Evolutionary Game Design

I don’t much care for designing sequels. Almost every game I have designed is a completely new idea. I did write a sequel to Balance of Power, at the behest of Mindscape. I also released a second edition of Eastern Front (1941). Otherwise, my stuff has always been new.

This is an uncommon approach to game design. Most designers stick with something that works for them. We have the Ultima series of games from Richard Garriott, the graphic adventures from Sierra, the flight simulations from Larry Holland, and so forth. The phenomenon is particularly clear with the series of games culminating in Doom II. Since many designers have played this series, it provides us with a useful example of the problems and possibilities of evolutionary design.

The basic series of games began with Wolfenstein 3D, then continued through Spear of Destiny, Doom, and Doom II. I have played all of these save Spear of Destiny.

The central selling feature, of course, was the fast 3D engine. You could move around in three dimensions with a frame rate so high that the motion appeared to be fluid. That 3D engine is what got everybody talking and playing.

The game itself, as I have described in a design review of Doom, was quite simple but nevertheless well-executed. This is one of the design elements of the series that few people have paid much attention to: it’s not only a good 3D engine, it’s a competently executed game.

What’s interesting to me is the evolution of the design. With each new release, the designers chose to advance some elements of the design. We can learn a lot by examining the elements that the designers chose to advance. We can learn even more by considering those elements that they chose NOT to advance.

There are two basic means for improving a game design. One approach is to refine the code, tweaking and improving the critical algorithms so that they run faster and better. The other approach is to take advantage of newer hardware, faster processors, bigger RAM sizes, and so forth. Most designers will use some combination of both approaches. It’s interesting to see how the Doom system expanded. The original Wolfenstein 3D used only 98K for its main executable (although there may have been some other executable code buried in other segments), and the total hard disk consumption was only 2.4 MB. With Doom, the total executable code shot up to about 700K, and the hard disk consumption more than doubled to 5.2 MB. With Doom II, the main executable gained only about 100K of code, but the overall hard disk consumption tripled to 16 MB! This demonstrates a common trend in many evolutionary designs: the object code grows slowly, but the associated data explodes in size.

What they improved
3D technology: Since the basis of competitive advantage was the 3D engine, we’d expect that this was the element that received the most attention in moving through the series, and it is obvious from the results that the 3D engine received a great deal of refinement. In moving from Wolfenstein 3D to Doom, they added a vertical rolling motion as the player walks, which greatly enhances the perception of motion. They also added curving walls, a big improvement from the rectangular geometry of Wolfenstein 3D.

Vertical movement: This was perhaps the biggest change from Wolfenstein 3D to Doom. In the former, you moved in a flat two-dimensional space, but in the latter, you could move up and down, climbing stairs, jumping off ledges, and so forth. This made possible several important design innovations: elevators and jumps.

Jumps: This was one of the big design improvements in the series. You could jump downwards. More important, your forward momentum would carry you some distance. This made possible a brilliant maneuver where you jump off of a rising pillar and your momentum carries you over a ledge and into an otherwise inaccessible area. The designers used jumps only sparsely in Doom, but in Doom II they started to understand the possibilities and they sprinkled such jumps liberally through the design.

Puzzle Cues: This is a good example of the designers correcting a mistake. In Wolfenstein 3D, there were no visual cues to secret doors. The player had to push on every single wall section to locate hidden doors. This became very boring very quickly. It was a design blunder, and there were quite a few complaints. The designers had the good sense to correct the mistake with Doom, putting visual cues to help the player recognize secret doors. However, I believe that some of those cues were just too subtle. I remember staring at a wall that I knew contained a secret door, trying to discern the visual cue, and never figuring it out.

Textures and more diverse imagery: The walls in Wolfenstein 3D were quite plain. There were about four basic styles of walls, each with a few variations. Obviously, the programmers didn’t want to eat up too much memory on wall images. But with Doom they pushed it further, coming up with a greater range of wall types. These of course consumed much more RAM. With Doom II, they went even further, adding a great variety of wall images. In the process, they broke the 4 MB limit; the game runs in 4 MB, but not well.

Ammo utilization: This was one of the big disappointments in Doom. You had so much ammo that you could pretty well blast anything that moved. The chainsaw was useless. With Doom II, they made ammo conservation an important strategic consideration. You really have to carefully weigh every single combat; in many cases it’s better to wade in with the chainsaw and take some injury (which can readily be repaired with medical supplies) rather than waste precious ammo.

Weapons differentiation: This was another element that showed continuing refinement as the series advanced. In Wolfenstein 3D, weapons are used recklessly; there isn’t much point in thinking through their optimal utilization. But in Doom II, each weapon has its own particular application. The chainsaw is best for knocking off individual weak monsters; the shotgun is good for short-range combat when you can squeeze off a shot and duck behind cover; the chain-gun works best for long-range work or more intense firefights; the rocket launcher should be used only for long-range pummeling of tough guys; the plasma gun should be reserved for intense confrontations with particularly nasty bad guys; and the BFG should only be used to clear a room of lots of bad guys.

Puzzles: The designers showed increasing cleverness in assembling the basic components of the system into interesting puzzles. I was quite impressed with one situation, in which you round a corner to find a room filled with imps who then teleport to a location just behind you, catching you in a terrible crossfire. The solution, it turns out, is to activate the imps and use the chainsaw. They can’t teleport into the position if you’re standing on it, so you stand on it, step back, and dispatch them one at a time. That’s a particularly clever combination of elements. Another bit of cleverness was right next door, in a room packed with a whole army of very dangerous monsters, presided over by the nastiest, meanest monster of all. I doubt that you could carry enough ammunition to finish off this horde. The solution, it turns out, is to dash into the room, grab an "invulnerability sphere" before they can vaporize you, then get the hell out of there, leaving them to turn on each other. You’ll still have to deal with a few who teleport out, and those few are a quite a handful, but when the dust settles, most of the killing will have been done for you. That was a clever puzzle; my compliments go to the designer of that level.

Monsters: Yes, they added more monsters as the series progressed. Now, I tend to take a dim view of this design approach; the fact is, adding monsters is a really trite way to spice up a game. In the case of Doom and Doom II, though, the designers demonstrated an admirable restraint. Their monsters are not petty variations on each other. Each one needs to be tackled differently. I swear, if they’d had purple imps, brown imps, and green imps, I would have shut the game down and never played again. They came close to making this mistake with the three kinds of soldiers in Doom II. Particularly impressive are the "lost soul" monsters with their heart-stopping lunge at the player.

Modem and network play: This was another big improvement from Wolfenstein 3D. It took a lot of coding work, I’m sure; perhaps that’s one of the reasons for the big expansion in executable code size in Doom. In any event, the modem and network play capabilities catapulted the game into a higher level of fun. Stories of intense personal combat are now part of the lore of Doom.

What they didn’t improve
Screen Layout: The basic screen layout has not changed throughout the series. There’s the first-person view dominating the screen and a status bar along the bottom. I see no problem with this; if it works, don’t fix it.

Input structure: There wasn’t much to change here. Movement commands, fire commands, and commands to change weapons that’s about all there is. The designers made no changes in the input structures during the evolution of the series. I think this was a sound decision. There’s this temptation that few designers can resist: we’ve got 101 keys on this keyboard, and right now we’re only using a few of them; why don’t we add a few special keys that do really neat things. Flight simulators are the worst offenders here. The Doom designers had the good sense to leave well enough alone.

Object geometry: One of the most striking shortcomings of the 3D engine is its treatment of stationary objects (dead monsters, weapons, medical supplies, and ammo). No matter how you move around, they always face you with the same orientation. Now, I am not criticizing this design decision; if you think about it, this is a clean solution to a messy problem, and you really don’t need to see the back side of a medikit. But it is noteworthy and I think a correct decision -- that the designers made no attempt to soup up this part of the 3D engine.

Monster AI: Another area that received little attention in the evolution of the series was the artificial intelligence that the monsters use. Basically, they wait at their assigned locations until the player comes into their field of view; then they attack. Their attack patterns are not too smart. I noticed a slight increase in the aggressiveness of the monsters in Doom compared with the enemies in Wolfenstein 3D, but it really wasn’t much of a change.

This disappointed me; the designers could have done so much to make the monsters more interesting. Particularly distressing is the fact that monsters will not respond to obvious tactical opportunities. For example, suppose that three monsters are standing together in a room, waiting for me to come around the corner. I carefully edge up to the corner, just enough to get the first one in my sights. I blast him; he cries out and falls dead. Do the others do anything? No, they just stand there. So I edge a bit further around the corner and blast the second one. Still the third one remains in position. Then I pop around the corner and nail the third guy. It would not take a great deal of AI to give the monsters the ability to activate when something happens in their room. Apparently, the Doom designers did not see much need to do so.

Instead of creating smarter monsters, they just made stronger monsters. This is nothing more than the old dungeon crawl stunt of making ever-more intimidating monsters and then giving the player ever more powerful weapons with which to dispatch them. I will admit that the monsters are somewhat interesting, especially the one that revives dead monsters but they really should have gotten a little smarter.

We must admit to ourselves that opponent AI is almost always the weakest part of our designs. Coming up with good opponents who do interesting things is an important skill, yet I see very few designs that show any talent in this area.

Hardware issues
The biggest changes in the series appear to be attempts to take advantage of improved hardware. This immediately put the designers in a no-win situation. By basing their design advantage on their technical prowess, they insured that they had to keep up with the latest hardware. This in turn forced them to require increasingly powerful machines. Wolfenstein 3D runs very smoothly on my 386/33. Doom runs quite well, with only the rare stutter when lots of monsters appear. But Doom II brings my machine to its knees. Even with all the options set to the most conservative settings, I experience frame rates of about 1 fps. That’s terrible!

Indeed, this problem goes beyond aesthetics. The game is unplayable on my 386/33, even though the box declares that it will work on a 386 with 4 MB of RAM. (the box recommends a 486, but doesn’t rule out a 386). The problem comes with some of their timed puzzles, in which you must activate a door or elevator and then run or jump to it before it reverts to its original state. In several cases, I found that it was simply impossible to get the job done in the allotted time. This was not a matter of FumbleFingers Crawford missing the boat by a fraction of a second; in the case that I studied, I missed the timing window by about 1.5 seconds.

I have a hunch as to what went wrong with the design. I’d guess that the program keeps an internal RAM heap of imagery, and constantly updates that heap as the player moves. This is why turning corners to large scenes causes the program to stutter. In the timing case I experienced, the program had a large amount of scenery behind me as I faced the critical switch. When I hit the switch, I turn and start running but the program has to reload all that scenery into the heap, which takes up perhaps two seconds. Those two seconds are just enough to kill my chances of making the timing window.

The right way to solve this problem would be to run all system timing through the master clock, but that would itself involve some overhead. Since some of the game’s activities are unclocked, the designers have to hope that everybody’s system is fast enough to get them through; this was not the case with my 386/33.

There really wasn’t much that the designers could do to solve the problem. When you base your product’s strengths on its speed, you have no choice but to assume a fast machine. But then you have to back this decision up with a firm box statement: "REQUIRES 486 PROCESSOR!" What’ll you bet that the designers asked for just that, and somewhere along the way a marketing person talked them into compromising the specification down to a 386?

Problems of difficulty
There is no question that Doom II is much harder than the original Doom, which was only marginally more difficult than Wolfenstein 3D. Why the sudden jump in difficulty? I suspect that the designers fell prey to a common mistake. It works like this: you’re paying close attention to the feedback that you get from customers. You get lots of complaints that the game is too easy to win, and few complaints that it’s too difficult. So you make it harder to win. Seems reasonable, right?

But think about the feedback process. What kind of people are going to give you feedback? What kind of people will refrain from giving you feedback? The answer, of course, is that the games aficionados, the guys who live and breathe the game, will give you feedback. And the poor slob who gets clobbered at the outset is not going to publicize his incompetence by writing the company and complaining that the game is too hard to play. He’ll just quietly remove the thing from his hard disk and never buy another such game.

We’ve seen this process over and over with games. A genre starts off appealing to a wide range of people, but after a while it specializes and moves away from the mainstream, becoming the preserve of the experts. Doom was a mass-market game that appealled to a lot of people; on that basis, many will buy Doom II. But I think that Doom II will be the last mass-market product in the series. After that, we’ll see an ardent group of Doomsters happily playing their Doom III, Doom IV, Ultimate Doom, Final Doom, One More Doom, and so forth. It’ll continue to be a commercial success -- but I suspect that the steep growth curve will level off quickly.