Real-Time Play: Tactical Patterns

Jan 29, 2011

Inexperienced Though Not Expert

The majority of players are experienced though not experts. However, when we study play we tend to emphasize either inexperienced or expert play. The study of inexperienced players is ultimately the study of learning, and the study of expert players is the study of algorithm-like optimization, neither of which reflects how most players spend most of their play time.

Experts account for knowledge such as the mechanical algorithms underlying Pac-Man ghost behavior, and, when compatible with how a particular game works, will train and drill specific patterns exploiting that expert knowledge.

The term behavioral heuristic was introduced in Four Aspects and Interpretation, where it was defined as, “a shortcut for decision making, which is applied because there is not sufficient time for evaluating ideal decisions in every circumstance.”

One example behavioral heuristic that I covered in detail was: “Because there is a timer, I should rush.” I suggested “sacrifice any piece to protect the queen,” as a behavioral heuristic we might expect for an amateur chess player.

In this entry, I will provide several more examples of behavioral heuristics, then introduce the emergent result of combining multiple behavioral heuristics: tactical patterns. It will be suggested that this style of play is applicable to a variety of genres, and is a sign of experienced, non-expert play.

Additional B.H. Examples


Behavioral Heuristic: Because there is some use for a resource, and the player is able to increase their stores of that resource, they should accumulate more of it any time more becomes available.

Examples: Special ammunition in FPS games, money in RTS/adventure/RPG games

Conflict: This may lead to accumulation well beyond any practical use, since it’s built up not to satisfy any planned need, but instead to remove the need to build it up later.

Clear Every Enemy

Behavioral Heuristic: To deal with unknowns about the AI’s tracking capabilities and/or uncertainty having to do with which areas of a level may need to be revisited, the player goes out of his or her way to remove every enemy seen.

Examples: Common in stealth action or survival horror games, especially when playing an area for the first time or on hard mode.

Conflict: This frequently puts the player in unnecessary danger, and wastes limited ammunition, more in exchange for a sense of security than for any significant impact on the chances of mission success.

Never Use an Easily Countered Unit

Behavioral Heuristic: Because an effective, direct counter exists to a particular type of unit, the player develops a blind spot for the role that unit could play tactically.

Examples: Common in RTS. For example, because infantry can be easily defeated by flame weapons, and the opposing team has access to flame units, a player employing this behavioral heuristic would avoid deploying infantry.

Conflict: If the other player is not in a situation to build the counter for said unit – due to lack of resources, incompatible team selection (an army lacking the counter unit), inadequate time, or inexperience – it could serve a valuable function in play. In fact, because a designer is likely to recognize and aim to offset a clear weakness, there’s a good chance that situations where the unit can be used offer a significant advantage by using it.

Tactical Patterns

The above examples each demonstrate an all-or-nothing persistent consideration. One tier of greater player sophistication is to switch between which behavioral heuristics are applied based on the current context. Still, the player is employing heuristics, by behaving in a patterned way without thoroughly accounting for particulars and probable outcomes, however the player is no longer so easily toppled.

RTS Tactical Patterns

An RTS example is easiest to step through, although similar considers can be made to how an FPS player starting a new level, or what MMORPG player does upon joining a new server:

  1. Player begins, needs to prioritize finding a place to set up a base.
  2. Once first building in base is started, power is needed. Player prioritizes this until a certain amount of power is reached.
  3. Establishing cash flow is the next priority, until enough is coming in.
  4. Next the user creates means of production for various unit types.
  5. Finally, the player switches into a set of mid-game behaviors, prioritizing based on local buildup and observed enemy activity.

The list above of 4+ patterns reads like a sequence – and it can often be treated as such – but how and when players employ these behaviors later in a round reveal that they’re associated more with minimum satisfying conditions than with an order, except to the extent that they’re prioritized when multiple needs apply. If an enemy player sabotages the player’s power sources, means of production, or disrupts cash flow, the behavioral heuristics attached to correcting for those issues (which is likely similar to the initial behavior employed to address those concerns) kicks in as top priority until the issue is addressed.

Each piece of that plan can be adjusted, independent of the rest. For example, a person might adjust their sense of how much cash flow is an acceptable minimum, while addressing the issue the same way, just taking longer before advancing steps. The player could modify how they handle a particular issue, for example changing the order or types of buildings made to produce units, without affecting the conditions under which that behavior takes focus. Lastly, new patterns can be added, attached to their own satisfying conditions and priority levels, for example sending out expendable units to all corners of the map as a low priority, given that the map has not yet been explored to clear fog of war.

Behavioral Heuristic Connection

With RTS games, the connection between behavioral heuristics and tactical patterns is not immediately obvious. The circumstances are more complicated, and the behaviors are more involved, which conceals the relationship.

Although tactical patterns are derived and adjusted by trial and error, the behaviors attempted first are products of behavioral heuristics suggested by the Four Aspects. Players initially respond to the instructions, assumptions based on visuals/audio (impression), and what the interface emphasizes, though for a player to become an expert their primary interest needs to be implementation (all causal relationships, timing, and tuning in the game).

As a result of that effect, similar local maxima are likely to be found by many players. Small deviations from those behaviors will show poorer results. Expert players avoid this plateau by conditioning their play behaviors through some other activity (ex. coordinated drills, study of tactics used by more successful players, wider exploration in trial and error, etc.).

Shmups Tactical Patterns

Shoot’em Ups, also called “shmups” for short, or “bullet hell” (games including R-Type, Gradius, Raiden, and Ikaruga) require combining very simple considerations:

  1. Because enemies come from the top, stay on the bottom half of the screen.
  2. Because forward/backward movement gives the player an additional dimension for dodging, but one of those directions is blocked when up against a screen edge, don’t spend long touching the very bottom of the screen.
  3. Because enemy shots tend to move primarily vertically and toward where the player’s was when fired, constantly move side-to-side.
  4. Because some shots are angled or tracking, rock forward/backward to let them pass.
  5. Because ammunition is unlimited, rate of fire is high, enemies are in dense formations all in one direction of the player, and shooting enemies reduces the number of projectiles to dodge, fire constantly.
  6. Because power-ups increase a player’s ability to defeat enemies, but appear where enemies are defeated (mostly likely to happen where enemies are most dense on the screen) and vanish if not collected quickly enough, the player often should dodge very briefly upward into danger to collect power-ups when they appear.

Each of those is a behavioral heuristic, a rule of thumb tying a constant behavior not to particular enemy or bullet placements, but instead to general considerations about such things. The last one listed, about power-ups, is in fuzzy area between behavioral heuristic and tactical pattern, since on the one hand it’s a reason to weave closer to where enemies are dense (less gap to close when power-ups are dropped), but on the other hand it incorporates a different set of considerations that become prioritized in a particular situation (it is composed of behavioral heuristics which become most relevant when a power-up appears).

The total result of these considerations is that the player generally winds up moving in a jagged figure 8 in the bottom half of the screen, firing constantly, swerving to acquire power-ups if they’re close enough.

Homemade Shmups

Boss fights in shmups add variety to the gameplay by introducing one-of-a-kind attack patterns, driving the player to derive different behavioral heuristics and tactical patterns than were useful in general level play. Where it’s safe to be, how it’s safe to move, and when it helps to fire all change. (Video of boss fight in Fraxy.)

Higher Dimension Example

In Warhawk, or another combat flight game – evasive maneuvers when “Missile Locked On” comes up on the screen temporarily make attacking impractical, as a tradeoff for becoming harder to hit. Otherwise, a search behavioral pattern is enacted, combing the map at a safe altitude and range, switching to an aggressive dogfighting behavioral pattern when a vulnerable target is identified.

Note that the success of an evasive maneuver in dodging a heat-seeking missile, whether in 2D or 3D, is often not a directly programmed result, but an emergent effect of the chase algorithm used, and constraints on ship/missile changes to speed and heading. (Though there are exceptions to this – Star Fox on SNES provides a counter example, in which performing a barrel roll made the player ship temporarily invulnerable to enemy fire, a directly programmed cause/effect.)

The Importance of Being Overwhelmed

A challenging game requires the player to be at least slightly overwhelmed, and therefore unable to make an optimal decision – too much information to respond to for the amount of time available for the decision. Chess forces us to resort to heuristics because there are too many moves to consider, even given unlimited time per move; on the other end of the spectrum, a shmup or RTS game challenges us because in the time we spend trying to figure out how to perfectly account for all factors, the enemy will have acted on pattern to disrupt the circumstances we’re evaluating.

(Those familiar with game-related psychology, or Jenova Chen’s thesis, will recognize the stretching of our abilities as related to Mihály Csíkszentmihályi’s concept of flow. See also: Mihaly Csikszentmihalyi TED Talk on Flow.)

Without being overwhelmed, there is no need to resort to adopting and evolving heuristic behaviors. Instead, algorithmic behaviors – strict steps able to be determined and executed with precision – would allow better results, replacing the dynamic act of play with the straightforward act of creating and following instructions (ex. Tic-Tac-Toe).

In the case of a shoot’em up game, the player’s attention is constantly pulled to two or more areas of the screen: to the player’s ship for purposes of dodging, to the enemy ships for purposes of aiming, and to power-ups for purposes of updating navigation to intersect with them. This results in the mix of overlapping behavioral heuristics battling for priority that were outlined above.

Note that if the shots, enemies, or player ship in Raiden moved differently, a different pattern of movement would be preferable to the one shown here.

Classic Doom AI poses similarly conflicting pressures on behavioral patterns, from which the game’s combat derives its richness…

Some enemy types have guns that require obstructing line of sight in sync with their firing rhythm.

Some enemy types shoot slow projectiles that do more damage but can be dodged.

Some enemy types run fast but have no range which forces the player out from cover.

Some characters fire slow weapons that can cause collateral damage if dodged too close to a wall, some characters have rapid fire to force lateral approaches.

One enemy type can bring dead enemies back to life. (Doom 2 only)

Most enemies run toward the player in a zig-zag pattern, making them difficult to shoot until they’re also firing.

The player is introduced to each of these characters gradually, then these different types get thrown together in various combinations in visually varied settings each having unique cover opportunities, traps, ambush points, forms of exit, lighting conditions.. The world plays out a bit like a chessboard involving 30 moves a second without distinct squares, forming poetry with the simple grammar and vocabulary taught to the player in how the various enemies attack.

Even an inability to see the entire world at once – offscreen, behind the player, or behind walls and doors – creates a need for behavior heuristics. In case an enemy we cannot see begins firing on us, or is lining up a shot to do so, shouldn’t we already be in motion?

Adaptive Pattern

Adaptive pattern is a special case of tactical patterns, in which one or more patterns is exploratory, experimental, even random, and operates in some fashion as to create, modify, or reject other tactical patterns.

Team Fortress and other team-based multiplayer games where the players have the option of serving many different positions or classes (going for the flag, providing certain types of defense, healing teammates). This keeps adaptive patterns alive, regardless of how experienced the players involved are, since the combination and styles of players on both teams are always in flux. The weighting of which role on the team a player might fill becomes not only a function of their own ability, but is also impacted by who else is playing at the time on both sides. If the player’s team already has one or more members successfully dealing with snipers, but no one is capturing the flag, the player might switch to Scout class to serve as a flag runner, or vice versa if the already has a flag runner who is having difficulty getting past enemy snipers. The potential for current players to switch out, and for new players to switch in, means part of the player’s pattern needs to include occasional reassessment of what role they can fill that will best serve the team’s needs.

In contrast to the earlier Warhawk video, which depicted a player serving their team as a pilot, this gameplay video shows a player helping their team as infantry – purposefully passing up opportunities to use parked jets, tanks, and SAM sites in favor of helping disrupt enemy vehicles as a smaller, more agile target.

Why This Matters

This entry, and Four Aspects and Interpretation before it, are early attempts to make clearer sense of moment-to-moment player actions in real-time games, as steps toward understanding the type of learning that takes place during videogame playing. This entry presents a way of framing non-expert game play as adaptive, contextualized trail-and-error originating with representationally educated guesses – importantly, that is a way in which players improve at videogames without constructing a more accurate mental model of how the videogames operate.

(What follows is mostly conjecture. I’m still working on filling in between where I’m at and being able to more rigorously support or test this next round of ideas.)

When developers bury meaning in a videogame, deeply rather than superficially, we often do so by planning out systems relationships, under the assumption that through practice players will internally reconstruct the underlying system. This is likely incorrect, as it seems like assuming that an athlete could gain a doctor’s understanding of anatomy by playing sports, or that a pilot could come to understand the device’s mechanical engineering by practice flying planes.

To the extent that tactical patterns and behavioral heuristics do form a system, it is (a.) complementary to, not a mirror of, the system they are iteratively shaped by, and (b.) is not necessarily understood by the player as system, which would presumably make generalization or transfer of the knowledge more difficult. If these experiences mean or teach something, it seems probable that they mean or teach something other than what we intend or think that they mean or teach; being better equipped to recognize and account for such disparity could improve our ability to embed deliberate meaning or information into the gameplay of real-time videogames.

Learn and practice team game development with Gamkedo Club.
Membership worldwide. Professional support. Proven process.

Subscribe by e-mail to receive weekly updates with Gamkedo.Community interviews and YouTube training videos for game developers!

All contents Copyright ©2017 Chris DeLeon.

Site production by Ryan Burrell.