As it was originally conceived, Rollers was a more straightforward arcade-style game, but we veered from that path pretty early on as we dreamt about a larger-scope project. We experimented and added features and unique scripts one at a time, playing with them and throwing out ones that weren't as fun.
We ended up with a wide variety of character stats and abilities that sat on the separate character objects. When we got production funding, we planned to beef up stats and abilities, but never took a step back to look at what we had -- we were planning to add and work them all into a cohesive externalized system. This made design, implementation, testing and balance more difficult, but we only realized our mistake when it was too late to change it.
This even haunted our UI process -- it made our Port (where you hire characters and upgrade them) UI/UX a challenge, and we went through several major versions. Our final version is the best of the bunch, but we're still not 100 percent happy with it.
Most of the game's major systems aren't wired up to external data sources, and are pretty hard-coded, so we were never able to implement a difficulty system or tweak game balance remotely. Future games won't get past prototyping without a nice neat external data system.
When we agreed to add the PS Vita as a launch platform, no one on the team had ever developed for the PS Vita before and we vastly overestimated its abilities. We originally estimated that 30 percent of our time in the last stretch would go to Vita-specific optimization. The reality was that about 80 percent of our time went to optimizations, from graphics to physics and other systems. This was even after we transferred extra resources onto the project to help.
Choosing to develop with Unity, while great for so many reasons, also hurt here... because we didn't have access to the Unity source code, this meant that when we needed to comb through the game looking for deeper and deeper optimizations, Unity itself was a black box.
As a result, we missed the mark on polish (all platforms) and Sony hardware-specific features. We didn't end up having time to fully exploit the capabilities of each platform, so our PC and PS4 experiences are not as visually robust as we would have liked (we're using Vita-friendly shaders and effects on all platforms), even if the performance is excellent due to extreme optimization.
We took a big risk in combining two very different game forms. We didn't expect to satisfy hardcore pinball or hardcore RPG fans -- we figured the game would appeal to mid-core gamers. Based on early reviews, opinion differs greatly (we've got superfans, haters, and everything in between). We're not sure if trying to find a new audience with this mash-up was the best way to go for our first major title.
As mentioned in our "what went right" section, our playtesting provided valuable insight for balancing and audience appeal, but the testing was mostly limited to the first few chapters. We misjudged the game's difficulty curve and the willingness of players to solve some of the game's strategy puzzles or manage party members and their upgrades.
On its face, the game appears to be similar enough to pinball -- many people believe anyone can play pinball, and this is true -- but mastery is another story. Our game doesn't look like it requires a lot of skill, but it does.
Our final boss battle was, for most players, punishingly hard. To add insult to injury, failing the last boss in the series could cost about 30 minutes of time and a replay of the entire battle. This left a sour taste for many players and reviewers. We added checkpoints in the final battle in an early patch for Steam and hope to bring this and other improvements to the Sony platforms soon.
Our "System Failure" didn't make it easy on ourselves to tune the overall balance, let alone do it remotely (via external data, which would have allowed us to nerf a few things without patching). We won't be making this mistake in future projects (this is something we've already institutionalized).
We were eager to move from our original 6-month prototype into full production -- maybe too eager. We front-loaded resources to try to overcome our potential Unity learning curve, but everyone picked up Unity so quickly we ended up with a bottleneck in programming. Our designers were waiting to port over workable ideas to the Unity project from our XNA prototype and our creatives had to wait to test content before the basic framework was in place.
We pushed ahead with design before the core set of scripts and combat system was complete, which contributed later to awkward workarounds (both in design and programming) that were difficult to QA -- many of these were redone from scratch when the system was completed, which ultimately was a waste. Obviously some better production planning and resourcing could have helped here.
Throughout the course of the entire project, we created a number of demos related to industry events such as GDC, and sometimes ended up rushing features that we later wouldn't have time to properly polish. In the future, we'll plan our production schedule around key demos.
On the design side, our team worked through the game in chronological order, which meant the first half of the game got the most polish and finesse (as well as the most audience testing / feedback) whereas the last half of the game didn't get as much love. We might have solved some of this by picking the "star" levels from each chapter and working on those first, then filling in the gaps later.
It took almost exactly three years from conception to launch worldwide on all platforms. While a long production cycle proved necessary, it also gave time for issues to bloom.
We had planned to launch the PC version of Rollers in late 2013 / early 2014, but we pushed this when we partnered with Atlus USA and added two additional launch platforms (PS4 & Vita). Having the production timeline stretch out meant Rollers rubbed against the work-for-hire projects that we lined up because we need to keep the studio lights on.
We needed to split our resources into two project teams. Phantom Compass often works on multiple projects at once, and in many ways this is a strength, but in this case it caused more strain on finishing up Rollers than we would have liked and hurt our ability to polish.
Staffing was also challenging: From our original six-month concept through to finding financing for the full game we lost many of our original team members as they found full-time industry jobs. Onboarding dozens of part-timers and contractors in the early days of our prototyping was an expected, but nevertheless a strain.
Our lead programmer position was also a bit of a revolving door, with four lead programmers coming on and off the project over the development cycle. This cost time in bringing each replacement up to speed and resulted in legacy code kicking around that they were wary of removing.
We really couldn't avoid the long development cycle for this project, but there is value in keeping them shorter when possible.
Was it worth it? Definitely. This game gave us many "firsts." We have all learned so much. We love this game and feel grateful that we are able to share it with others now who might love it too.
What will we miss the most now that production is over? Having the word "balls" come up hundreds of times per day.
What's next? We have a three-year plan to slowly wean ourselves off of work for hire projects and transition to creating on our own IP 100 percent of the time. We love the Rollers characters and universe a lot, and would like to keep them alive in another game... maybe a sequel, maybe something else. But for now we'll keep on rolling.
● Phantom Compass
● Nov 18th, 2014 (Steam, PS4 and PS Vita SCEA)
● Nov 26th, 2014 (PS4 and PS Vita SCEE)
● Steam (Windows)
● PS Vita
Number of Developers
● 3-7 full time
● 30 part-time contractors over time
● 15 voice actors, including Willow the Dog
Length of Development
● 2-D Prototype ~ 6 months
○ First Meeting Nov 9, 2011
○ Prototype ready for GDC March 2012
○ Polished up for submission for Funding spring 2012
● Unity 3D Full Production ~ 2 years 2 months
○ September 2012- November 2014
● Approx. $700,000
Lines of Code
64225 Lines of Phantom Compass Written Code
+ 56680 Lines of Code from Plugins
= 120905 Total Lines of Code (ignores shaders)
● Unity 3D