Difficulty Tuning Roundup
There are many elements of gameplay that are worth tweaking
for harder difficulty modes, provided there is time to test each of these
thoroughly against the other modes, and with players of varying skills. I'll investigate
a few here.
Time limit. This is a classic difficulty tuner for racing
games, and a few platform games-essentially inspiring the player to get to the
end of a level or challenge by X time. This usually forces the player into
optimal route-finding play and looking at their navigational tools differently.
Had Prince of Persia:
Sands of Time embraced non-linearity, we may have had a time limit goal
challenge to recontextualize the levels in some very interesting ways.
Generally, most racers have a medal-based reward structure where the gold medal
will be a very tight time to achieve. Games that have other play types involved
in the racing -- F-Zero has combat for example -- end up forcing the player to also
optimize those play types, alongside finding the ideal path.
Damage Dynamics. This is the easiest thing to tweak when
making a game harder-just make everything hurt more. The problem is that
without careful tuning of other factors, such as enemy placement and resources,
this can bring about one-hit kill syndrome, which in a 5-10 minute stretch of
play can come across lazy and unfair. It's also the easiest way to change the
feel of the game from an evaluate/solve game to a predominantly memory-play
based affair.
The best difficult modes in games tend to share the same
formula for upping the difficulty, which is contextual to the level portion.
Sometimes this involves more enemy spawns, and most of the time there's
additional damage-and in some places, less ammo.
Simon says. If the game is dependent on button presses, like
DDR, Beatmania, or Guitar Hero, then demanding more presses per time unit makes
sense as a difficulty increase. Pressure is increased through the players
having to parse what they see, and combine that with what their currently
engaged feet, fingers, or other digits are already doing.
Increased AI aggression. Halo 1 and 3 are perfect examples
of how to pull this off, as are the Infinity Ward installments of the Call of
Duty series. Sometimes, the enemy just behaves more aggressively, forcing the
player to act more quickly and confidently, rather than wait around trying to
aim a perfect head shot. In Halo, the grunts will do suicide kills in the
harder difficulties.

The enemy AI in Halo shows increased aggression at higher difficulty settings.
Make your enemy AI take more risks, but on the flipside
don't make the enemy AI be able to see the player from a million miles away
with the aim of a laser-eye-surgeried hawk, or you'll get a Mega Man-style
scenario in which players feel cheated through an unknown and unseen kill, such
as from an off-screen bullet. In some contexts it's realistic, but in most
games of this sort that I've played, it's rarely fun. It could just be me, but
it made me stop playing Medal of Honor: Airborne.
Reduced neutral zones. This is when safe areas and unused
areas are removed or replaced with hazards, such as spikes you can't stand on
or instant-kill zones. The Challenge levels in Valve's Portal are a great
example. This has the added side-affect of pin-holing the player's focus on
very few options, and with good level design can create a high-pressure
puzzling environment.
HUD restrictions. In some games, challenge can be tuned by
altering the interface. Games with multiple feedback systems work best, as
numbing or reducing just one of your in-game senses can cause player
decision-making to be more tense. The Metal Gear Solid series has done this
since the beginning, where harder difficulties would remove the in-game radar,
meaning the player had to rely not only on their memory of the game's layout
but also their reflexes and knowledge of the AI behaviors to survive.
Resource stinginess. The Resident Evil series chose to limit
ammo as difficulty levels ramped up. This results in extreme tension, helping
to add to the horror, but also greater pressure on the player to be
resourceful. Many skillful Professional Mode Resident Evil 4 players rely on
sharp-shooting kneecaps to bring the enemy to its knees, then finish it off
with the knife to get through the game on a minimal bullet-budget.
The Illusion of Fairness
It's a common perception that the key to keeping a challenge
fun for many is to make it feel fair to the player. Of course, classic video
game scenarios-such as one man against an army of aliens, or a flying saucer
against numerous battalions of space ships-is completely unfair.
Few games, if any, are "fair." The illusion of
fairness to the player is what's important, and this comes in the form of
convenience mechanics. This includes things like health packs placed before and
after sections with large enemy counts, generously placed ammo pick-ups,
strength/weakness matrices for certain weapons a player may have versus present
enemies onscreen, and smart checkpointing that checks in regularly, usually
before and after these kinds of areas. Above all this, communication and
feedback to the player is necessary. Here are some examples.
Physical feedback. This is the staple of many FPSes, but
Resistance: Fall of Man lacked it (the game has since been patched for Sony's
new rumble controller), and in my opinion really suffered for it. Aside from
obvious elements like weapons firing and vehicles moving, most games nowadays
tend to use feedback in the controller to communicate how close you are to
death.
For example, if the controller is at a light stage of rumble, the player
is being attacked, but is not about to die, or the enemy is attacking from afar
with weak munitions. If the rumble is intense, chances are that the enemy is
attacking at close range, or with an attack that is bringing the player nearer
to death.
Hazard feedback. Flashing red markers onscreen and other
screen filter effects are great for games where your avatar is in 3D space and
can be attacked from all angles. Half-Life pioneered this with 4-way red
flashes to indicate if damage was greeting you from the sides, behind or in
front, but later games like Call of Duty enhanced it by specifying exactly
which angle.
Loss feedback. This is somewhat rare in games. Usually, this
feature is considered taken care of if the game is tightly designed. Some
games, such as the critically acclaimed racing game Burnout, use an instant
replay function. Team Fortress 2 also does a great job of telling players how
they died. Upon being killed, a player's assailant is displayed onscreen
immediately crushing any illusions that the player was robbed of a victory by a
poorly built game-an illusion suffered by many sore losers (myself included).
|
One additional suggestion is to turn the problem of difficulty adjustment on its head: rather than designing the game to be relatively easy and increasing the challenge by adding problems or reducing resources, set the default balance to be "really hard" and provide dynamic tools that help the player overcome the challenges.
Some basic examples of such tools would be spawning more ammo and/or health powerups, reducing the number of enemy units, and degrading enemy AI. But a more interesting approach could be for the game to be able to detect when the player is having trouble with one section of the game -- perhaps they frequently die and reload in the same location. As this condition is detected, the game could begin to spawn friendly NPCs with ever-increasing abilities to support the player. "I'll cover you -- you take out that generator!"
The approach has the advantage of allowing the game to be tuned with a consistently difficult level of challenge throughout, while dynamically providing just enough help when and where it's needed to get players through sections they're having particular trouble with.
Another way to make "Simon Says" harder is to redefine the victory condition. The best example is Beatmania's "normal clear" and "hard clear". The evaluation of your success is very different depending on which lifebar you pick. With the normal lifebar you're immortal and victory is defined as a solid performance. But with the hard lifebar not only can you die, but you die easily. Mistakes are still allowed, but you'll lose as soon as the game catches you 'faking it'.
The particular example that comes to mind as an exemplification of this is the arcade shooting game Gunroar, by Kenta Cho (http://www.asahi-net.or.jp/~cs8k-cyu/windows/gr_e.html freeware for download. There's also a Linux binary out there). The game always progresses forward AT LEAST at a certain rate and can never get below that. This is the bottom. However, at the player's option he or she might choose to proceed faster and thereby increase the game multiplier faster than it would otherwise do on its own autoscroll. This has two effects: The game gets harder and the score for downing an enemy ship is greater. This allows someone with no experience with the game to gradually come to understand what expectations to have, and allows an experienced player to jump right in and ramp up their personal difficulty in no time at all.
Another interesting facet of Gunroar (and many independent Japanese-made games, if I'm to be frank) is the ability of one to watch replay of one's own gameplay. Some games allow you to save them, but Cho's games tend to use the replay as the game's demo. That is, when you lose in Gunroar, the title screen will have your latest game in the background for you to watch at your leisure. While this doesn't provide the improved functionality of a saved file (you can't speed up, skip, or rewind), it does give immediate feedback on how you played. These games in particular tend to give one tunnel vision because the player's single-shot-death boat is much more important than destroying every enemy. The instant replay allows a wider view of the field while the experience is still fresh in their mind. This replay mechanic serves, then, two roles. One, to help improve the score of the player by giving them the opportunity to better understand where and how to take risks for score; and two, to offer some loss feedback-- know where the bullet originated from and how it moved and why you were there in the first place allowing you to possibly learn from the experience of loss.
I'm of a very strong mind that, with the right balance, dynamic difficulty could eliminate this concept of granular "levels" in the classic sense and we're starting to approach the point where it's computationally feasible to do so. I'm not at all certain that I WANT the concept to die, but that's a discussion for a different day.
One key I've always found is think about difficulty not as making the game more difficult but instead think about it as making the game more challenging.
This could implore ideas of tactics and strategy of AI and combat encounters. More flanking, more aggressive behaviors, AI can hear you better, they track you for a longer amount of time etc...
Yes I agree increasing the hitpoints in npc's is bad, but often used technique. Increasing accuracy can also be quirky. Inc the deadlies of each npc might be a better tuning strategy.
One technique for an action game, for example, might be to make enemies slower or leave longer openings between attacks or have less armour as you, the player, fail more. In sports games, decrease the stamina or some other attribute of the winner at halftime. Make the enemy guns less or more accurate in an FPS, make the enemies less aggressive in an adventuring title, increase the amount of wear on the opposition's tyres as you race about the circuit. There are multitudes of things that can be done to "help" the player surreptitiously without resorting to giving them more than they could have earned normally. Giving handouts in a game is rather irresponsible, in my opinion.
Some of these, at a glance, may seem rather drastic. Indeed, some of them, when hustled hard, could be game-breakingly terrible. But two factors should be considered here. First off, at the point where you're gaming the system like that, you're no longer playing the game you paid for-- you're playing a game of your own devising wherein you purposely attempt to break the core mechanics as thoroughly as possible. You're playing it like an MMO. Second, one iteration of change should have effects that are only noticeable if you know to look for them and then only barely. Subtlety is key in my theory, here (use your imagination. S'what you get paid for, no? :).
But Brad's ideas are still interesting, and I could see them as now being possible in certain genres (hell, you can do anything if you throw enough code at it. It might not be good, but you can do it). Dynamic difficulty, as Anonymous above me notes, is definitely a touchy subject, and one I've given a lot of thought to. I'll still say that I definitely think it's doable, and even a good idea to try. Using subtle gradations rather than large leaps should have a net effect of keeping the pressure on the player and actually nurturing their growth (we can count on powergaming munchkins existing, but if that's how they want to play...well, the customer is always right :/ ).
It's just another piece of the balancing formula, just as difficulty has always been.