What’s Up With the Mod Scene and Independent Bots?

Alex J. Champandard on January 8, 2008

Why has the mod scene become less active over the years? Have next-gen games grown so big that the sheer size of their codebase prevents independent game developers and AI enthusiasts from experimenting with new ideas?

This week’s developer discussion is based on a comment by Chris Dirks on Getting Started with AI and Programming in Games:

“I’m interested in AI design, but not really looking to build my own game. Which existing games or frameworks would you recommend to start building AI mods for? […] FPS is a slight preference but I’m open, and I’m not picky about languages either.

There was an active dev community for counterstrike bots back before CS Source, but it seems to have died away and most of the sites have vanished.”

Sadly, it seems like a common observation of the industry.

Where Have All the Modders Gone?

After Doom III came out, John Carmack was initially disappointed that the mod scene was “slacking” compared to previous Id games. More recently though [1], he’s realized there’s a trend here:

“In general, all the technology progress has been essentially reducing the ability of a mod team to do something significant and competitive. We’ve certainly seen this over the last 10 years, where, in the early days of somebody messing with DOOM or QUAKE, you could take essentially a pure concept idea, put it in, and see how the game play evolved there.

But doing a mod now, if you’re making new models, new animation, you essentially need to be a game studio doing something for free to do something that’s going to be the significant equivalent. And almost nobody even considers doing a total conversion anymore. Anything like this that allows more media effort to be spent, probably does not help the mods.”

It’s therefore no surprise that modders are increasingly getting absorbed into the AAA treadmill:

  • The teams behind the original Enemy Territory, Team Fortress and Counter Strike have all been hired by Id and Valve respectively.

  • Any coder capable of understand a game’s codebase enough to hack a mod (with little or no documentation or tutoring) deserves a job as a programmer in industry!

If the head of the mod scene is increasingly going professional, and its long tail is disappearing, then what does that mean for AI?

Independent AI Bots

Quake 2 Bot

Around the time of Quake 1/2 and the original Half-Life, there was a wealth of bots available: Reaper Bot, Gladiator, Erasor, HPB Bot — not forgetting sites like Bot Epidemic. Now even those bots are hard to find, let alone new ones!

  1. Are the default AI bots with commercial games good enough already?

  2. Has the advent of broadband internet and network play canceled out the need for AI bots?

  3. Is it too hard to build a bot from scratch in such detailed levels, or using complex engines?

Is there any other reason that independent bots are very uncommon these days?

What Do You Suggest?

Do you have any advice for Chris, or anyone else looking for a starting point making AI mods? What can we do in general to improve the state of independent bot development? Do you know any exceptions to these trends in the modding scene?

Post a comment below or in this thread the forums!

Discussion 9 Comments

alexjc on January 8th, 2008

Maybe I spoke too soon? I just found this: [URL=""]2007 Mod of the Year Award[/URL]. I'm impressed that the older engines like Q2 or HL1 are still being used... In fact, those seem like the most ambitious mods code-wise. Most of the recent mods, however, seem to be rathermore content-based total conversions. Not much sign of AI though! Alex

Jare on January 8th, 2008

The WoW mod scene is insanely active, or at least it was back when I was playing the game. Even in a relatively niche game like Supreme Commander, I see [url=]a lot of stuff[/url] happening: units, maps, game modes, interfaces, gameplay tweaks, and even AIs. It may be true that some mod areas have diminished, probably for all the reasons listed, but new areas have grown (interface mods, user movies). Nothing stays the same forever, and that's a good thing. :)

Andrew on January 8th, 2008

Well, there [i]is[/i] a lot of modding going on (a lot not at the ModDB too), and if a mod is a total conversion, be sure there is AI in there. AI mods by themselves sometimes get attention, but the Moddb isn't very good at accommodating them since they're a tad small for it. I think taking each of your proposed points one by one: 1. [i]Are the default AI bots with commercial games good enough already?[/i] To a degree, and this is highly dependent on the game. I'd say Unreal Tournament 3's bots are actually [i]too accurate[/i], given their levels of "skill". However, no, I don't think many are "as good as possible", or even near "good enough", because many games don't even include multiplayer AI. ;) 2. [i]Has the advent of broadband internet and network play canceled out the need for AI bots?[/i] No, not as such. I am sure that a large percentage of players will play many bot-based games offline, at least for practice. I am sorely missing bots in Call of Duty 4 for instance - took me a good few days to learn the game, when it might have been a few hours with bots and a tutorial or to. And going past this; playing with friends, teaming up against "the AI" is a valid form of entertainment - even if they are not as proficient as the players (especially if the AI level is turned down) it allows larger scale games with less humans. Real time strategy games have this problem especially (keeping 7 other players online for the duration of a 1 hour game? you must be joking!). 3. [i]Is it too hard to build a bot from scratch in such detailed levels, or using complex engines?[/i] I think this isn't a problem as such, since most bots are never built from scratch (or if they are; there is a AI framework for easy modifications to behaviour). They are built on the existing AI code, modified (perhaps until the original is unrecognisable). I think an important point to add to this one is that; [i]A lot of game AI's have become completely hardcoded.[/i] The amount of SDK's which lack any way to change the AI (notably; these are more "Map and model editors" then anything) is getting larger. Even more so; there is no way to handle multiple "AI personalities" (eg; Age of Empires 2 had this) or easily merge bot AI into a multiplayer game which includes no bots to start with. Many singleplayer games never even release SDK's. Console released games usually give PC ones less of a priority too. Some do get bots added even if no multiplayer code exists for it however, TFC as an example, Half Life via. Sven Co-op as another, and others. Tough to do from scratch, but doable. I'd not say AI has got much more advanced or needs to be more advanced then TFC's bot AI was. [b]Counterpoints[/b] There [i]is[/i] a heavy amount of mods being made, as I've said. There is likely an AI mod for almost ever recent game which has easily accessible AI functions. [b]Answering the persons original question[/b] I think out of my experience, and it's not got any FPS bot experience (except when I tweaked an existing bot in TFC), these are some possible games with "easily accessible AI". [list] [*]RPG: Neverwinter Nights. Completely scripted in "round based" AI events for all combat and non combat behaviours. You could code almost anything into it since there is DB support, allowing many academic things to be put into the game (see recent AI Game Dev articles). I've got combat AI in the works, but totally on the backburner since I'm working on other things. The game is also very cheap now, and allows for a lot of non-combat behaviour implementations (eg; ambient stuff going on, chatting between NPC's, tag games, etc.) [*]RTS: Dawn of War. Since I moderate the DOW modding section on the Relicnews fansite, I've gotta bring this game up. It's AI is run via. a set of LUA AI files, and if you get to the bare bones, simply calls a new tick every 1/8th of a second to check what to do next. Some parts are basic - there are bugs in some AI functions - but outstanding work by modders on the [url=]AI Skirmish Mod[/url] has got it working to a proficiency I've not seen by many RTS AI's. Main downside is lack of player interaction (no dynamic alliances, etc.) and some bugs which makes AI pathing somewhat poor. [*]RTS: Age of Empires 2. I can't even run this game anymore, it for some reason won't run and crashes before loading (???). Anyway, back in the day, it was a prime contender for showing off RTS AI. The best part is it is very simple, and entirely rule based. There is a huge selection of existing AI's for it, and most of it is highly documented. It's also very cheap to buy, and has random maps making the AI need to accommodate many styles of play. If you're interested, Starcraft had a similar rules-based AI implemented but hardcoded, so it can have some parallels drawn to that. [*]TBS: Civilization 4. I'm going to wait until my copy of "Complete" comes through, but there is now SDK tools to at least look at the AI in the game, and then modify it. Turn based tactical games like Civilization, with random maps, and needing long term plans but also a need for personality and differences between upwards of 20 AI's running at once, is a difficult problem to tackle. I certainly need to look at this myself. Some modding has gone on to implement new campaigns, scenarios, time periods, factions and other things - the original game includes some nice varieties of scenario too. [/list] FPS wise, I might suggest Half Life 1 and check out the Sven Co-Op mod, which improves AI creatures and is basically "multiplayer singleplayer" against NPC's. I've not looked at the AI in the game myself, but it's a good mod and will have the AI available. One map specifically is very AI based, and a new release of it coming out soon (see it's homepage) looks a lot more impressive then previous versions. Other then that, FPS bot games, Unreal Tournament (original or 2004 probably) and Quake III Arena (which you can get the source of now) are classic ones which already contain bots.

JonBWalsh on January 8th, 2008

It's definitely diminished though. I think there's a combination of factors. 1. Technology advancements make it harder to get into a mod and get things done. You need more people working more hours to make progress. 2. More robust selection of games. Think about many of the popular mods that used to exist (such as counter strike, team fortress, etc.). Now there are mods or games around that fill the same game types or concepts that mod makers used to fill. Now you see mods that seem to go further out there in concepts are fill more of a niche game play (Zombie Panic, Gary's Mod, the ship, that mod where you build your own fort out of panels before the game begins). 3. Growing gaming industry and advancements in education. I think the fact that the gaming industry has gotten a lot larger and the education around computer science and game development has caused interest in mods to decline. Now people who are passionate about game development likely have a much better chance to both get into the industry (maybe not a AAA studio but there's plenty of smaller ones now a days) or can more readily go for an education that will lead them down the path towards game development. As for AI bots specifically I think the need has declined. With more people having access to high speed internet, better community features and game browsing, and a larger spread of single player games people aren't as likely to want to play against bot AIs. It's so easy nowadays to play against human opponents (the best AI of all :P) that the motivation to create custom bots to play against is pretty much all dried up. The only reason to do so now is because of a passion for AI and game modding. Even then though creating game content that no one (or very few people) are going to use can be pretty bleak work.

DoctorMike on January 9th, 2008

This is definitely one area where students on A.I. or Games courses - we run one called "Games Development and A.I." which we think is the only one where both are done together! - have a definite chance to contribute. Making games is an important part of the latter type of course, and I believe that there is a real need for A.I. programmers in the industry; hence why we launched the course a few years ago. Maybe student projects will be the last remnants of the scene, unless people are really committed either to using mods as a spring board into the industry, or are that peculiar type of masochist that likes to code in the jungle of K-LOCs. Mike

DoctorMike on January 9th, 2008

Shameless plug: for details.

Ronin on January 9th, 2008

I can't speak for RTS & FPS modding. But when it comes to RPGs there's alot of modding going on. Including total conversions, AI tweaks, extra content and such. As mentioned above, Neverwinter Nights(NWN) has a very good modding community, though the 1st one is bigger than the 2nd one. And The Elder Scrolls series. Especially Morrowind and Oblivion has tons of mods being made in all categories. One could say that multiplayer has killed modding, but NWN is played online, and there exists tons of mods for multiplayer mode too. Like one total conversion I tried, set in Middle-earth. So that leaves out the multiplayer factor for RPGs at least. The one thing the NWN and TES:series share though is that they both have incredible toolsets and freedom to make any content you would want. Including changing most if not all the rules of the game. And they also have advanced AI scripting for NPCs, which makes the whole modding process a lot easier with several other SDKs. I guess more advanced SDKs tend to scare away the casual modder, while toolsets are more easy to get into and gives fast results (depending on what you make ofc.. ) Note: I mention TES and NWN only cause those 2 mod communities are the ones I've got most experience with.

DoctorMike on July 14th, 2008

There is a netbeans based server for networked AI for UT2K4. One of my students used it for his final year project. I will look up the details and post them here tomorrow.

DoctorMike on July 14th, 2008

I agree about the boards, etc, as my students often beat their heads against walls trying to do A.I. in their engine of choice. Questions are often asked, but answers are rare, sadly, when raised on SDK or related forums. People want to keep things close to their chests, rather than sharing the knowledge. I will try to knock something up on HL2 this Summer. As for NetBeans, it is a java extension. The "thing I cannot remember the name of" then allows a server to run that UT2k4 can connect to, and A.I.s can be written in C++ as well as java, using VS2005 or better, but I will root out details tomorrow.

If you'd like to add a comment or question on this page, simply log-in to the site. You can create an account from the sign-up page if necessary... It takes less than a minute!