations.

By combining information about movement, positioning, encounter locations, pacing, spells, loadouts, and strategies, we can build a broader picture of how different systems influence each other.

AI Agents provide another layer of gameplay information alongside traditional testing and real player feedback.

The value is not simply in generating more gameplay. It is in using that gameplay to better understand how the game behaves as a system.