Tuesday, May 30, 2006

Defending a Point: Examples of Positive Feedback Loops

I'm consistently amazed at the way people plan - or fail to plan - their games. These are people who spend hundreds of thousands to make a particular game, but they don't even take the basic, well-known fundamentals of game design into account.

Let's take the charming Bang! Howdy. We'll leave off on the design of their game world and community, and focus strictly on their tactical game.

They have two variants of a specific tactical game for 2-4 players. One is called Claim Jump. The other, Gold Rush. Both involve collecting nuggets, but Claim Jumping lets you steal nuggets from another player, whereas Gold Rush does not.

The other difference is that one has a hideous positive feedback loop, and the other doesn't. You know all the information you need to know to see the feedback loop. Can you see which - and why?


The answer is that Claim Jumping has the positive feedback loop.

The reason is simple, once you think about it. Stealing from the enemy means going to their base, killing their units, and taking their money. Since the game is largely opportunistic, once one player starts attacking your base, the other players follow suit because you're now an easy target.

This means that every time anyone steals any gold from you, there is a chance (30-60%) that you will go down to zero regardless of how much gold you have, since the players will continue chaining theft after theft. It's impossible to defend yourself or stop them, because your units respawn, one by one, right in their feild of fire.

Worse, these players cooperate to steal your gold - the way the game is set up, unless you have a sniper, it's vastly more efficient to steal and run rather than steal and try to kill the other person who is also stealing.

The rest of the time, the other players attack the now-undefended base of the player who is attacking you, rather than teaming up with him. This results in the kill potential simply chaining to the next player, and it becomes likely that his gold will all get stolen and his score will drop to zero.

On the other hand, the Gold Rush scenario doesn't allow theft. Therefore, you race, and the vast majority of the combat happens between bases rather than at bases. Even if someone does camp on your base, they are losing more money than you because they aren't collecting cash and they have the units available to collect it. You, presumably, do not, or they wouldn't be at your base.

This means there is no positive feedback loop, because there's no advantage to suppressing someone. The only advantage is to kill whatever units they have deployed, meaning that your respawning units are not immediately crushed and, more importantly, your score doesn't actually drop while you are down.

Of course, the dynamics of Gold Rush also leave something to be desired, because whoever has the fastest units almost universally wins. But it is their best game.

Their other game is the cattle-branding game.

The cattle game also has a positive feedback loop embedded in it, because the cattle run from any soldier that pulls up next to them. This means that if you get an edge, you drive the cattle towards your base: further from your enemy and closer to the seat of your power. This is functionally the reverse of Claim Jumping: once you start to get behind, you have a harder and harder time getting points. But the result is the same: the weak get weaker.

I suppose it is theoretically possible that several weak players could team up to raid the cattle of a stronger player who has herded them into his base. However, every time it happens, the two invaders war amongst each other, giving the defender enough time to respawn and retake the cattle from the two shattered, weakened armies. This is because the way this game is set up, it is more effective to kill the enemy's hero rather than just running with whatever cattle you can grab.

It really is astonishing. The dynamics of Claim Jumping and Cattle Rousting are ideal for maximizing the positive feedback loop. It's almost like they did know basic game design, and set out to create the least balanced games possible.

Anyhow, I'm picking on Bang! Howdy, but most other games are at least as bad.

It doesn't take a genius to do this kind of basic game design. As people spending hundreds of thousands of dollars, you would think they would carefully balance their games.


Craig Perko said...

An interesting way to fix most of the negative feedback loop would be to make units respawn in an amount of time relative to the distance they died from your base.

So, if your people die defending your base, it takes two seconds for them to respawn. If they die at the other guy's base, it takes a minute.

This means defending against a dramatically superior force possible.

Exactly what the times would be would have to be researched, of course.

Patrick Dugan said...

Yeah, I was going to suggest they implement such a negative feedback mechanism. Positive feedback is what makes dynamics interesting, but only when balanced by negative feedback. Its yin and yang, very basic stuff.

I'm wondering, however, how to detect and adjust feedback loops in models where the verb-set is inconsistent by context.

Craig Perko said...

The answer is to rephrase the question by rephrasing the way you talk about your game.

If you think of your game as "verbs in context", you are not talking about a logical system. If you convert it over into a more transparent paradigm - or a number of more transparent paradigms - you can ask the question again and get a useful answer.