Open Editorial

Was BULLETSTORM Over-Engineered? An Argument for Technical Innovation

Alex J. Champandard on April 11, 2012

At the Paris Shooter Symposium last year, the BULLETSTORM team made a great impression on many of the AI programmers and developers in the audience. Mieszko Zielinski showed off an event-driven behavior tree, the game's tactical reasoning system, and Jaroslaw Ciupinski discussed the animation-driven locomotion and the planning-based path following. Each of those, taken on their own, push the boundaries of what's been done before in games. BULLETSTORM as a whole does this too, with its innovative Skill-Shot design and ambitious use of physics simulations at different speeds in local spaces...

While the game was received very well critically, it sadly hasn't met similar commercial success, and Epic recently divulged that the sequel was discontinued. From that perspective, you can understand a question that came up after the Shooter Symposium 2011; one Lead AI Programmer discretely asked me whether BULLETSTORM was over-engineered and was this amazing cutting-edge technology something optional.

Here's the short version my answer: hell no! I figured this was an issue important enough to not treat it discretely, and since we've just released the standalone recordings from the Shooter Symposium 2011 too, it made even more sense to blog about this publicly :-)

Screenshot 1: Locomotion in BULLETSTORM is done by a multi-pass smoothing algorithm that starts at the navigation mesh level, and ends up with smooth looking curves. (Slide from the Paris Shooter Symposium 2011.)

The Benefits of Experience

The first thing to realize is that the innovation on BULLETSTORM's AI didn't come from nowhere... Mieszko for example worked at Crytek (among others) before as an AI Programmer, and no doubt expanded his knowledge of behavior trees and combat reasoning there. Crytek has been researching and developing its behavior selection trees for many years now, with each iteration improving upon the previous one. The same can be said for its combat reasoning system, for instance the cover system in CRYSIS 2 that Márcio Martins demonstrated at the very same symposium. Not to mention that Mieszko either attended previous year's Game/AI Conference, or watched the proceedings online here at

“By the time the team started on new technology, they had a huge head start.”

By the time People Can Fly started on new technology, with or without the Unreal Engine, they already had a huge head start. Not only could the team leverage its previous experiences, they also didn't have to maintain legacy systems and could easily avoid bad decisions made on previous projects. This is the ideal scenario for any developer, and emphasizes the importance of hiring experienced developers and investing in their continuing education.

Experienced developers don't go far out of their way to innovate.

Screenshot 2: Part of an event-driven behavior tree that was built for BULLETSTORM's enemies. The editor is built within the Unreal Engine. (Slide from the Paris Shooter Symposium 2011.)

A Programmer's Passion

Another reason that BULLETSTORM wasn't over-engineered, is that it was passion-driven. Each of the systems presented at the Paris Shooter Symposium was built by a programmer very passionate about solving those problems specific to their games, and pushing the limits of what they had seen done before in other games and engines. From the influence-map style of reasoning used in HITMAN 5, the useful debug visualizations in GHOST RECON: FUTURE SOLDIER, to the tactical navigation in BRINK... the pattern was the same.

“Why not let developers do what they are good at?”

Of course, there are some risks to following passion blindly and disconnecting from the rest of the team. However, when channeled within the context of the game, that energy has mostly (or only) benefits. It's that kind of passion that companies hire developers for, so why not let them do what they're good at?

Passionate developers innovate without doing any work!

Screenshot 3: The tactical query system in BULLETSTORM uses a combination of scoring and filtering to get the most easily tuned results and highest performance. (Slide from the Paris Shooter Symposium 2011.)

Recruiting by Leading

Leading the field by technical innovation also attracts talent, both up-and-coming developers and experienced veterans. The students and helpers at the Paris Game/AI Conference last year were all very keen to speak with Mieszko and Jaroslaw, and many discussions were raised among the Senior- and Lead-AI Programmers in the audience too. People Can Fly seems to be one of the few European studios not actively looking for AI Programmers; I'm sure they got many applicants!

“Most studios realize the benefits of being open about technology.”

Over the past couple years, we've seen most studios realize the benefits of being more open about their technology. In fact, the studios that keep their technology close to their chest rarely do so for technical reasons... But that's another article. If you're going to build the technology, your studio might as well leverage it outside of the game too!

Innovative developers attract other skilled and passionate developers.

Mitigating the Risks

There are of course, risks to technical innovation. Luckily most of those can be mitigated, in the same way that People Can Fly managed to pull it off with BULLETSTORM.

  • Grounding — The biggest way to reduce innovation risk is to use a known starting point! Stay on top of recent games and their technology, use specialized sites to do that...
  • Direction — Another problem is to avoid heading in the wrong direction, which tends to require interaction and feedback from other developers, for instance at specialized conferences.

At, it's our goal to provide all this for developers passionate about artificial intelligence, character animation and behavior in general...

Shooter Symposium 2011 Recordings

If you've reached this part of the article and are thinking to yourself you missed out on the Paris Shooter Symposium last year, then you're probably right! That said, you're in luck. We've compiled all the recorded videos and slides together, and made them into an amazing standalone product with a full day of high-quality videos split into four segments: animation & locomotion, behavior logic, combat reasoning, and design & authoring.

There are a few ways you can get your hands on it:

  1. You can purchase it standalone from the store for the price of 397€; it's at 20% introductory discount until May 1st.
  2. As per popular request, year-long PREMIUM memberships without subscription are now available, at 997€ for GOLD. You get the Symposium for free until May 1st!
  3. If you're currently a PLATINUM Studio member, or you've been a GOLD Studio member for over a year, then the symposium is now accessible for you at the Symposium page

Your Thoughts?

Do you think it's useful to let programmers work on what they are passionate about? Does technical innovation have benefits for studios, both within the game and outside?

Post a comment below or in the forums associated with this article!

Discussion 2 Comments

bvanevery on April 15th, 2012

Without reading anything else about the game, this says to me that AI technology cannot by itself make a game profitable, no matter how innovative it is. It seems quite possible that a studio could overemphasize AI at the expense of other game virtues. That's a risk, and a reason not to just turn the passionate AI programmers loose. The internal studio judgment would have to be, does the AI team have too much influence on the direction the product is taking?

hoose on October 8th, 2012

Do you think the non help fron within some companies can be ocasionally has been created cause of the paranoic thinking of the final owners?,i mean,money,or when you talk about the free or open software,are you adding to piracy the fact of non earn money with some features because they are open ....thing so helpfully by the way,i mean,if i sell a game and earn what is supossed to be a good price,can i leave for users or others developers some parts of technology?...well...if you do,your next game will be 2 or 3 times bigger...i mean,i will use ideas and even code lines from a lot of people,i will know what people or buyers want...i can let millions of persons around the world extracting even the lines from the enviroment...the big one you buy to built your game...well...they(companies) will not pay...users will see what they dont let us to do this....a game should be well priced....the pc should be well priced...them will be no piracy...they invoking the excuse they need to make,how many advances in graphical aspect without having to use 3 or 4 conected pcs do you think can add to enviroments?...players,advanced players,want time cycles,want shoots on walls and leaves moving with the air,they want the dust when walking...but,they,for sure,want the most aproximated real reaction from NPCs,i don't know if the enviroments are created some years ago when people only wanted to use their new 2 GB video cards...i have played with a lot of graphical efects and once you play that way,you need that image in your screen...i mean...visually the game must be the best...when talking of AI...i really don't understand why you separate AI in 2 or 3 ways when tactical or combat ascpects are like when NPCs don't fight,you should be able to set combat within non combat reaction..thats all...don know how many options or capacities a NPC is programed to do...if you let your imagination fly,NPC will even recolect visuall information,but you need a lot of process i guess,when a NPC look to the front,what calculates(or think)?...just what they see in front confronting to what they need,for what reason they are in the game?,what they do? can write and write thousands and thousand code is my bet,how many memory they will use?...30 NPCs calculating like thinking will eat your pc...just a few,no,or what game you have in mind?,a sofisticated war game with a lot of thinking soldiers inside?,some NPCs better programed than others?,a better AI tree to let them resolve the screen just using,exactly,the needed line in that moment?When a NPC resolve something are you using the hole AI tree...or the AI tree is not running except those lines?...let your brain do what you want play.........I would say that I wish you luck but I think what I want for you is the chance to work on those desire aspects without hurry to earn money

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!