I'm looking at the tentative feature list for the next version of Unangband, and this consists for the most part of game features that have already been implemented in other Angband variants. In particular, I want to implement the 256 colour code and extended Latin-1 character sets (with custom glyphs) that feature in several other variants. Now, at a surface level, you'd expect it to be easy to adapt this code across, due to the similar nature of the code bases (they all have a common ancestor, and the source in all Angband variants is open).

Instead, you end up porting the code across by hand, because the diffs will never merge successfully. You also port a whole class of very subtle bugs that occur because the set of assumptions in the original code base does not match the set of assumptions you have made in your game. For instance, Nick is currently feeling the pain in having adapted a in-game notes patch that features in the many variants to First Age Angband - he still hasn't squashed all the bugs with what consists of a benign user interface change.

This is why my policy for Unangband has been for the most part to 'go my own way' and write code from scratch, using the ideas that may have been generated elsewhere, but with my own hand-rolled code. I end up having to do more work up front, but this reduces the number of subtle bugs I have to deal with in the code base (but sadly not to zero). This hasn't always been the case - for instance, the dungeon generation and AI code is heavily indebted to other variants, but for the most part it seems to work.

But the problem is more serious than just the short term pain of having to adapt and debug someone else's code, or come up with your own version of someone else's ideas. As soon as you start down the path of writing an amateur game in an established genre, you are putting yourself in competition with every other amateur game writer in that genre. In fact, you're putting yourself in competition with all of them, cumulatively. You're (usually) only one developer, and, in the Angband case, at anyone time there are at least 5 or 6 other variants actively being developed, plus the core Angband development team. You have no chance of being able to produce more features in your game than all these other developers, and yet it feels at the same time that you have to, in order to 'keep-up-with-the-Joneses' and 'capture' part of the overall genre player base.

But it's not just the small pool of Angband and variant developers I have to deal with. There are freakish coding prodigies lurking at the edge of the boundaries of this relatively small inlet, working on rogue-likes that are spoken of with whispered breaths in the dim dark corners of the Angband forums. I will admit to being driven close to tears by one helpful newbie who suggested that 'Unangband was kind of a good start, but I should look at Incursion: Halls of the Goblin King, figure out what they're doing and copy it'. I'm an amateur game developer, working on a game in what little spare time I have. I don't have time to play any other Angband variants, let alone anything else... and forget adopting other people's ideas in addition to my own.

Then there is Dwarf Fortress...

(Let's just say I stared deeply into my own soul after seeing its layers of features upon features and the abyss didn't answer... I'm sure peers to James Joyce felt the same way after reading Ulysses)

Attempting to adopt other game's features is the first dangerous step on the path to a mad game development arms race, where you try and include every feature from every other possible game that might apply to yours. You may, through insane acts of desperate endurance, somehow succeed at this - look at the huge 'feature count' for Unangband for an example of this. Having done so, you'll realise that during the time you took to do it, so much more has improved in every other game you were comparing yourself with, that the work outstanding will have grown exponentially. And even if you're strong enough resist the urge to try to feature add and not to compare your 'feature list' to other games in the genre, your player base will.

It's not just Angband variants, or other amateur game developers, who are vulnerable to this problem. Anyone considering creating a first person shooter should read Dsylexci's Tactical Gaming Done Right before proceeding. And then figure out if they're willing to write Operation Armed Flashassault III. I can't go back to Grand Theft Autos earlier than San Andreas, simply because having no climb or swim mechanic sours the taste in my mouth - I've been given better free-roaming and to take a step backwards is like throwing away the keys to my cell door. I'm sure people who played Crackdown feel the same way.

I suspect that this arms race which can only result in mutually assured destruction will be survived in one of three ways:

1. Collaborative worlds big enough to contain every possible game feature anyway (Because no single player game could justify the development budget).
2. Self-contained mini-games with a simple enough rule set that adding features will detract from the game play.
3. Games with a strong narrative drive that only require a feature set large enough to move the narrative forward.

But the inexorable pull of large collaborative worlds, with the gravity of massive budgets and huge player base will draw types 2 and 3 closer to them. The only thing holding back total collapse will be the difficulty of integrating 'pick-up-and-play' into the overall world space - and when Blizzard realises that they already have a segmented player base, and they could just sell levels 1-30 Dranei start as a stand alone game in itself, you'll find the likes of World of All Possible Games Craft taking over.

[Andrew Doull is an IT manager from New Zealand who spent the last 5 and a half years working in the United Kingdom. He's just emigrated to Sydney, Australia, and spends his free time developing Unangband, a rogue-like game, and blogging at Ascii Dreams. He recently covered the Edinburgh Interactive Festival for Gamasutra magazine and has just started an irregular column for GameSetWatch.]