What the enemy is actually doing

An enemy in a tactics game has a strange job. It is not there to win, because it can win trivially by being given more health and better aim. It is there to make the player's decisions matter, which means it has to be dangerous in ways a player can anticipate, prepare for, and then be punished for ignoring.

That is a harder brief than being clever, and the genre has solved it three main ways.

Commandos: Behind Enemy Lines (Pyro Studios, 1998) has almost no artificial intelligence in the usual sense. Every guard stands where a designer put him and walks a route somebody timed. The enemy is a mechanism.

This is not a limitation, it is the design. Because the mechanism is identical every attempt, the player can learn it, and learning it is the game. The skill being tested is observation and timing, and both of those require the world to keep its promises. A guard who wandered randomly would make the whole thing unplayable.

The cost is obvious and worth paying: once solved, a level is finished forever. Hand-built, deterministic enemies make the best puzzles and the worst replay value.

X-COM: UFO Defense (Mythos Games and MicroProse, 1994) runs the other model. Aliens are not following routes. They have simple behaviours, they have weapons, and crucially they have a view of a map that is mostly dark to you.

The result feels intelligent because it is unpredictable in the right way. Nothing coordinates a sophisticated attack on you, but a thing you did not know was there shoots somebody in the dark, and the mind supplies a plan it did not have. Terror is cheap when the player has less information than the enemy.

Jagged Alliance 2 (Sir-Tech Canada, 1999) is the high-water mark of this family and is still praised for it. Its soldiers use cover, throw grenades at groups, fire to keep heads down and go round the side, which is more coordinated behaviour than most games twenty years younger manage.

XCOM: Enemy Unknown (Firaxis Games, 2012) introduced the structure most modern tactics games use. Enemies exist in small groups, they are dormant until something sees them, and on being activated they get a move to scatter into cover before you can act.

It is a compromise with real benefits. It keeps the pacing of a mission under the designer's control, it makes each fight a discrete problem, and it means the computer only has to think about a few units at a time. It is also the single most complained-about feature in modern tactics design, because that free scatter move punishes the player for the act of looking, and because a map full of patient dormant groups is a map that does not behave like a place.

The honest reading is that pods solved a production problem and a readability problem at the same time, and charged the player for both.

There is an engineering reason the enemy rarely outthinks you. The number of things a single unit could do with a full budget of action points on a real map is enormous, and it multiplies for every additional unit acting in the same turn. Searching that properly is expensive, and it has to happen in the seconds a player will tolerate waiting.

So nearly every game scores options one unit at a time: how good is this tile, this target, this shot, for this individual, right now. That produces enemies who are individually sensible and collectively thoughtless. It is why a genuine pincer movement is rare, and why when one does happen players remember it for years.

Every difficulty setting in this genre does one of three things, and players react to them completely differently.

Give the enemy more. More health, more units, better gear. Players accept this without much complaint, because it is visible and it can be planned for.

Give the enemy better information. Let it know roughly where you are, or let it stop behaving as if it is searching. Players notice this immediately and it makes them furious, because it breaks the one contract the genre runs on: that hiding works, and that what the enemy knows came from somewhere.

Adjust the dice quietly. Hidden modifiers on hit chances, in either direction. This is the most common and the least defensible, because a player who suspects it has no way to check, and once the suspicion exists every miss is evidence.

The best enemy in a tactics game is one whose next move you could have guessed, placed somewhere you did not expect. Surprise should come from the arrangement, not from the rules changing.

Into the Breach (Subset Games, 2018) proves the point by going all the way. Every enemy shows exactly what it will do on its next turn, with no randomness at all, and the game is still difficult, because knowing the plan and being able to stop it are different problems. It is the clearest evidence in the genre that predictability and difficulty are not opposites.

RAD Tactics is a squad tactics game for Windows, real time with pause, set after a collapse, with zombies and with factions fighting over what is left. It is in development, and we are not going to describe anything in it that is not finished.

The line we care about is the one in the middle of this page. An enemy may be stronger than the player, and it may be better equipped, and it may outnumber them. It should not know anything it was not told by something that can see.