Melee Combat 101 - Rough Draft

By Seijin8Seijin8, on 24 Jan 2014 15:12
Last updated at 31 Jan 2014 10:46

Incomplete work, but (marginally) ready for critique

Melee Combat 101

The Challenge of Creating Realistic Swordplay in Video Games


The purpose of this article is to acclimate the reader to real-world melee combat to better illustrate the challenges in crafting a game interface that captures those realities. This is meant to spark discussion about what makes the best, most intuitive and fun game interface, and is not meant to convince the reader that we want a computer simulation that perfectly mirrors the real thing.

For sake of clarity, I will be limiting the discussion to a medieval era, multi-cultural1 historical game without any "magical" components. The goal is to generate an immersive melee gaming experience. It is necessary to note that melee training and immersion are both highly subjective terms, and so there may be experts in the field who disagree with some of the "realities" presented here. Criticism of these themes is welcomed in order to produce a more broadly appealing snapshot of what this (entirely fictitious) melee game would evolve into.

This article is broadly divided into three parts:

  • Part 1: dealing with incoming sensory data and the internal wiring of the body during combat.
  • Part 2: dealing with idealized control inputs and broad tactical outlines of what the game's characters should be capable of doing.
  • Part 3: discussing feedback from the game's events, mechanisms integrating conditions discussed in parts one and two, and looking towards future mechanisms for improving our game interface.

Part 1: The Subjective Experience of Life-and-Death Struggle

“Detached reflection cannot be expected in the presence of an upraised knife." – Oliver Wendell Holmes

Let's begin with a discussion of what melee combat is probably like for the majority of people who find themselves involved in it. For now, we will ignore the individual's background or physical characteristics and will play the role of "Joe Average: Wandering Swordsman", facing a single individual with similar armament and hostile intent.

Finding yourself facing the prospect of an ugly death at the hands of an enemy, your biochemistry and senses undergo a sharp transformation. The experience is both terrifying and exhilarating, senses elevated and diluted, with much of the body responding in heightened fashion, honed by countless prior generations of conditioning23.

Incoming sounds might be diminished or amplified. Your sight could be red-tinged, narrowed to a tunnel, or could be shocking in both clarity and intensity. Events might pass by swiftly and unnoticed or time could seemingly slow to a crawl. All of these are potential effects derived from the potent chemical cocktail that survival stress causes to be released into the bloodstream. This state of being goes by many names: "Adrenaline dump", "high stress state", "survival stress response", "amygdala hijack", "helmet fire", and "in the zone" to name just a few. The state is no guarantee of combative capacity, and can just as readily lead to running or freezing in place. Depending on genetics, experience, circumstance or random chance, this high-stress state might force a crushing weight of perceptions and loss of equilibrium, or conversely, could provide an uplifting sense of invulnerability and unnatural coherence of thought. The sensation could be welcoming and soothing, or utterly alien and unnerving.

Frequently, high-stress state manifests in some disorienting, less-than-perfect hybrid of the aforementioned perceptions4.

Through this sudden miasma of sensory warp, you can see that the foe you face is undergoing much the same experience. They are likely unaware of their rictus grin and the spittle they emanate. Don't think the crazy smile is joy in any sense; more likely just excess muscle tension peeling the lips back. You5 probably look much the same, with an unnatural pallor6 and hairs standing on end7. The pinpoint-small pupils the adversary aims at you might be seeing everything in perfect clarity, or may not see you at all.

Adrenaline will definitely make you both jittery and uncoordinated, but also amplifies the speed and power of well-honed activities8. This biochemical stress response could allow you to withstand great injury as though it were nothing… for a time. You might be able to run, jump, fight, and hurl objects with a great deal more force and speed than you could consciously summon.

This roller-coaster ride will burn you out quickly9, and when it fades, the euphoria often turns to nausea and a fatigue so profound that you might opt to lay down for a nap, even amid a raging battle1011.

Part 1A: Internal Wiring and its Relationship to Response Speed

Our introduction to the first giant hurdle that cannot be jumped"

How does adrenaline get triggered? It starts with a perception that there may be danger. This perception is initially received by a part of the brain called the midbrain12. The purpose of the midbrain13 is to preserve your life. It will act to do so without conscious thought and can go on "autopilot" if there is a plan in place for countering the perceived threat. Note that running and fighting are the two template plans it has for dealing with most threats, and this pair of opposing responses is where "fight or flight reflexes" come from.

Before we can examine the tactics and control mechanisms of melee combat, it is important to recognize just how severely high-stress state affects the body and mind of someone confronted with battle. Heart rate may momentarily spike to 220 beats per minute, stabilizing a few moments later at 150-180bpm for the duration of the threat141516. Hyperventilation is also common at this point and may rob you of the ability to speak17. The muscles become infused with blood and often shake uncontrollably. This is particularly noticeable in the legs, with the thigh muscles quivering in preparation for movement, causing the knees to move around a lot to remain upright. Typically, you will keep balance by crouching, often in preparation for forward movement18. It is very common for a combatant to crouch aggressively just to offset this muscular imbalance19. The muscles of the chest and back are equally tense, causing the shoulders to hunch20 and making the elbows project out to the sides. The net effect of all this postural adjustment is that the combatant seems wider and lower to the ground, sort of like a wild animal21.

Returning to a first-person perspective, here you are, bent over and hyperventilating like a crazy person, sword in hand22 and ready to engage the adversary (who is similarly postured). Without some kind of special training to modify these behaviors23, you are likely facing chest toward the threat, wide-legged and staring straight at them. You may be numb or feeling every slight breeze like it was sandpaper. While you can probably still hear24, most sounds are muted except for your ragged breathing and the thunder of your own heart pounding like a jackhammer.

In this adrenal stress condition your internal wiring has changed substantially. Signals coming into your senses are being diverted and processed in a more rapid (but less precise) fashion, causing a lot of the weirdness you're experiencing. The midbrain is a faster processing mechanism that the forebrain25. Its speed allows it to gather data and respond fast enough to (hopefully) save your life.

The midbrain operates on a very different layer of reality from the forebrain. It is utterly devoted to speed of action over correct action26.

To better illustrate the differences in reaction speed between the forebrain and midbrain, let's examine the normal27 process for "seeing" something and then acting upon it:

  • Light enters the eyes
  • The eyes relay the signal to the visual cortex
  • The visual cortex receives the signal, converts it to a rough image (which is immediately dispatched to the midbrain), and then finishes the image and sends the refined, high-def version to the forebrain
  • Images travel along neurological relays to the midbrain (short trip) and the forebrain (longer trip)
  • If the midbrain does not detect a threat in its rough sketch, it remains dormant
  • The forebrain receives the image, decides what it means (OODA loop, detailed later) and what to do about it
  • The forebrain relays the required actions to the motor controls (hindbrain)
  • Hindbrain receives data and signals the body to execute action (further delay depending on bodily conditions)

This whole process — assuming nothing too complex needs to be decided upon by the forebrain — takes upwards of .4 seconds, possibly a full second for more complex (but still basic, computationally speaking) decisions. Much of this time is devoured by the various neurological relays the information has to pass through, and the forebrain eats up the rest of it with the decision process.

Now, that doesn't mean you can act on this data within .4 seconds, just that the body begins responding in .4 seconds. Time to complete a simple action (dependent on the action's orientation to target and execution time) could be from .5 seconds to a full second or more.

That is a hell of a long time when your life is in danger. If the forebrain managed our threat responses, we would never have survived as a species.

Luckily, the midbrain handles threat situations and does all this (perception, decision, rapid execution of action) in less than half this time, sometimes as quickly as .3 seconds, and occasionally even faster.28. This rapidity of action is mainly due to the midbrain being situated next to all the major parts of the puzzle: it is near the visual cortex, memory centers and hindbrain, and has the switch marked "adrenaline pump" to kick off events when a red alert arises2930.

That means the by the time you consciously31 perceive danger the midbrain is already moving your body to deal with it, using whatever training it has received. Adrenaline ensures that the body will be able to respond quickly to the midbrain's demands, and the massive spike in heart rate makes sure the adrenaline32 is quickly fed into the musculature.

In a lot of ways, it is like you've been thrown into the passenger seat of your own body while the midbrain takes over. And that is a good thing! Your conscious self cannot respond at this level and has no business trying to. It would just get you killed33.

Part 1B: Sustained Response Speeds and the Instinct for Tactile Engagement

Since we are talking response speeds — and grinding the body's gears to get into this crazy-fast response envelope around 1/4 second — the stimuli method starts to be really important. Up to this point, we've discussed vision as the input, but two other senses come into play as well: auditory and tactile (sound and touch).

Though exact measurements will vary from person to person and type of stimuli, a general hierarchy of sensory response times (under midbrain or well-trained forebrain control) would have sight around .225 seconds, sound around .2 seconds and touch around .125 seconds3435. Though all of these are very fast response times (to our forebrain at least), touch is by far the fastest, around double the response speed of sight.

This becomes crucial when we get to the last part of the reaction speed puzzle: The Boyd cycle, more commonly called the "OODA loop". OODA stands for "Observation, Orientation, Decision, Action" and arose from a military logistical process. Though it was created to describe the response mechanisms within a large organization, it works just as well on a personal level and has become a fixture of modern combatives training.

The conceptual mechanism of the OODA loop allows us to analyze what happens after a danger has appeared and a response has been initiated. We then have to perceive36 the result of that action, consider our options based on present arrangement of objects we can work with37, decide what is the best option38, and then act on it39.

A complication comes if a new, contrary observation arrives before an action has been taken. This resets the cycle, calling for new orientation, decision and action. This breach of the cycle is sometimes called perceptual separation, where the OODA loop's newest observation is in conflict with its current decision to act40. By itself, this is problematic, but against a capable enemy, the visually conflicting signals are rapidly joined by tactile signals and pain…

Recalling that the midbrain can complete some version of this cycle within .25 seconds, how much more effective would it be if the midbrain's first action on perceiving a threat visually was to rush ahead and make direct contact with it?41 Now the midbrain has a faster signal process to work with (touch), and can complete its OODA loop before the foe's forebrain knows the fight has even begun. As cool as that is, it also means that your brain cannot receive a contraindicating visual stimuli before a tactile OODA loop has completed. This virtually removes sight from the battlescape.

There is a reason enraged people reflexively grab onto one another: it is to create this condition42.

We are wired to want to do this under stress, and while training and circumstance can certainly alter this behavior, there is no way to produce a faster OODA loop than to be in constant contact with the foe. While this technically works both ways, often the first to latch on and start bludgeoning43 will be the first to get inside the other guy's OODA loop. This is frequently decisive.

Part 1C: Contrasting Sensory Perceptions with the Video Game Interface

So here's the fundamental problem:

  • Video games are visual and auditory, forebrain-dominated experiences with minimal bodily movement. The player operates in reaction loops longer than .4 seconds.
  • Melee combat is a visual and tactile, midbrain-dominated event that ranks among the most intense whole-body experiences possible. The midbrain contains base reflexes that will seek to establish reaction loops no longer than than .2 seconds, accomplished through tactile feedback.

Viewed in this way, the gulf is simply too much to bridge with any modern sensory interface. No matter how "immersive" a melee game attempts to be, it will simply never be anything but a weak simulation. Hell, even playing with action figures gives some kind of tactile feedback and a truer sense of situational awareness.

Even if we could bring ourselves into a fully immersive midbrain experience of this sort, would we want to? Personally, I don't like being nauseous and jittery, and I doubt a game making me that way would be a welcome addition to my life. Given the countless hours I've poured into perfecting my audio-visual experience in Skyrim, it should be no surprise that I wouldn't want a game that simulated a lot of tunnel-vision and auditory exclusion on a regular basis, no matter how "real" it might seem.

Let's be honest: we don't want our games to be "real" at all. We want them to be fun, thought-provoking experiences that allow us to be the hero (or villain, if that's your thing) of a world fundamentally different from our own.

So why the extraordinary effort to push towards "reality" in gaming? If I had to guess, it is because our forebrains can tell the difference, and want a reality that seems "real enough" to get temporarily lost within.

So if we can't (or don't want to) feel melee combat, what parts of it do we want to simulate? What sort of mechanism can we use to get some of the feel without going overboard into a nightmarish experience?

As we transition into a discussion of physical technique, note that real-world tactics must function within the midbrain's exceedingly tight response times, and the best tactics make a mockery of even the midbrain's processing times. The most significant differences between "real" and "cinematic" maneuvers are found by analyzing their inherent OODA loops and triggering mechanisms. Most cinematic movements take a long time to develop because the participants (stuntmen and actors) are physically distant, reliant on visual stimuli, and are playing to an audience whose own visual responses are measured in forebrain time.

From a pure gameplay perspective, if the AI opponents were operating in midbrain time, you probably would never see or be able to respond to their attacks, and this would not make for a very fun game experience. Simply "slowing the game down" doesn't work because all of the characters on screen would be obviously reacting slowly, and your own control interface would have to be equally unresponsive/sluggish to mirror that reaction event. This is especially sticky when you consider that halving the game's measure merely gives your forebrain-dominated reaction loop adequate time to perceive the threat, not necessarily to convert that into an action that translates smoothly into the game — a game that feels unresponsive with such slow-motion timing.

Gun-time events in games are often awkward, even when they are anticipated. It sometimes works in first-person shooters simply because we have no expectation of outrunning a bullet, and squeezing a trigger is a close relative to clicking a mouse. Not so with a more kinetic action. Melee combat at half-speed is still too damn fast for the complex dance that the fingers must execute to translate our response to the game avatar.

Why is it that the most effective method for playing most 2D arcade fighting games is to simply slap the controls as quickly as possible? Because it gets inside the other player's OODA loop. Even with a very few controls, hands optimally positioned, and responding to largely cinematic actions, it can still be too damn fast to counter.

One of the great things about the combat in Batman: Arkham Asylum and its sequels is how rapidly Bats engages and dispatches his foes, smoothly aligning to the next target and using that action to set-up the next44. It is simultaneously satisfying (forebrain realizes this is actually fast-paced, yay!) and infuriating. Your responses are largely limited to keeping a rhythm and watching for the visual cues that indicate a different type of attack needing a new counter. But really, are you controlling The Dark Knight here? No. You are giving him permission to attack on your cue, but the context and mechanism of attack is outside of your control except for those moments where you can execute a special action. Those actions often engage some sort of slow-motion to allow your forebrain to perceive it. Functionally, it is as much a rhythm music game as a melee simulator.

And yet, that is a truer combat pacing than most games can even attempt. In fact, I suspect it is so damnably fun and smooth to play because it is a pretty good simulator for your forebrain's degree of involvement in an actual life-and-death battle, limited to a harried "go, go, go, look out!, go, go, go…" as the midbrain (everyone's personal Caped Crusader) does all the work in its own time at a near imperceptible pace.

Part 2: The Bodily Experience and Tactics Overview

Transitioning the discussion from the mental/sensory aspects to the more physical events, we have three areas to examine:

  • The methods of moving the body through a battlescape
  • Samples of battle tactics that arise from the combination of technology, movement and sensory perceptions, and some guidelines on how these may be applied in a game aiming for "realistic" melee combat
  • Supply mechanisms and other limiting factors that will help determine the specific tactics for inclusion in the game.

We will examine these hand-in-hand with the control interface we would use in gaming.

Part 2A: Movement

Whatever control interface we are using — gamepad or mouse and keyboard — we are greatly limited in our options. At best, we are trying to use just two hands/ten digits to control the entire body, marionette-like. Kinetic control mechanisms may make this "more real", but absent the speed of midbrain intervention, the ability to generate momentum through larger-distance movements, and most importantly, tactile feedback from the opponent, it will never be anything but a slow and sloppy mess.

For now, regular controls are better than current full-body simulations. While some specialized systems may give a broader sense of interaction, they are too costly to be used for a consumer-level game, and are outside the parameters of this discussion45.

What we need is a control interface that allows for a broad range of movements with both the body and the arms. A minimum of three separate multi-directional controls are needed: One for the lower body, one for the upper body/arms, and one for head/camera. The existing mouse + WASD control mechanism isn't half bad for this, save that it requires your "movement hand" (usually left) to be doing full-duty just with movement. Any other action needs to delay the movement in some way as the fingers go hunting for the next control input. My own control interface at home is a G13 game board, which allows my thumb to be the WASD interface with my other four digits freed up for other actions46. There are plenty of gamepad/gameboard types out there, but I believe this type is the minimum control mechanism needed for a melee combat game of this sort. While every control input discussed in this article will be possible with a keyboard (or maybe even a console-style gamepad), it will never be as efficient or responsive4748.

We need a minimum of the following actions to be available49:

Lower Body

  • Movements in all directions (WASD or the arrow keys) with some method for differentiating desired degree of movement (double-taps, analog stick, Shift key, etc.). Ideally, we have three degrees of activity available, such as 1) short-step, 2) large step/lunge, 3) emergency dive/dodge/roll (with attached recovery time before any coordinated action could be attempted). For this, an analog thumbstick would be the best option, but I don't know of any gameboards that have them.
  • Jumping actions may exist, though it is my express desire for these to be minimized except as weaponless obstacle-clearing/climbing options, perhaps as context-sensitive controls.
  • Forward-based movements would inherently travel larger distances, with maximum forward pressure indicating a full sprint.
  • Backward movements would have tripping possibilities in cluttered environments.
  • Combined movement toggles could be used to indicate special actions, such as back-then-front dropping to a wider stance with a lowered center of gravity, receiving a charge. These actions would always be triggered on oppositional movements to avoid accidentally activating them through normal/frenzied movements.
  • The medium and maximum movements would have a degree of inertia to them and could not be reversed, though they could be modified through 90° input changes, such as a hard sideways dodge (3rd degree right input) converting to a collapsing dodge (then press back input to let the legs collapse, laying momentarily flat). My goal with this would be to have the left analog stick (or ministick + 1 or 2 button presses) be able to nearly function as a parkour simulator with minimal other input.

Upper Body

  • The leftmost keys (Ins & Del) control the left arm, the rightmost (PgUp & PgDn) the right, and the middle (Home & End) are modifying actions.
  • The modifying button press can be a context-sensitive key as well as a is a "charging" event where the weapon is moved to gain momentum. Hold longer = more momentum/swing. This allows you to tell the game that you intend a hard swing prior to initiating the action to swing the weapon at all. Holding a modifier and then moving (thumbstick/arrows) notifies the game that you intend to couple movement directly to path of action. Perhaps moving then a quick tap of the modifier (or other movement key) would communicate the desire to plant the foot? At some stage, the number of contextual controls will become unmanageable, and to make the game approachable for every player, these options would have to be on a sliding scale, allowing as much granularity of action as the player behind the interface could manage.
  • PgUp = high jab, whip, feint or quick parry
  • PgUp + Home = high quick slash (right to center/left)
  • PgUp + Home + PgUp (either as a press & hold mechanism or triple tap) would bring a high right-left windup for a left-right slash50.
  • Every action is distinct and delineated by specific-to-direction key-presses that are sensitive to both duration of hold and rhythm in combination with other keys. This way, even if the controls are fumbled and the action that arises isn't what the player intended, it also isn't far-off from it.
  • PgUp always brings the right arm up
  • PgUp + Home always moves the arm/weapon toward the right and center, even if only in preparation for another action.
  • Home as a modifier always represents a degree of reserve to the movement being executed. Holding Home indicates the weapon/arm should be contained to an inner perimeter of action and not over-committed. (And yes, there are instances where releasing that restraint is desirable.)
  • Home + PgUp would always indicate a center rise of the weapon into a left-right swipe51
  • A link would exist between the degree of button-mashing and the commitment level of the action. A two-handed weapon receiving a simultaneous Ins + Home + PgUp brings the weapon high overhead, presumably for a power-chop on the Del + End + PgDn… but maybe not, since a simple End press resets the weapon to center, allowing the counter-attacker to skewer themselves… and a rapid release into a quick "finger roll" across PgUp + Home + Ins would instead hew a powerful slash from right-to-left.
  • There needs to be a clear contextual linkage between the positions of the buttons and their results on-screen to help bridge the gap between the fingers and a whole-body movement.
  • Save for modifier buttons, every button-press has a visible, time-sensitive movement event, not all of which are attacks or blocks. Some may simply be baiting for a response, or be pre-movement wind-ups to create the needed orientation for an action.
  • Direct linkage between weapon weight and responsiveness, visible on screen as levering effect. Swinging a heavy-bladed sword will always lead with the pommel, requiring a preparatory movement before the swing commences. Consider that the modifier may be used to adjust the depth of the swing. A two-handed right-left swing that draws the elbows in tight prior to the apex of the swing loses reach and a bit of height, but accelerates the tip of the blade much faster52. Modifier keys allow us to inform the game that this is what we want to have happen.
  • The mouse — operating primarily as a camera-adjustment device, is only marginally responsible for aiming. This allows us to have our line of vision slightly decoupled from the attack's parameters, so that we can track targets in the background and keep situational awareness even as we attack53.

Head/Camera/Field of View

  • Whether the game is first-person or third-person, there will need to be a control for aiming the point of view, and this is traditionally the mouse. I doubt a more effective mechanism will be available for this, and even if it were, the time needed to retrain gamers in its use would be non-trivial. The mouse serves this function intuitively and well if the controls have adjustable depth of field, look sensitivity, and variable acceleration.
  • The mouse wheel could serve the function of shifting from first-person to over-the-shoulder third-person, to a wider field of third-person view, perhaps even an overhead view of some sort54. The user would be able to configure their preferred "bounds" for camera movement to avoid overshooting the chosen view in the heat of battle.
  • While the mouse buttons traditionally control attacks, I've never felt that this was ideal in anything but a first-person shooter. If the arms are manipulated jointly by left-and-right hand controls, this requires significant coordination between the player's arms. (Don't even get me started about how the left mouse controls the right arm, and right mouse is blocking with the left. Arggh, my head hurts….) Instead, the left mouse might control "focal points" that are the targets for the arm actions. Better yet would be to have the mouse buttons trigger context options and unit orders based on them. If the mouse is to be a viewpoint controller, lets link it just to things affecting sight.
  • It is important to note the side effects of various view control schemes. If using the mouse as an aiming device, and not just a viewpoint control, then first-person gameplay will be inherently more accurate, with the hostiles occupying the largest percentage of screen space, relative to any other view.
  • With a third-person mechanism, the player has a blind spot directly in front of them that can only be negotiated be decoupling the camera or avatar model from solid reference points. We could maneuver the camera to the sides, and view the battle from that angle, but a blind spot would always remain, and it would decouple our control interface from its natural left/right anatomical indicators. Placing the camera into a third-person over the shoulder view would still occupy much of the field of view with the avatar model, and creates blindspots to whichever side the avatar is on.
  • Arguably, first-person perspective gives the most "realistic" representation of tunnel-vision by its very nature. A first-person game with a wide (or variable) field of view and depth of focus55 is probably the most realistic and immersive option, but would require some mechanism for shifting view elastically to simulate a head-turn absent bodily reorientation, similar to the "POV hat" used in many flight simulators.


  • The above suggestions could as easily be done with a three-button mouse, with the mouse click-and-press controlling the arms (with a variation of the six-button arm map), or even have the mouse controlling the right arm while a smaller button interface (1, 2, 3, Q, E, for instance) controls left arm options.
  • I have seen many recent simulators that linked attacks (such as sword swipes) with movement, but this is a poor simulator for anything that doesn't "charge" into attacks. While right move + attack could well create a rightward slash, it could as easily indicate a pivot and counter stab (which would travel the opposite direction)57.
  • While it is necessary for momentum to be available to drive a weapon into a target, that momentum can be supplied by either (or both) combatants.

To close this section on movement: Consider the over-the-shoulder resting position, two-handed sword slash (baseball bat swing). Depending on distance/orientation to target, intention and blade design, this could have several possible points of contact: the lead elbow, pommel, doubled fists on grip, basket hilt/cestus, cross guard, forte, parrying hooks58, "sweet spot" of the blade59, and then blade end/tip. It is likely that multiple impacts could occur in a single, rapid path-of-action attack60, using the elbow to create space enough for the pommel to strike, smashing the opponent's head backward61 and exposing the greatest possible body plane62 and finally the diagonal slash that either lays them completely open or chops them nape to midriff63.

Yeah, the control interface needs to be able to do that without it being a preordained context action. It has to be intentional and key-press generated, in this case nothing more than Press and hold Home (tells game to get ready for a directional attack input, starting high), Press and hold PgUp (motion begins from high right hold, so weapon shifts to high-hold/over-the-shoulder posture), press and hold Delete (action proceeds to low left), release PgUp (completes motion indication, blade will stay low left without attempting to return), release Home (consent to attack, any automatic reflex defenses temporarily gone). Upon initiating the movement, the opponent is struck close by the elbow, so the collision physics return to a point similar to the start, quick-tap Home and again (repeat action, low-left direction still pressed), this time the pommel smashes in. Finally, quick-tap Home again to recommence attack and thumbstick back briefly (stage 1 movement, literally one step backward, context of action dictates which leg steps) to open the distance and allow the blade to blast through the target…

Continuing from the fallen foe above, we64 see another attacker approaching from the left65.
Problem: the blade is stuck in the first opponent. He went fetal66 and the blade snagged in his pericardium, bound equally by splintered ribs and errant chainmail links.

The new foe is rushing to make up distance, but is still several body lengths away. Let's finish with a kick to get the victim off the blade. Release all arm controls (indicates entirely new action originating from current arm/blade position) then Press and hold End (indicates low context-sensitive option). Tap thumbstick forward (context is leg-based) and a foot-shaped cue shows up to indicate that the selected option is to kick. Release End to commence and Joe plants his foot on the fallen enemy's face. Press and hold Home and then tap Insert and PgUp repeatedly to "wiggle" the blade out. The game knows that condition "blade lodged in target" plus context "kick" and the indicated arm actions should result in the blade coming free. The animation triggers, all too slowly as the second attacker closes the distance and starts winding up, their own sword brought up and over their right shoulder and continuing to far above their head.

Just as you are about to abandon the blade (via triple-tap of End button), the weapon comes free. You already see that the opponent's overhead swing is commencing and you will not be able to get your sword into position in time. So, you hit the blocking action and crouch in behind the blade, leaving the tip low and pommel high (Press and hold Insert, then Home, then PgDn in the six-button control scheme) to deflect the force of the blow to the right side67. The screen dims a bit on the hit, but the dimming is towards black-and-white instead of red-tinged, so there is no real injury. With the opponent committed into the swing, their momentum carries them forward and to your right. Drawing the right leg back to turn the hips as you stand (Thumbstick back and circle it left), the blade rises overhead, cross guard and pommel framing your shoulders (press and hold Home, press and hold Insert), you whirl to face the attacker and strike high right to low left across their spine and right arm (still holding the prior keys, press and hold PgDn and release Insert to charge the attack's new direction and then release Home to commence the swing). The opponent — bent over and stumbling for balance — is ripped open from mid-spine to armpit, pieces of triceps blasted away as well. As they collapse, you continue turning (thumbstick nudges left and forward circling) to see if there are any more foes in your blind-spot. Reflexively, you go Home + PgDn to have the weapon swing up center with blade downward-facing as a right shielding action, just in case…

That entire battle sequence could be done with just the left hand on the gameboard (assuming a properly sensitive thumbstick control) and only six keys (G13 has twenty-something, but 10+ is standard). The entire time, the mouse is being used to specify targets (point and left-click) and look at new areas, possibly providing movement context in conjunction with the thumbstick. Perhaps mouse click + mouse wheel would allow for dynamically altering stance, posture, level of commitment to movements, context options, or weapon drawing.

At some stage, we have to worry about the controls being too sensitive. A big problem with an interface this integrated is the accidental triggering of unintended actions. Part of the dynamic scaling system should be a selectable "threshold" of responsiveness in the controls.

Further, Joe's midbrain must play some role, possibly executing simple defensive actions or leaning dodges when an incoming attack can be avoided with such measures and no other control input is occurring. We don't want Joe's actions to jar the player's immersion or rob them of a sense of control, but the player (being forebrain driven regardless of Joe's in-game state) should be able to direct the battle without excessive effort. Rolling control motions (just drumming the fingers in a simple beat) should produce simple actions, as would simple "twists" or "circles" of a stick controller.

Part 2B: Technological Relationship to Tactics

I'll let you in on a little secret: Most martial arts are made up of a very small batch of "A-game" moves; Rarely more than five of them. Everything else in a given art is meant to support or disguise the use of those A-Game tools, or create avenues to utilize them against someone who is not in optimal orientation68. Exactly what those tools are — and the tricks for deploying them in less-than-optimal situations — is the key difference in martial arts systems. Two methods may deploy visually similar tactics with wildly different strategic aims. What you are likely seeing are the support tools being brought into play to generate a desired opening for the A-Game methods. Word to the wise: spinning, jumping, rolling, kicking attacks are virtually never the A-Game tools69.

Draeger's concept of the evolution of martial arts can be simplified to this:

  • Begin with a batch of otherwise unrelated tactics that seem to work (fighting techniques)
  • Generate a framework for using these tactics together (fighting system)
  • Solidify the framework with a series of optimized methods for bringing the most effective tools to bear (martial science)
  • Allow natural variations into flourish within the method to promote constant evolution and growth (martial art)70

The core strategies of this game should evolve in much the same way. A basic method of fighting would be developed through gameplay, and just as the character learns new options and tactics (perhaps by first seeing them in use, and/or having to spend time training), so too does the player establish their preferred methods of battle. With each new encounter, the existing toolset is validated or disrupted, and the core methodology may splinter into context-sensitive subsets, or become tighter in its core function, reinforcing the aspects that make it work.

Ideally, the player steers the character's progress so that the two are in sync with one another. The character might be able to spend XP/perk points/whatever mechanic on "automatic" actions (i.e., Joe's midbrain does stuff on its own) allowing the player to "patch holes in their game"71, or augment actions they find themselves using regularly.

In a sufficiently diverse game, there would be many different mechanisms of attack and defense, and a large array of opponents and armaments. If the game were to hew close to reality, some things to keep in mind:

  • Something to consider is that thrusting attacks are much harder to detect than slashes. A thrust aimed at your face betrays little movement left-right-up-or-down. It simply gets BIGGER. And by then, it is usually too late to do much about it. There isn't a lot of body motion to reveal the attack, either. Compare this to a sword slash where a large swinging event can be seen throughout the body, and it is relatively easy to read the attack coming. Additionally, the human body is relatively safe from slashing attacks that do not strike an artery. Slashes will glance off of bone or other hard contours and may lose much of their force getting through muscle and bone so that internal penetration is limited. Thrusts can penetrate deeply and glide or float past bone, causing massive internal trauma. A thrusting action that is not halted by its deliverer will often pass right through the body, skewering the target. Absent any other factors, thrusting weapons are superior killers.
  • Shields paired with spears and/or quick swords are vastly superior to almost any other armament that can be carried. You could come up with ways to discourage this method or enhance others in order to create an arbitrary "balance" point, but all other things being equal, this is the king. Only in multiple aggressor scenarios with foes to all sides does this methodology fail, and frankly, so do almost all others in those conditions. If the game had a sophisticated encumbrance system, and was open-world and exploration-based, the shield might be discouraged from use, but opponents would occasionally have these tools.
  • Dual-wielding (weapon in each hand) is an extremely difficult skill to master, and in the heat of battle, you are quite limited with these tools. You would lack the leverage and reach to defend against spears, have too little mass in your weapons to properly block heavy attacks (though you could rely on quick footwork and deflections), and there is significant risk of one of your arms crossing the path of another in a way that leads to injuring yourself. While this methodology is popular in games and cinema, it is often not a good real-world method. The only martial cultures to use it successfully were unarmored tribal cultures, and their weapons always served multiple survival uses. Invariably, these weapons were hand axes/tomahawks, short swords/machetes, and/or daggers.Though I wouldn't want to actively discourage players (or martial artists) from attempting such a style, it simply isn't effective on a battlefield when other72 options are available.73
  • Having said that, using only one normal-sized weapon against someone wielding two (and without some kind of heavy armor to compensate) is unwise. One weapon against two equal-mass tools is never done right in cinema. If it were, the two-weapon stylist would almost invariably win. Real two-weapon stylists do not double-up attacks on a single angle, nor execute "scissoring" motions or any other similar crap. More often, it will be a slash with a thrust on a completely different angle. Usually, the thrust is the killer, with the slash being large and visible and telling the opponent where to put themselves so that the thrust skewers them. However it is done, one attack is meant to generate a defense, which the other exploits, often to bring both weapons back into play against an undefended and exposed target area distant from the initial wound point.
  • Absent shields, two-handed spears are the most ancient and prevalent large battlefield tools. They are simple to use, work well alone or in massed formation, serve survival purposes (so they are never dead-weight), and penetrate ably light and medium armors74. Someone who has good footwork and a spear is a nightmare to aggress upon. As a matter of practical function, spears are always taller than the people using them simply to avoid accidental self-skewering in uneven terrain7576777879.
  • When fighting against heavy armors with good deflection properties, only the joints are likely to be vulnerable to attack. Contrary to popular belief, most people who wore such heavy armors in battle were quite maneuverable and able to get up rather swiftly after falling. Warriors in both Europe and Japan devised methods to trip up a foe and then lay their own armored body weight on the pommel of a sword to drive it through the joins in the armor, usually at abdomen80, armpit81, or throat82. Femoral83 arteries, perineum and groin are also valid targets, as are the hamstring tendons in some situations.
  • When considering armor, cost, materials, technology and craftsmanship play an enormous role. Nothing of the medieval European era was better than contoured full-plate custom-built for the warrior that included removable segments to lessen weight and enhance mobility. Such suits were expensive in the extreme and the people able to create them were few and far between84
  • Historically, the chain mail hauberk was a favored armor for thousands of years, yet all modern empirical tests call its function into question. It simply wasn't an effective protection system against piercing and even many slashing-type weapons. Popular theory held that plate mail, being heavier and rigid, was more restrictive of movement, but it appears to be at least equal in this respect. The rigidity meant no unexpected shifts in momentum, custom joints did not hinder movement, the suits were often easier to wear since the weight was evenly distributed85. The sole advantages conferred by chainmail over full-plate alternatives were the rapidity of donning it, and the simplicity of tools needed for its construction. While both suits were labor-intensive to create, chain required no great expertise, and could be constructed largely "on the move" at a camp with no forge.
  • Many large historical armies opted for a composite armor that might have consisted of leather-backed chain with a plate cuirass86, probably finding this to be an adequate compromise of materials, cost, labor and protection. Even such utilitarian designs were difficult to produce and required expertise to craft.

The ideal armament and tactics will always depend on the defenses being faced. Spears are exceptional until you have to deal with heavy armor that can wade through the strikes. Then it takes larger, heavier slashing and thrusting tools, which requires closer range to control their use, which demands equal armor for both participants, which usually means that weight has reached a point where shields are abandoned in favor of two-handed weapons and close-range grappling to render the foe momentarily immobile for the killing blow… all of which can be undone by a few mobile, lightly-armored footmen with low-cost stabbing devices and a bit of courage.8788

Part 2C: Supply Systems and Specific Tactics

If we are trying to craft a realistic melee experience, the tactics would be heavily reliant on the setting; the customs, technology, disposition of large forces and supplies in the game world. It doesn't matter if every blacksmith has the requisite skill to craft full-plate in record time if the materials to do so are not available. It doesn't matter that the hero of the story is exceptional with sword and shield in an area dominated by roving gangs of mobile spearmen who have mastered pack tactics. The "correct" strategies and tactics must evolve from the game's setting to be valid.

We want a gaming experience that provides more than one "correct" strategy, without also including a host of ineffective options. Balancing the options to find the "correct" strategies would principally involve the threat profiles that we expected to face, the game setting's own resources (and access to training if this mechanic was used) and the game's encumbrance mechanism.


The most common encumbrance systems for games are a gross-weight method (can carry X weight of gear), and geometry-based methods placing object shapes into an inventory space (common in MMOs and adventure games)89. Since load-bearing in the real world is reliant on both factors (weight and shape), let us use a combined method for this game. Even with this established, the total weight/space is open to a wide range of options, from realistic (and thus highly limited) to wildly unrealistic (to the point where a shape-based method becomes nothing but a visual UI for finding the gear).

Most games have highly unrealistic encumbrance systems9091929394. In order to have adequate gear for all expected threats, we might need (at minimum):

  • A large crushing/chopping instrument for dealing with heavy armor
  • A shield and shorter stabbing implement for medium-mobility engagements
  • A ranged device of some type95
  • Two or more types of armor, ranging from light and mobile to heavy, slow and tank-like.

So, before anything as mundane as food and shelter96 are considered, we are toting around 150-200lbs of combat gear. Even a 300lb gear limit doesn't leave much overhead for "new acquisitions"979899100.

Threat Profiles

Assuming we are aiming for a fairly realistic representation, the combination of supply/encumbrance and setting/technology/culture will determine a range of optimal/"A-Game" and contingency/"B-Game" tactics for Joe/us to use. There may be different styles favored culturally in some areas that need to be adapted for with the use of direct response tactics101, or proactive tactics/strategies of a more comprehensive nature102. Ideally, a combination of reactive/proactive methods are used, with Joe carrying a primary set of tools and some standbys for odd situations, but using mobility, observation and brains to overcome many threats.

Proactive gameplay takes more time and requires the game areas and options to be sufficiently robust that multiple avenues are available. For those of use who always wonder "why didn't the bandits post a guard there?", the available options should be such that the existing enemy forces cannot cover them adequately, creating patrol gaps or other weaknesses (begging the question of why the bad guys chose this place to set up camp).103

To get specific about tactics, we'll need specific threat profiles to draw from.104

  • The game world is split broadly into two primary zones: a central and developed group of city-states with connected travel routes, and a peripheral and wild region of small tribal communities with travel options dictated by nature.
  • The wild areas favor survival-based tools such as medium (throw-able) spears, shortbows (with broadhead arrows, good for hunting but faring poorly against metallic armor), daggers, slings and hatchets. When armor is used, it is hide, fur and leather with some warriors wearing cuirasses taken from fallen enemies, but rarely carried in day-to-day activities. These hunter-gatherers are communal, insulated and stealthy. Their settlements are designed to defend against animal and other tribal intrusions, and everyone is a potential combatant. Training is generally poor as they have subsidence living conditions and cannot spare the time or calorie expenditure for complex, organized warfare training. Their morale is top-notch as they are literally guarding their homes and loved ones. Their senses are exceptional, and many of them will be able to track or note Joe's passage through their territory, so stealth must be balanced with mobility in these lands and Joe will need to pack light to remain quick and quiet. Because every tribe has their own territory and complex relationships, the time and distance any one group has to track Joe down is limited. Potential attackers will likely be alone or in small groups. Living among an area populated by other predators, these warriors exhibit animal cunning and pack tactics. Assume that if you/Joe see one approaching, they are serving to delay or distract you as another gets into killing position.
  • The "cultured" areas have solid metallurgy and supply with complex trade relationships being the deciding factor in what is available to whom, for what price, and at what time. Because most gear is full-metal and custom made, many pieces of armor that are found will not be readily adaptable to Joe's own needs. The threats of the area are split into three general categories: highwaymen (lightly armed and armored, hit-and-run attackers with ready transport waiting nearby), patrolling guards (medium armor of a standard set and style, polearms are the favored tools, able to cope with horse and armor equally, and appear imposing to keep citizenry in line, tactics are oriented towards surrounding a foe to bring their slower weapons to bear and they understand principles of zone control, so they will take their time to develop a killing positional arrangement), and finally professional soldiers (acting as weapon-system components with specific tasks, creating a coherent threat matrix that no single individual could cope with; speed and rapid killing actions would be needed before they could bring their full weight to bear, and their commanders would be primary targets of opportunity to disrupt their coherency). Civilians are of no threat, and settlements rely on fortifications with multiple traversal points for trade and day-to-day traffic, so defensive access is meaningless on an individual level. Battles here will be heavily influenced by considerations of legality and allegiance, and the environment has many hard obstacles that can favor a more mobile, close-range approach to warfare. Conversely, travel by horse/wagon is viable, and so a broader range of gear is available. Further, the developed nature of the area means that survival needs are more easily met through trade and barter, so a larger volume of inventory space may be devoted to gear, if Joe so desires.
Tactics for Consideration

Each of these two environments will have a few optimal (and many definitely sub-optimal) tactics available. Perhaps Joe finds a set of options that can work in all of these, or focuses on specific tactics for each. Some considerations:

  • The tribal lands favor mostly thrusting/penetrating weapons in cluttered, low-visibility environments (due to flora). A closed-face helm is a must (though peripheral vision may be needed in some instances, so a hinged-face helm is a better option). Armor should include a cuirass, pauldrons, and perhaps a brevet/gorget, but leg armor would get easily caught up in brush and uneven ground105, and gauntlets are largely unnecessary. Because Joe will have to move forward or away from most perceived threats, packing light is a necessity, and greaves/heavy boots are a hindrance. A back-plate would also be unnecessary. In terms of armament, the natives have the right idea. Polearms are obstructed, shields would have to be smaller to avoid getting caught up on brush, two-handed swinging/crushing weapons serve no survival purpose and are too slow for this environment. Medium spear, hatchet, dagger are the way to go for melee tools. A shield may come in handy against arrow fire and hostile spears, but it would have to be considered against other encumbrance factors.
  • The civilized lands offer and require a larger set of options. Joe could try and adapt the "tribal lands" tactics to this environment, perhaps meeting with some success, but his offensive and defensive arsenal would prove inadequate in many cases. Spears, daggers and hatchets will fare poorly against the developed culture's plate armors, and incidental impacts from the standard weapons of this area are likely to be bone-crushing without metallic armor protection. These battles could occur in multiple ranges, cover may be easy to find or unavailable, mobility may be restricted by several factors in both "open" terrain and cities, but there are avenues for high-mobility tactics if Joe can seek them out. Disposition of threats may vary widely. There is no option to "do as the natives do" because the natives have no single tool and must rely on specific measure/countermeasure options. A shield is a must, sturdy armor is a great idea, but may hinder escape in environments where attacking threats multiply rapidly (guards call for backup, highwaymen's allies flank and ambush, patrols surround, delay and wait for more patrols). There is no perfect strategy in this environment, but multiple methods for traversing the area allows for proactive reduction of some threats.
  • Put simply, Joe cannot possess a single tactic/weapon/armor set for all situations, but may be able to develop skill in tribal methods and then augment those options with specialized tools to leverage those same skills in the civilized regions. Conversely, Joe could have two separate methods for each that share only a few skills and tools, or simply eschew a singular/modular battle strategy altogether and become proficient (if not expert) with whatever tools are readily available at the time, making small preparations prior to every expected battle and existing in a perpetual countermeasure philosophy of battle.

From a (fun-seeking) gameplay perspective, we don't want Joe to be rendered helpless by well-considered prior choices that turn out to be ineffective in random/unpredictable situations106. Whatever methods Joe develops, a perk/experience/midbrain system should exist to help counter threats that fall somewhat outside the tactical envelope. Fundamentally, there is a single tactical stratagem that works pretty much regardless of weapon or environment: Collide and grapple.

In this way, the weapon attacks are meant to produce openings for decisive close-range finishes. It is important to recognize that unlike every damn movie out there, enemy #2 will not be waiting to see what happens to enemy #1. They will be charging into the fray with killing intent. Thus, timing and distance manipulation need to be mastered to produce the needed opening with enough time/room to complete the attack before enemy #2 can cut you down.

This theme can best be viewed with disarmament tactics, and thankfully YouTube has no shortage of them to analyze. (I have selected disarmament as a focal point because the reader knows what that is, and there are plenty of examples of it in games and cinema. This should not be read to indicate that disarmament is a major focal tactic in melee battles, though it could be.)

Disarmament in the game would be a contextual option based on close-range grappling107. Disarmament is almost always an offense-based action triggered by tactile reflex, where the opponent's defense is placed into proper geometry when it counters a certain angle of attack. Throughout the descriptions, I will be using the term "attacker" and "defender" with the "attacker" always being the guy executing the disarm.

  • Notice how kinetic this engagement is? They are coming together with some force, and the moment the blades lock, the attacker's support hand is doing its thing, threading into the open space with a gross-motor action. Without losing any momentum, the attacker is already pivoting. Note that the key piece of movement that makes this whole thing work is the attacker pulling the arms/weapon frame in conjunction with a pivoting step that causes the defender to place their foot into a vulnerable position for a trip.108 This is an excellent disarm example showing the clear technique of western medieval era sword tactics. Should the move fail, the foe is still tripped, shoved or off-balanced with their back facing the attacker. Even failing all of those things materializing, the pivot step takes the attacker off the line of charge and could as easily be used to maneuver in a way that keeps a second enemy on the opposite side of the defender.109.
  • Note that this type of technique is most common with European methods. While these kids aren't combat experts (and are hamming it up a bit for fun), it serves to demonstrate how functional midbrain-dominated grappling actions can be. The initial wrapping motion is viable in an armored environment, and is reliant on gross-motor actions to succeed. Note the back-and-forth counters involved, each of them a viable technique in their own right. The countering disarm executed by the defender is a simple grab to the extended handle of the attacker's sword, but was anticipated in this case (the sword may have been meant as a distraction technique in any event). I suspect this derives from a thematic battle where the attacker's sword was broken. Something to keep in mind is that this tactic would still work if both were furiously charging and smashed into one another. Collision creates the desired tactile OODA loop that allows this to occur very rapidly (I would say they are going about half-speed).
  • A very straightforward samurai-styled disarm. Nothing much to it, yet highly effective; a hallmark of the best tactics, and not at all cinematic. Again, collision creates the needed range and this would function well with gross-body action. The finishing blow is too short and would likely not function well in a heavily-armored environment, but against a foe with no head or neck protection this would be excellent, and the sword may get disarmed either way.

Recalling from Part 1, grappling methods are far more common and reliable than cinematic treatments of melee combat would have us believe. Accordingly, many of Joe's tactics will be geared towards colliding into a range where these context options materialize. It would be up to the combat AI to avoid letting that happen, while simultaneously trying to fuel killing attacks with momentum in order to generate its own contextual attacks. Distance manipulation, baiting, feints and timing would play a major role, and as the level or armor increases, so does the validity of these sort of disarmament tactics. Note from the above clips how the more western methods often smashed into one another on the charge. This would be something to model accurately in the game while not turning it into a "smash and win" tackle-fest. Getting tripped in this environment is a severe hazard, and if you are the recipient of such an action and the camera perspective is tied to anything but a free-roaming 3rd person, such actions would rock your situational awareness badly. Oftentimes, a multi-axis takedown on someone who has tunnel vision is as effective as a cloaking device. In the time it takes the thrown individual to search for their foe's position, there is no shortage of carnage that could be delivered via sword, spear, shield or dagger.

And consider also how the game would model these interactions: Much of this is not visual, and in a first-person perspective would not be on-screen without the game steering your view, which is both immersion-breaking and damaging to situational awareness.

Implementing the Core Strategy of Attack

All of this this is not to say that basic attacks would be merely opening actions that had no lethality of their own. Good timing and orientation paired with accurate deployment could make a "set-up" action lethal all by itself. A foe charging into grappling parameters could be side-stepped with the sword thrusting deep into the gap between gorget110 and cuirass111.

A fully developed fight strategy would maximize the lethality of the set-up options while ensuring that the attack pattern forced the foe into unsuitable defensive posture or caused them to have to close range into the grapple, where your attacks have already laid the groundwork for their demise by forcing them into parameters that you have trained to be optimal for your armament and skills.

I have a high art; I hurt with cruelty those who would damage me. — Archilochus

One of the biggest hurdles in real-life combatives instruction is getting the student NOT to focus on the movement being executed, but instead to think to the next action, their surroundings, and be in perpetual motion. Too frequently, people engage in "posing" where they complete an action, adding flourish, watching their own handiwork unfold and wasting valuable (life-saving) time. This is a completely natural phenomenon that combatives instructors have to constantly attempt to decondition from their understudies.

Consider how this translates to a game: We want to see our handiwork because it is gratifying and fun. A battlefield expert wants only to turn around, seeking the next target and finding none, with all foes dead or dying on the ground. The expert defocuses within a chaotic symphony of death-howls and gurgling, attacks and countermeasures narrowly perceived, felt but unseen, acted upon instinctively without ever being fully comprehended, furiously seeking the next recipient of their "high art" until none remain to engage against112.

This leaves no time for admiration or pride in one's efforts and runs contrary to expectations of gameplay. There is no time for exultation, no slow-motion kill-cams, and the sense of gratification exists only after the fact.

Perhaps a post-battle "recording" could aid in the self-aggrandizement department. An "afterglow" recollection could be part of the skill-training mechanic with key presses being executed at particular times to indicate a desire for improvement in some facet of the battle. In this way, the fun-cam could serve as a post-battle assessment. I would imagine that if the majority of fight scenes in the game were properly cinematic and enjoyable on reflection that this would be a useful mechanism, but I fear as a player, I would grow bored with having to relive the events time and time again when the ultimate aim of gameplay should be what this battle accomplished or opened the way for; i.e. the story.113

Maybe a more suitable pacing mechanism would be to have such recollections "trainable" only for short, expertly-executed sequences, and the longer, more involved battles would have a "review" option, but would not be compulsory and would not be required to review for skill advancement.

Tactics We Won't Be Emphasizing

To close this section, let's examine a few very commonly seen tactics that would definitely not be part of a "realistic" model of melee gaming:

  • Deflecting arrows with shields: Strictly, a shield has too much inertia to act as a reactive mechanism against arrow-fire (the arrow's profile remains invisible until it is well inside the visual OODA loop), and it is up to the archer whether or not a shot strikes shield or body. Proper shield positioning can reduce the likelihood of this event, but never completely eliminate it. Real-world archery isn't accurate enough in a melee battle to place shots so effectively on a moving target, so this is largely random. Coupled with some decent greaves and a good helm, a shielded warrior may be virtually immune to arrow fire, but by nature won't be able to catch the archer either (since archers are lightly-equipped skirmishers and not stupid enough to hold their ground against a charging, armed and armored warrior).
  • Kicking: Vastly over-used in cinematic combat. In a grueling, armored battle, balance is hard enough to come by with both feet solidly planted. We are unstable beings at the best of times, and our bipedal structure is easily compromised unless both legs are maneuvering to align against the directions of force we meet. Further, consider how foolish it is to place your leg in a position to be deflected by a four-foot razor blade. There are few instances where the payoff would be worth the risk of such a movement. To deliver a solid front kick, you must be firmly planted (read as: stationary) or walking forward. Not running, charging, pivoting or avoiding, and not aligned to deal with incoming force of any sort. Turning (roundhouse) and side kicks are often impossible to execute in armor due to joint limitations and load-bearing against the knee which does not fare well under such conditions114115.
  • Jumping: I'm not much of a jumper to begin with, and have never worn full-plate armor, but I can tell you from experience that loaded under class IV tactical gear with a trauma plate, my vertical clearance on a jump wouldn't be able to avoid a cockroach, let alone a sword attack or similar low threat. Even if I could gain adequate altitude, so what? Going straight up only to drop back into the same place I started is hardly a good use of time or energy. To hurl that mass in a balanced fashion through the air towards or away from something, I could only expect to collapse on the landing. I doubt my legs would enjoy having to arrest that much unbalanced mass, and with armor (and armament) being inherently top-heavy, it isn't plausible to have the legs leading the jump at any stage. The only reasonable ending to that situation is a collapse and roll into a low, semi-stable posture. Not useless, but its necessity represents a near-unforgivable tactical failure in allowing circumstances where that bullshit move was the best thing I had available. At best, such actions would be included in the game only as "midbrain saving throws" where Joe's amygdala takes the wheel for a hail-Mary survival action and leaves the player with a mess to clean up and scant seconds to do it in. Good for drama, perhaps.
  • Spinning: A full step and turn-in-place is fine and natural, and often necessary when ablating indirect collision energy. "Spinning around" is really an about-face with a 180° maximum intended rotation116. Anything intentionally done with more than 180° of turn is silly. It removes your eyes from threats, cannot be executed in a balanced fashion in a grappling encounter, and inherently takes your arms out-of-play for a moment117. In other words, during that half-second or more of spin (which is popcorn and a movie time for the foe's midbrain-driven OODA loop), you are defenseless and your next action is occurring on a predictable path of action, easily defended. Oh, and by the way, that action is most often a backhand motion, so it is likely that you will complete your pirouette with the foe stepping behind you. Good luck in your next life!

Part 3: Putting it All Together

How would we go about producing discrete tactics that were compliant with what we understand about the reaction delays our interface will impose upon lightning-fast techniques governed by processes and senses we cannot use to play the game?

(Because the problem itself is such a jumbled mess, don't expect the suggestions here to have a great deal of coherency. I'm spit-balling again.)

Context Cues

In the example battle given in Part 2118, the Home and End keys "automatically know" what context options we want. It is rather convenient in that battle scene, but what about another, quite different situation?

Something many games have done well is the use of context melee options. It isn't hard to presuppose that getting the blade free might be an important option. There are many equally practical and necessary tactics available, but how do we determine which ones to use?

Andre Panza's Kick Boxing allowed for the creation of a unique set of movements for your fighter, drawn from a large batch of them. This, or a similar perk-based mechanism would work well in a melee combat game, allowing the player to generate a special context arsenal that fit their play style and triggered on the events they wished.

There would be an upper-end to the number of such context actions that was beneficial without their trigger events tripping over one another. While a player could choose to invest heavily in these so that a truly wide array of options existed, almost randomly selected as the context options showed up, it would make for disorienting gameplay if each possible option had both granularity to their success, as well as potential perspective shifts. Just like in real-life martial arts, you really only need a few such techniques and the excess just detract from training time (XP/perk points) and trigger unreliably (harming situational awareness).


The game would need excellent collision detection with depth of penetration being a factor in wounding. It would go hand-in-hand with a solid grappling mechanic to create "takedown and impale" options, and having many "weapon bind" events use the legs or forward momentum to trip up the foe.

Most grappling is just a push/pull exercise with some part of the opponent's balance compromised to prohibit corrective adjustments in pressure. If the animation/context system were solid, a very similar directional input scheme could produce wide-ranging but predictable and actionable results.

Section added 27JAN2014

The Game Mechanics of Injury

Most games default to a few systems for tracking damage from melee combat:

  • The Health Bar Mechanic, which presupposes injuries have an immediately noticeable and measurable effect, and generally that the majority of wounds are neither trivial nor fatal. Reality: Most wounds suffered from a blade or piercing weapon in in melee combat bleed a lot. Even if the wound itself isn't debilitating, blood loss certainly can be. When dealing with armored combat, many otherwise dangerous wounds can be rendered trivial and would more likely effect stamina than health, simply wearing one down under weight of repeated blows. For a health bar to be relevant, it has to be utilized in a game where the injured party can immediately assess the damage they have received, and that the majority of hits are non-fatal. Though the mechanic itself is not a problem, its use in a realistic game environment serves to eliminate medical management tactics, such as keeping an injured limb in a position where the bleeding can be controlled, but that doesn't remove the injured limb's functionality in the selected guard or tactic. The health bar is — in my opinion — best used to measure blood loss only, and not some accumulated wound measurement. The use of a heartbeat sound and some red bloody-red flashing at the screen edges would serve to indicate this without adding a bar to the display.
  • The Binary Wound Mechanic is where wounds suffered are fatal or immediately debilitating to the injured body part. While this system is somewhat more realistic for large weapons, it largely ignores stunning effects, smaller weapons, and the very real damage reduction (not outright elimination) from armor. It also doesn't address blood loss as an ongoing fatigue factor. Of the systems available, it is adequate as long as the collision modeling is properly done.
  • The Call of Duty Mechanic, where all wounds heal to full capacity over a short span of time. In my opinion, this system works excellently as a stamina-based method. Most games use highly unrealistic systems for "healing", so this is the most honest treatment of the gamey-ness of such mechanics.

The optimal system (to me) would be a combination of these. The Health Bar option (rendered as some heartbeat/pulsating audio-visual cue) would serve to indicate blood loss while the majority of wounds fell into a Binary Mechanic, and stamina-effecting / perceptual separation events could be handled with the CoD mechanism (combined with blurred vision, deep (maybe ragged) respiration, and stumbling around).

Of course, none of these mean a damn if the collision system isn't good…

Collision Physics

One of the immediate problems to overcome with a wound system is the physics and collision modeling. This inherently carries another problem with it of how to communicate the results/effects to the player. No matter how well-developed the method

  • Modeling granularity of result, where an object "in the way" is determined to be partially effective as a defense. Many games simply have a block/no block mechanism, which does not properly deal with the staggering forces that can be levied against an otherwise adequate defense. How much force a block can deflect or contest relative to the receiver's stance and stamina is both difficult to accurately model, and its success or failure cannot be confusing to the player. The results should be transparent enough to be anticipated and planned for.
  • Depth of action and how to communicate this meaningfully to the player. How far away is the target? Is there an optimal range to be in? For both swinging and many shorter thrusting weapons, the optimal range is along the center of percussion, roughly the same distance from the grip to the pommel as from the tip to the vibrational node or "sweet spot". Striking beyond this still produces a solid impact, but may also stun the arm of the attacker. Striking closer than this point reduces velocity and cutting power, and eventually as the range closes, cutting stops being the most effective use of the tool, supplanted by a myriad of alternate methods (which in a game would almost have to be context-triggered; there just isn't any control mechanism sophisticated enough to manage the sheer number of options available at this point).
  • How to model and communicate more elaborate, close-range situations where weapons/lines of attack may be entwined or overlapping, for instance, when a too-close strike is blocked, only to have the attacking weapon give way for a pommel strike or flip into a new line of attack.

A martial artist might well spend more time on such close-range, overlapped approaches than on their basic strikes. It would be practically impossible for a game interface to capture anything more than a small fraction of such options. When these appear in games at all, they are essentially "kill moves" triggered at random.

And yet, the greatest breadth of material — the real meat of many fighting systems — is found right there.

Wound Effects
Post-Battle Medical Management

There needs to be an unintrusive way to inform the player of the degree and type of injury their avatar has sustained. Certain injuries (decapitation, dismemberment, fatal impalement) leave no gameplay options and could result in whatever cinematic prcess the designer wished to implement.

Some display mechanism needs to exist to give information about injuries and their general characteristics, such as fractured or broken bones, incapacitated limbs, bleeding or any number of life-threatening torso injuries. There isn't a lot of value in telling the player that they have sustained a torso injury that they have neither the capacity nor tools to repair, and simply waiting a long time for them to die. May as well just shift these injuries to the more cinematic methods of expiration or quick fade-to-black.

Ultimately, how far a game designer chooses to take this mechanic depends on the setting, time intervals between game events, options for healing, etc. With longer time intervals at hand, we can deal with realistic recovery, but also can confront infection and disease as wound consequences. The limits to how far we can take this depend entirely on what the game is trying to accomplish.

Healing and long-term medical consequence are some of those "gamey" things that may be best left that way.

Returning to the immediate interface, less severe wounds might need attention from the player when an opportunity presents itself. It is possible119 to apply a tourniquet to oneself in the heat of battle before facing yet another attacker. It would not be necessary to have complex medical interactions for addressing the avatar's own injuries, but it would be very interesting to have that same option for stricken allies — the ability to diagnose and apply various medical treatments to them.

If we have this super-accurate wound effect model… why not really use it?

In modern battlefields, the most common injuries that can be treated in the field are bleeding, airway impedance, tension pneumothorax and shock, which the US Navy SEALS helpfully acronymed as "BATS". Though some of the specific treatments for these may not have been available in a medieval battlefield, the basic field-treatable injuries would still exist.

Given that we know people don't immediately expire from bleeding injuries, it would be interesting in a sandbox environment to have the choice to save a fallen enemy. Though it exceeds the bounds of this article, it is certainly within the present discussion-space about video game violence. These "nameless, faceless" enemies may pop up giving us no option but to murder them, but with adequate ability, we may be able to save a few as well.

And if we are going to go down that path, then it would be silly to inflict a bleeding injury on a foe and not be able to offer them a chance to surrender…

I hope the future of games120 gives us such a scale of options to work with.

Section added 29JAN2014

(Partial) Bibliography

Artwohl, Dr. Alexis & Christensen, Loren W. Deadly Force Encounters: What Cops Need to Know to Mentally and Physically Prepare for and Survive a Gunfight, Paladin Press, 1997, link

ABCs of Street Attacks, Tony Blauer, 1990 (?), link

deBecker, Gavin, The Gift of Fear: Survival Signals that Protect Us from Violence, Little, Brown & Company, 1997, link

Draeger, Donn F. , Smith, Robert W. Comprehensive Asian Fighting Arts. Kodansha USA, 1981, link

Grossman, Lt Col Dave & Christensen, Loren W. On Combat: The Psychology and Physiology of Deadly Conflict in War and Peace, PPCT Research Publications, 2004, link

Quinn, Peyton, Real Fighting: Adrenaline Stress Conditioning through Scenario-Based Training Paladin Press, 1996, link

Siddle, Bruce K. Sharpening The Warrior’s Edge, PPCT Research Publications, 1995, link

Sprague, Martina. Fighting Science: The Laws of Physics for Martial Artists Turtle Press, 2002, link

Section added 30JAN2014

VR Control Inputs

(And why these will be nothing but a disappointment for melee gamers.)

There has been a lot of discussion about VR gaming of late. Rightly so with the Oculus Rift on the horizon, and many other prototype systems being brought into the light. Because of how visually linked a gaming interface is, it seems a natural thing to think of the monitor/TV as a limitation.

In my opinion121 this is a dead-end for melee gaming.

We've already seen how different the OODA reaction loop is for the forebrain and midbrain. Playing a game through an interface — no matter how compelling — is still primarily a forebrain activity122. It requires you to not only perceive the events of the game, but to reason what those mean and how to (unnaturally) translate those movements into key presses.

Just adding your head/vision to another level of reality isn't going to alleviate that. While it does remove the need for a head-tracking/view control (essentially making your neck into a control device for the camera), it also decouples vision and movement. You will now need some additional keys for bringing body alignment to match camera alignment. For me, this would be an incredibly distracting event.

Let me explain: In a multiple-adversary battle, one big advantage of making contact with enemy #1 is that you can now look to where enemy #2 is without impeding your reactivity to enemy #1 — after all, you have a tactile OODA linkage to work with there. You can feel what that guy is trying to do, and it lets you take a gander at where the next threat is coming form. Done expertly, you will always be looking at the next danger in line as you deal with the one you are in contact with.

Within a game, if the avatar's movement is linked to camera position, then using vision to seek targets during an engaging action is not possible.

Having created this linkage between avatar body and player head movement, are we going to allow the avatar to pivot as rapidly as we can whip our heads around?123. If not, then how will we know when the avatar is properly aligned? With an FPS, we would probably see the arms come into position; Arms that we cannot easily see around without additional key presses/movements. One of the big advantages of 1st person view is the lack of avatar-induced-blind spot, but now we have two of them — one for each arm. Add a shield to the left arm and much of the view is filled with useless visuals. Consider that the Oculus Rift and similar VR sets aimed at the consumer market are going to be splitting a 1080p image in half, and that there is no peripheral vision124 to be had with a "goggle-like" viewing field and we are losing as much as we are gaining.

(Many martial arts systems make a habit of attacking through areas blocked from view by their target's own arms. Such "zones of obscurity"125 are critical to manage during a battle, and any competent AI opponent would be programmed to initiate attacks into these areas.)

Without tactile feedback, the needed processing speed for real battle actions won't be there. We are responding to attacks and injuries to the avatar using vision and maybe some auditory cues. Even with the midbrain engaged, we are still at the longest version of its OODA loop, and can be defeated through attacks that the AI should be able to feel its way through. How do we know that a given context/grappling action is available if we cannot feel it? Visual cues are nice, but needing a longer reaction loop during a movement inherently means the movement must slow down to allow for this event to trigger.

Imagine watching someone tackling another person. Imagine now that in the moments of contact, everything slows down to allow the guy getting tackled to visually perceive the event. Once this "visual window of opportunity" closes, everything returns to normal speed. Now picture every combat event you experience in a game going through this same rubber-banding quick-time event. Gamers generally hate QTEs, and now our melee interface requires one on many — if not all — ongoing combat actions.

We are very sensitive to when movements look unnatural. Keeping these events from being unnatural in appearance would require the entire thing to enter a slow, dreamlike state. Personally, I would like it if the game had this option, able to run at a slower pace for some gamers. But many people would hate such a mechanism if it were forced upon them, and the need for these time dilations would become grating very rapidly.

So, do we remove all such tactile options from the game? Create an invisible collision boundary through which the actors cannot move? If our goal is realism, that definitely isn't it.

Stretch Reflex Can Be a Bitch

Let's take this VR thing a step farther and add a kinetic interface so that our body movements are visible to the game. Through practice with this mechanism we start to get very quick at dealing with targets, and our midbrain begins to pick up its designated role in this process. Block, dodge, parry, riposte, all done at a very near-to-real speed.

Then, suddenly, you lash out at a target and feel your biceps disconnect. The pain hits around the same time that the bulge of torn muscle tissue rides under your skin to smack into your shoulder. Sounds terrible, huh? It happens sometimes in combat training when going too quickly without anything on the other end of the movement to absorb the outgoing force.

The midbrain has access to a function called stretch reflex wherein the antagonistic muscles are relaxed during a movement to increase acceleration. Sounds great, right? Anyone who has walked on a slippery surface, momentarily lost their balance and had their leg shoot out way ahead to catch the fall has experienced stretch reflex. This is a move we couldn't consciously perform in time, both because of reaction speed and the physical speed needed to get the leg there at all. That ability is denied to the forebrain because it has a very high potential to outrun the elasticity of the antagonistic muscle fibers. In other words, midbrain perceives threat, moves rapidly, and the muscle may tear. If this stretch reflex happens once a year, no big deal, the muscle tears will be small and easily repaired (hopefully). Fifty times a day playing your favorite VR melee sim? You'll be visiting the emergency room in no time.

Speak with a martial artist from a method that emphasizes speed and ask about torn biceps. It happens because there is no object to strike, and the full force of the outgoing motion has to be arrested entirely by antagonistic muscle tissue. Even for otherwise athletic and healthy martial artists, this can cause injuries of the sort that require surgery to repair.

I would expect exactly these same injuries with any VR interface that didn't have you actually hitting something126. If your midbrain starts to think this VR represents a real threat (and it can start to perceive things that way), then it may engage stretch reflex to counter the threat.

How do we keep this from being a problem? Just make an interface that doesn't look convincing or immersive.

What do you mean you wouldn't pay for that game?127

More than any other factor, it is this kind of thing that will keep VR melee games from materializing128. I suspect that this is the underlying reason that the Wii and Kinect interfaces are so damnably slow to my sense of timing. I go too fast and the game doesn't read it. If the game did read moves that fast, it would ultimately be encouraging exactly the kinds of injuries I am talking about, so there isn't much reason to "correct" that design fault.

There is substantial liability involved in teaching the midbrain to react on this level, to say nothing of a very real case for creating "murder simulators". There are only about five hundred things I would love to do with a knife in a VR sim that I can only shallowly and non-intrusively perform on fellow martial artists. There are maneuvers combatives practitioners cannot do at speed because we will kill or cripple our training partners. Takedowns that drive a blade into someone on the fall cannot be practiced, even with a rubber training knife, because of the massive amount of impact force being driven into the point of the blade. At the very least, it can cause internal injuries and broken ribs. There are real-world cases of death resulting from such training.

Would I love to have a VR game that let me do these things? Yes, I would. Do I think this would be a good thing to hand out to the public at large? No.

A couple of closing observations:

  • A lightweight kinetic controller still doesn't translate inertia properly (empty-handed "sword" movements are too swift, maneuverability is limited to the confined space and is more responsive than armor would allow while also being unable to generate the desired lateral movement).
  • With VR headsets, depth perception cues may be perfect at one distance and terrible at another. Even in real-life it takes a few probing actions to be sure of distance to target, measured by a combination of footwork, visual guides and tactile feedback from "swatting" at the other guy's guard. The professional fighter's "feeling-out" process is nothing but an attempt to get the midbrain to feel comfortable in a non-grappling exchange.
  • To get the needed speed/reactivity to make "real" combat tactics functional requires a minimal forebrain reaction loop, best achieved by pre-optimized control orientations. In other words, hands already touching controls that are instantly responsive, i.e.: mouse and keyboard/gameboard. Anything else requires the forebrain (and increasingly, the midbrain) to respond to these new sensations at a pace that the body normally reserves for life-and-death encounters, and as mentioned above, this easily leads to injury.

Section Added 31JAN2014

Why I Am Sooooo Tired of Stamina

The ubiquitous stamina bar! Fixture of semi-realistic melee games the world over. I would love to tell you that it has no place, is one more little thingy on the screen that can be eliminated in favor of more blood stains, but stamina is a real thing, and is often decisive in battle.

(I still hate the bars, though.)

Okay, so stamina is real, yes. But so is flatulence and spittle and not saying the right things under stress and acne and greasy hair and impotence. We go out of our way to avoid these things with our avatars in a game world, so why hinder the action with a mechanic that exists only to promote weakness in the hero?

I mean, I get what it is trying to accomplish. It gives the illusion of another layer of possible gameplay, another statistic to make gamers wonder where to stick the experience points, etc. But stamina isn't all it is cracked up to be.

Muscle comes in two basic types: fast twitch and slow twitch129. Slow twitch would be high-stamina muscle, able to operate for longer periods of time at a moderate level. The other kind of muscle — fast twitch — is where power comes from. It fatigues rapidly, but packs a wallop. And while a person can train themselves to a very high standard, having both power and endurance, you reach a point of diminishing returns very quickly.

Look at high-level mixed martial artists: At the championship level, they are expected to be able to perform for five rounds of five minutes each with a one minute rest period between rounds. Many of the knockout artists (when they fail to smash the consciousness out of their opponent) visibly slow as the fight goes into later rounds. They are reliant on fast twitch muscle for their power, but when it fails, they become sluggish.

Conversely, there are fighters who are cardiovascular machines, able to compete at the absolute top level for not just five rounds, but potentially many more. There is little noticeable difference between their quickness at the beginning of the fight as the end. After 25 minutes of intense back-and-forth exchanges, they seem none the worse for wear, often ending a round with a backflip or similar showy move to demonstrate that they have energy to burn, even after a performance that would leave most humans dying in a pool of their own vomit.

And these high-endurance fighters frequently get to showcase their cardiovascular capabilities, because very few of them can knock anyone out.

Stamina and power have an unequal co-existence. Gaining one means losing some measure of the other. Would you want to build them equally? Not generally. Render it to simplest terms: would you rather end a battle decisively now, or in a slower fashion much later? If I create a character with massive killing potential, then stamina hardly matters. Everything I was fighting died in one or two blows, and if I am facing so many targets that I cannot kill them all before stamina is gone, then what were my odds of taking them in a longer, slower battle anyway? If I create a character with high stamina…. um… that isn't a winning strategy if it takes forever to drop an adversary. The enemies are able to gang-up and outflank because I cannot dispatch them rapidly enough to prevent it.

Stamina as a Gradient System

If we are going to use stamina as a mechanism, then let's look at how it should be implemented. It isn't on/off. Lets say for the first half of your stamina reserve, you are moving normally, and then you gradually slow as you tap into the second half of it. By the "end" of it, you are very slow to move, and may not be able to fight. Standing in a defensive posture for twenty seconds isn't going to allow it to come back. You're going to need hydration, food and a few hours (or days) of solid rest.

What things would drain your stamina? With adrenaline/high-stress state, absolutely everything. Walking, running, swinging a weapon, looting, having to focus on the battle and give orders, standing still with gear in hand, restraining the midbrain from its more unsavory mid-battle desires130, breathing… everything. So it is a countdown of sorts that limits the duration of a battle scene. And fatigue usually isn't a factor…. until it is. By the time you know your stamina isn't going to last, it is usually to late to do anything about it.

If a game is trying to utilize stamina in a realistic sense, then it is subject to dozens of factors131. Maybe it should be implemented alongside systems like measuring the caloric value of your avatar's breakfast, and how many of those calories were complex carbohydrates? Because for stamina, that shit matters a lot. Are you at altitude? Stamina plummets (unless you are used to high-altitudes, then you would be fine for a while). What about humidity? Are your clothes conducting body heat in a way that assists perspiration or is it in some way inhibited? For stamina, that matters. And if we are going to ignore 98% of the things that go into stamina… are the remaining 2% worth the trouble it takes to program them?

For stamina to be meaningful in a realistic sense would require mind-numbing logistics to be taking place all the time.

So why use it at all?

The one argument I've repeatedly heard for stamina is that it limits how often your avatar can execute "power" moves. Most such power moves are inherently cinematic. and very dynamic, eating up lots of space and time. You know what else limits cinematic moves? Deadly competent enemies.

Stamina Bars = Poor AI

Stamina as it normally exists in games is a limitation with little practical purpose. It is a little green bar132133134 that attracts the eyes to it and causes the forebrain to have to juggle a few extra, largely meaningless bits of data instead of surveying the important details that should be abundant on the battlefield.

What should you be watching? Let's start with the foe's eyes and body movement.

  • What are they doing, why are they moving that way, are they concealing something, and if so is it a wound or a weapon?
  • If they are moving this way unconsciously, is there some way to turn that into an advantage?
  • If on purpose, is this movement indicating something about them? Are they diverting attention from another foe? Is this whole thing just a delaying tactic, and are you paralyzing yourself into inaction by wondering about all of this when cutting them in twain is the ultimate goal anyway?
  • Perhaps the mutual hesitation means you may not be foes after all? Are you both really tired?
  • Ooooh, the stamina bar is full135!

On the above list, something doesn't fit.

Among its many other (unimportant and largely unrealistic) effects, the stamina bar exists to distract you from the idiotic actions of the AI's extremely limited capabilities and lack of human affect. If they were true opponents of any caliber, you would want to study their movements, discern what you could from them, and try to turn what you have seen to your advantage. But instead, you look to a stamina bar and see if you have enough juice to take them on, or if you should hold back and wait for a couple moments.

And if you have to wait, it isn't because you're tired, but that your avatar is. To hell with your own choice of pacing, or narrative, or any sense of excitement you are currently experiencing, obey the136bar.

Let's Raise the Bar

Stamina bars, focus bars, mana bars… every one of these exists to limit what you can do with your avatar, mainly to stop you from mercilessly steamrolling the inadequate AI adversaries.

Let's picture that we have a "jab" bar that limits how many jabs we can take at an enemy before we are all jabbed out. A competent enemy probably would not let you jab, jab, jab, your way to victory. Likely, they would do everything in their power to stop you. AI usually does do everything it can to prevent you from killing them. It just sucks at it.

Now, obviously a dev could simply make the AI unassailably fast and powerful, and people would (rightly) bitch because it was near impossible to get anywhere. But let's not confuse fucked game mechanics and intentionally placing the enemy's timing inside the player's OODA loop for "competent" AI.

I want AI that forces me to become better by playing better against me. Am I constantly moving in a specific way (that I am not conscious of)? I should get dick-nailed when I do that for the third time. Did I catch the enemy when he was making that move again? He'd better make an adjustment.

This is all about pattern recognition. The AI should be able to utilize it, and it should also have patterns that it uses to engage us. Patterns that can change as needed.

This certainly isn't beyond programmers to do. The Skyrim mod Dragon Combat Overhaul does this marvelously, despite a multitude of limitations forced by the game's engine. It achieves this by allowing the dragon to assess how the battle is going, and adjust accordingly. Getting it's ass kicked on the ground? It won't land again anytime soon. This is a fairly simple concept, with only a few variables, and yet, it is a game-changer to have to fight one of these things. The difficulty goes way up, not because the dragons are tougher or faster (their stats are unchanged), but because they behave smarter. They fake, break their usual patterns, establish a behavior long enough that you adapt to it, and then change it again.

There is absolutely no reason that a melee-based AI could not be an order of magnitude more cunning than these modded dragons, but to be "realistic" it couldn't simply be infallible. As a player, we would need to have the ability to perceive its choices; Not just the result, but the formulation of those ideas. "See the wheels turning" as it were.

For instance, the AI starts an attack, then changes angle to observe what your countering response is. A few seconds later, it executes the same thing, observes that you do have a pattern… and on the third time, it tries to counter your expected reaction. Of course, you could anticipate this, and discern what the likely counteraction will be, prepared to take advantage of that yourself. This is a fairly one-dimensional exchange; most such actions are hidden within a few other attacks with the real goal being to see your reaction to a specific one amidst many.

The better turn-based strategy games already do this, with the civilization/nation/daimyo AI observing buildups on borders, weaknesses of structure, expected limits of mobility, etc. This would require a faster pace, and some way of discerning data from noise, but the game has an inbuilt cheat since you have to press the keys in a certain fashion to get the game to do what you want. It wouldn't be hard to log the player's selected discrete moves as this AI opponent perceived them, and begin building a database of intelligence about how the player will fight.

There will naturally be gradients of enemy AI. Some will be comparatively stupid. The real purpose of henchmen? To give the uber-bad guy time to observe the player and develop that database.

The danger of fame within the game world? People know how you operate and what to plan for.

This forces the player not only to adapt to countermeasures, but to be cognizant of their habits, and to develop multiple modes of bringing their chosen tactics on line… much like a real-world competitive fighter.

- Begin the rewrite
- Drop the culture sections and game settings data, keeping those concepts more abstracted
- Drop information on martial evolution - not needed for this article
- Lots of editing

Rate this article

rating: 0+x

Add a New Comment
Unless otherwise stated, the content of this page is licensed under Creative Commons Attribution-ShareAlike 3.0 License