AI Algorithm

Thanks to Forge (at Computer Programming and Magic: The Gathering) for the heads-up on this article on Temporal Difference Learning, a way to get an AI to "teach itself" to play games (in this case, Backgammon). I'm still getting my head around the math, but the possibilities are fascinating.

Game Competition

Dr Dobb's Journal and Microsoft are running a competition, Dr Dobb's Challenge II. I might enter, if I can get my Robot Sumo Wrestlers in a Maze assignment done.

Idea #7: Flight simulator

Anyone who knows me knows I love things-with-wings (a few months ago I spent several hours drooling at planes at the RAF Museum at Hendon). So naturally, I like flight simulators, especially combat-oriented ones. Most modern flight simulators are very accurate representations of air combat.

And therein lies the problem.

I like a realistic flight sim as much as the next prop-head, but after a long day at work/university, there are times when I just want to load up a sim and go blow stuff up. I don't want to be dealing with a highly complex flight engine -- I want something that takes some of the work of flying away, and lets me concentrate on the fighting.

I suppose this is partially a nostalgia kick. The flight sims that I most fondly remember -- F-29 Retaliator, Fighter Bomber, F/A-18 Interceptor, Falcon, Birds of Prey on the Amiga; F-19 Stealth Fighter, Their Finest Hour, Gunship 2000, A-10 Tank Killer, LHX, Harrier Jump Jet, Strike Commander, US Navy Fighters, ATF, USAF, WWII Fighters and even Falcon 3.0 on the PC -- had varying degrees of realism, but all were immensely fun while being realistic enough.

Now days, you have the über-realistic games like Lock-On: Modern Air Combat and Falcon 4.0, or the arcade shooters like Ace Combat 6.0. The middle ground is gone. Don't get me wrong, I like all three of these games, but sometimes I want something that's realistic enough (i.e., the F-16 can't carry 150 Sidewinders) but also fun (i.e., I don't have to spend 10 minutes flipping switches before I can even start to taxi).

I also believe that this failure to publish for this middle ground is slowly strangling the flight sim market. The market is currently serving the simulation fanatics or the arcade jocks, without doing anything to entice new players deeper into simulation territory. It's like you've just finished learning to crawl, only to be told that the next step is to run a marathon while wearing a 100 lb pack. Most newbies come in with the Ace Combats, will take one look at LOMAC, and stick with the arcadey games. Some may make the leap, but of those, how many will bail before reaching the top of the difficulty curve?

So this idea revolves around a flight sim that serves the middle ground: accurate enough, but not sacrificing gameplay to the point where it stops being a game and starts being a second job. I definitely like the idea of dynamic campaigns -- Falcon 3.0 did this particularly well -- and providing a catalog of aircraft for the player to choose from -- though maybe the 40 aircraft in Birds of Prey was a little excessive.

And having the sweet, sweet graphics of Ace Combat 6.0 wouldn't hurt...

Idea #6: Shoot-'em-up

R-Type horizontal-scrolling or Xevious vertical-scrolling shoot-'em-up. Nothing fancy: shooting baddies, collecting powerups and points, scrolling over a variety of backgrounds, and a boss baddie at the end of every level.

Idea #5: Epic fantasy game

I love the two Elder Scrolls games that I've played, Morrowind and Oblivion. The open worlds, the RPG mechanics, the stories... all combine to form games that really suck your time away (I've lost count of the number of times I've been playing one of them, and suddenly finding that it's 3am).

But the one thing that does annoy me is the lack of continuity between the games. Oblivion makes the assumption that the main quest from Morrowind was succesfully completed, but all references to events of the earlier game are very generic.

What I would like to see is an RPG series, where each chapter checks to see if earlier chapters have been played, and adjusts the game world appropriately. For example, if chapter 1 had the PC doing some heroic deed, in chapter 2, you'd find statues to your character that actually look like your character from chapter 1. Instead of vague references to "the Nerevarine" (to take an Elder Scrolls example), you'd find references to "Jim the Bosmer, who was revealed to be the Nerevarine".

Another example could be a strategic fortress, where the player in chapter 1 had a quest to (help) capture it. How well the player completes the quest (right down to abject failure) affects the strategic position of the world in chapters 2+.

Obviously, you would also need a "default" position, for those players who don't start the series at the beginning.

You could use the opportunity to create a true epic arc, for example, the little fishing village you start at in chapter 1 could be the capital city of a huge seafaring empire by chapter 5. Which does raise a salient point: how far apart should the chapters be set, geographically and temporally? I do like the idea of setting the chapters centuries apart, which gives a reasonable explanation for why you have to create a new character at level 1 each time, and gives you the opportunity to show how the world has changed since the last chapter. Geographically, you could take the Elder Scrolls route and set each chapter in a different part of the world, but I also like the idea of showing the change in the world between chapters. Possibly one solution would be for each chapter to be set in an area adjacent to a previous chapter's, expanding the world while showing the changes in the older parts.

Mentioning levels also reminds me of another problem in Elder Scrolls. In order to remain a challenge for the player, creatures level up along with the player: the wolf that gave you problems at level 1 still gives you trouble at level 10. This is whacked. At level 1, the wolf should give you problems. By level 10, the wolf should sense that you need to be left alone. By level 20, the wolf should be actively running away from you.

Urk

Urk.... between university and all the things that needed doing around the house, I haven't had much chance to either update this blog or do any gamedev. Heck, I haven't played a game more challenging than Freecell on my PC or X-box.

In the meantime, there's an interesting couple of articles on Wired about the length of computer games.

Idea #4: Detective game

There haven't been much in the way of police procedural games with lots of replayability, at least as far as I can recall. The Police Quest series has some -- mainly because of the easter eggs scattered through the game -- but other than that, what's there been?

Using a similar open world approach like Grand Theft Auto, I see this game as being set in a small North-Western American city. The game would dynamically create crimes, seeding the clues throughout the game world (but logically, of course), leaving the player to determine when and where they go, who to talk to, etc.

The player would assume the role of a newly minted detective who, because of the parlous state of the local police department, investigates all manner of crimes from car theft to homocide.

While most of the crimes would be dynamically generated, there would also be an overarching plot -- not necessarily crime-related, or maybe some dynamically generated crimes would hook in to the overall plot -- linking the beginning to the end.