You may have heard the news that The Experimental Gameplay Project is up and running again. For those who have never heard of it, it is a group that was founded by Kyle Gabler (World of Goo) and Kyle Gray (Henry Hatsworth) to rapidly prototype new game types. They take an idea and have seven days to create a small game around it to see if it works. Some of the stuff they have come up with is brilliant, I heartily recommend you browse through some of their other games.
One of the latest ones is Egg Worm Generator by Karl Gabler, originally designed as a shooter, after running short on time and with the AI not being good enough for the task the game was re-purposed into an evolution simulator. The goal is simple, all ‘creatures’ start out in the centre of a flat plane, on the right is the path to safety, on the left is certain death. The first four creatures to make it to safety ‘win’ and a new generation is born based on their characteristics. Things start out slowly, most barely making it over 2m\s but as things develop the newer generations get faster and faster.
Sadly, my first run reached it’s peak around generations 15-25, by generation 100 those original records still hadn’t been broken. So it doesn’t work brilliantly (although I have heard some people have run it for a lot longer, ~14,000 generations, and hard better results), but that wasn’t the point after all. The point was to come up with something new and whilst it isn’t exactly that either, it is certainly something that hasn’t been done in a while or quite the same. With more development it could perhaps evolve ( 
) into a playable game.
Another example of this type of simulation is actually mentioned by the creator, a simulation carried out way back in 1994 that developed ‘creatures’ for specific tasks such as swimming and running. It’s well worth a watch. (Source)
The other example that I know of is Darwin Pond. I was first introduced to this by a science teacher at secondary school, it simulates interactions of the specimens as they eat and breed. The slower swimmers often don’t make it to the food which means their poor design is discarded whilst the more efficient swimmers live to breed and pass on their advantages. Darwin Pond isn’t just a ‘let it run’ program like Egg Worm Generator, you can actually perform experiments, how would the swimmers do with more food or less food, you can also save swimmers for later so that you can test them against each other. It is available for free from the Darwin Pond website.
Going slightly off track, the closest (and more importantly gaming) example is Spore. It doesn’t have the generational aspect, but it does include development and from what I’ve read when your creature develops walking abilities the game can work out how to use whatever configuration of legs you may have given them. I haven’t actually played Spore so I can’t comment any further.
This post is pretty ambiguous with it’s relation to games, but then if a Flight Simulator can be called a game, why not these? The answer of course is interactivity (although that’s harder to explain away with Darwin Pond), that doesn’t stop the potential for incorporating these simulators into a fully fleged game however.
Spore features no evolution at all from what I have seen my brother play, you just create the creature, you are not creating a population of slightly random creatures and alowing the best to breed to make the next generation, that idea was discarded early on in the development when EA decided they wanted to make a mainstream game.
If you want a go at another one of these things, and one that gets the genetic algorithm better than eggworm) then you should have a look at breveCreatures it is a screensaver and remembers where it left off, so you can run it for weeks if you want.
That looks interesting, now if only my computer would make it to screensaver…
If an evolution algorithm were ever to become part of a game, I think it would have to be a very minor part. I can’t think of any real uses as it is besides allowing you to cull certain types, which would lose appeal unless there were more goals.
Well one use for a genetic algorithm in a game would be the enemy AI. As in the ones you defeat quickly don’t go on to breed so eventually they can always get you. Could be interesting for an RTS, and would prevent people using the same tactics over and over again to win. Obviously it would need the ability to be reset.
PCG pointed this out a few days back (last week if I remember correctly) and I was slightly intriqued. Certainly could have potential for something.
Anyway with regards to Spore Jon is right. After all the evolution is technically done by you thus not really rendering it much of a evolution simulator.