CAT | Puzzle Retreat

At the 2013 Game Developers Awards hosted by the GDAA, Puzzle Retreat was awarded with the Accessibility Award. Not only is it an awesome and prestigious award that we like showing off in our office, but it marks a very important milestone when it comes to game development. Film Victoria and Screen Australia now consider accessibility when it comes to providing funding and are now rewarding companies that excel in developing games that are accessible to a wide audience.’

We think this is pretty swell.

Accessibility in gaming has always been a topic of contention. How does one make a game that caters towards people with motor, cognitive, hearing, speech or vision impairments? Mainstream games usually shy away from this demographic in favour of the masses.

In terms of our games, we aim to make them accessible to those living with impairments. We believe everyone should experience the joy of gaming!

Puzzle Retreat was designed from the ground up with that philosophy in mind.

Sometimes games (particularly puzzle games) rely too heavily on language, small icons or graphics that are make it difficult for players with certain types of vision impairment or  difficulties with language to be able to understand and follow. We’ve attempted to alleviate the problem by using large and bold icons that can easily be differentiated. Furthermore we tried to make the game playable without understanding any written text.

There is an definitive association between time limits and penalties with puzzle games. I’m sure you’ve all felt the frustration of nearly completing a level, only to have the timer run out on you. We decided to take a different route when it comes to unforgiving scenarios.

We eliminated them entirely.

Puzzle Retreat allows players to take as much time on an individual puzzle as they’d like, reset it as many times as they want and even skip the puzzle entirely. Puzzle Retreat was designed to be a relaxing puzzle game, so it only felt right to dispose of time limits and penalties.

We’ve also tweaked the detection radius of the blocks so that its extremely forgiving when a player misses a block by a small margin. This feature, plus the removal of the timer allows players who don’t have a range of fine motor skills to be able to enjoy Puzzle Retreat.

We at The Voxel Agents are extremely excited when it comes to the future of gaming in Australia. With so many awesome studios producing games of such high quality and Film Victoria and Screen Australia providing consideration for funding to those who place emphasis on accessibility, we can’t wait to see what gets released in the future.

This is Agent Aiden, signing out.

, , , , , , , , Hide

Puzzle Retreat as we know it today is a far cry from where it began in May 2011. It started as a third-person game about felling trees, and it finished as a relaxing minimal board game about sliding ice blocks. In between, it ventured into a massive variety of themes and styles, including one where you were responsible for unfurling dragons by the pool so they could sunbake. It’s had explosions, bad guys, tractors, floating islands and even storylines. The game you play today was only possible with eighteen months of refinement, simplification and a whole lot of love from a creative team striving to make the ultimate logical puzzler for mobile. This is Puzzle Retreat’s game dev story.

The Pitch

Yangtian Li pitches an idea for a game

Lumberjacks cut wood.

Carpenters make things.

Yangtian Li, our in-house artist at the time, pitched to the team an elaborate design for a lumberjack-come-carpenter game. You fell trees in the forest, bring them home and make furniture. You can then unlock, sell and buy different design schematics, paints, flourishing details, and then trade what you make with other players online.

Add a Splash of Puzzle

Henrik Pettersson was immediately inspired by the puzzle potential of felling trees in a forest.  His first design was a puzzle game where the trees fall into each other and knock each successive tree down dominoes style. The second design, and eventual winner, focused on your player character who stands behind each tree to push it over. You must have enough space to stand behind the tree to push and there must be space for the tree to fall onto. This puzzle design requires you to find the right order to knock all the trees down whilst keeping the appropriate spaces free, and not locking yourself in.

The first concept, and it starts with a forest of trees!

We really liked the potential depth of puzzles this mechanic presented, and the simplicity of the interaction. Playtesters were scratching their heads and smiling, and we could feel the potential of this game really standing out – it’s a brain scratcher that can fit into a few minutes a day on a mobile.

Save the Ozone

Icons avoid the problem of tall trees


No tags Hide

Over the past year our studio has continued development on our TOP SECRET grid-based puzzle game.

One of the major and on-going challenges that our team has faced during development was the creation of high-quality handcrafted puzzles.

Agent Matt showing the rest of the team how one of the variations could work

Initially, we experimented with new puzzle variations with cardboard and a set of poker chips. It empowered designers to prototype rules earlier with no upfront code investment (while code was spent on building the actual game). We also couldn’t deny that it came with super cheap “save features” with the use of a handheld camera or pen and paper. However, there were downsides to creating puzzles with cardboard and chips. It took a great amount of time to test puzzles with oneself, within the studio and with playtesters on the streets of Melbourne City.

Pen and paper prototypes in notebook form
was another alternative to cardboard and chipery!

Designers had to make sure that each move made on cardboard was legal and because a computer wasn’t dictating how the moves were made, it was prone to human error and caused creation of unsolvable puzzles and puzzles with unintended solutions. Playtesting within the studio was also a lengthy process, it involved designers restarting the board manually by hand after each play. Since a computer wasn’t dictating the original layout of pieces, this process was also prone to human error and sometimes caused awkward moments when it was realised that a puzzle being tested was unsolvable. Getting our designs in game on a portable device to take to the Melbourne City streets for playtesting wasn’t easy either. We had to use one of our Legacy Flash Level Editors, which we were no longer supporting, to paint out our cardboard prototyped level and then export it, which took about 5 steps before being able to play it in game.

Our Legacy Flash Level Editor,
it was pweeeeetty and initially helped us get the levels in-game

Our first working Unity Level Editor, had paint, erase, load and save features (similar to our Legacy Flash Level Editor). It also saved puzzles in XML (a format that our Legacy Flash Level Editor supported). The Unity Level Editor had a first working solver, which made Unity crash a lot and our designers refused to use the Level Editor until it was fixed. We did see potential in it and persisted to shove in and rip out new features, one of the most significant features was the Solver.

 Our first working Unity Level Editor, we previously supported in-editor revisions!

Like a Phoenix, the Solver died temporarily (ie. removed), but soon after it was reborn (ie. reimplemented), but instead of turning into the same Phoenix it once was, it was reborn into this insanely powerful tool that changed the way designers now go about creating puzzles in our studio to date. The Solver had the ability to tell us how many solutions existed to solve the puzzle and if there were any solutions that were unintended, which we call illegal. A puzzle that was found to be illegal would never make it into the game. Designers were able to create super difficult puzzles that would have taken a day each to make and now they were being made in less than a hour. If a designer didn’t know how to solve the puzzle that they created themselves, they could request for the solution to be played out in Unity’s play-mode by the Solver.

The Solver was then made to play the game for us on device, it was an amusing sight to sit back admiring it’s beauty. It looked as if our studio was haunted by really clever poltergeists that have possessed a bunch of our iPads and was playing and solving each puzzle within a matter of seconds without making any incorrect moves.

A reenactment of the really clever poltergeists hard at work!  

After a couple weeks of non-stop puzzle creation our designers got really accustomed to the Unity Level Editor and following this we had a discussion about our puzzle creation process and where to take it next. From this meeting, we decided to optimise the process further by adding a list of solutions to the puzzle being edited within the Level Editor itself without designers having to go into Unity’s play-mode. It allowed designers to tweak their puzzles and to see the effect of their changes on the final solutions far more rapidly. Previously designers had to wait a whole minute each time they needed to test out a change that would have affected the solutions to a puzzle. Also the newly added visual representation of a solution communicates itself more quickly and clearly as to where each piece in the puzzle fits on the grid to the designer in comparison with how it was previously in play-mode.

Our Unity Level Editor, functionality over form.

If you found this blog post informative, please visit our Puzzle Retreat Facebook Page to Like or Comment.


Settling In!

We moved into our new office in Collingwood last week and we’ve been settling into the new space. We’ve had to deal with all kinds of things such as a brilliant coffee machine, a bit too much sunlight, and everyone having to work out how exactly to get to work.

The new space is great. We have a lot more room to move and it’s closer to home for most of us (except for Sam… sorry Sam) and it’s right near Melbourne’s funky Smith St. Maybe we need to start being more funky so that we all fit in?

Levels! A whole bank of levels!

We’ve been making a lot of levels for our next game. So many in fact that just managing all of them has become quite a chore (hundreds and hundreds of designs that could make their way into the final game).

To deal with these numbers, Agent Sam has made a really nifty tool that allows us to easily sort and structure our levels, give them ratings, and easily view information about each level. This was all done using Google Spreadsheets and some nifty Google Apps Script. We’ll have to write a full feature-blog post describing how great Google Apps Script is for making simple tools as it’s proven to be a very versatile weapon we can use in our quest for making better games.


Another cool thing we’ve made by hacking Google Spreadsheets is to create nice reports about who is actually playing our games and how long they are sticking around. This info is great because it helps us make our games a lot more fun, but it can sometimes be hard to really read the graphs that Flurry make. So we made our own graphs!

Even more awesomely, we were able to use Google Sites to automatically pull these graphs from the Spreadsheet, and display everything in a somewhat nicer format.

Leveraging Google’s services has proven to be a really great way for us to rapidly create useful tools that enhance our ability to make games. More on this in a future blog post feature.

We made some really hard levels, couldn’t solve them, and so made the computer do it for us

When designing levels, it’s often very difficult for the level designer to be able to keep all the variables and permutations of the level in their mind at once. It’s possible for us to make much higher quality levels if we have an automated “solver” that can solve our levels for us, as well as providing very useful information about a level (such as the number of possible solutions). We’re also kind of lazy.

So we made a “solver”. This has actually been a work in progress for quite some time, but it only just recently became awesome.

 This tool was made by extending Unity through it’s great editor features. We really strongly recommend that other Unity developers get in on this and start making wicked tools by extending Unity.

And then we made it even more awesome

The next step, of course, was to make it automatically show us the possible solutions in a way that was highly readable to a level designer. Agent Sam is doing a more detailed write up about this, but basically we can learn so much about how our levels are structured through this tool and it makes it much easier to make great content.

We quietly started a beta test

Agent Henrik began work on a beta test for our new game. It’s a closed beta at the moment and I am sorry but it is now closed.

Sorry about that.

On the bright side this means that our new game is getting close. We still don’t have a definite launch date but it will definitely will be released sometime.

We also started thinking: “What’s Next?”

The new project is in it’s final stages so we’ve started thinking about our next project. We have some new IP that we want to develop further into a new game, and so we’ll definitely start work on that.

But we also see a lot of potential to improve Train Conductor, so we’re going to explore that. It’s still early days so we don’t have totally concrete plans at the moment, but we’re thinking of making the newer “Challenge Mode” the focus of the game since that style of play seems to be much more the kind of play we were trying to create with the original Train Conductor.



The best art style for a logic puzzle doesn’t get in the way of a players ability to see and solve the puzzle itself.

Agent Tian and I researched what art principles were used in defining art styles for logical puzzle games. This information will be  relevant to other small development studios, as well as artists and students of game design.

We gathered data about the audiences of various games from the Facebook pages and user reviews of those titles. We found that Train Conductor 2, being a score based arcade game, attracts a much younger and tech savvy audience than is common for logical puzzle games.

Here are two typical sets of demographic data we’ve gathered from users interacting on the games facebook pages:
If you’re unfamiliar with Quell it’s a fantastic logical puzzle game that is performing well on Android phones. On their website you can read about the Making of Quell where you can read their development story.

Train Yards is a successfull logical puzzle game developed by Matt Rix. In his postmortem at GDC this year he talked about the difference between developing for casual players and harcore players. The slides are available for free here and you can listen to the full talk if you have access to the GDC vault here.

When developing a game for gamers you can skip several levels of teaching. With 10 years of gaming experience comes vast a priori knowledge of computer interfaces and how they usually respond to player input. For example, I’m sure you have, at some point in life, been completely mind boggled by the inability of an old relative to move files across folders.

We’ve seen endless cases, while playtesting with a casual audience, where the play tester develops the most obscure theories of what the game rules are. The cause of these theories is simply miscommunication by us, the developers. By stripping the game of unnecessary art assets we can greatly reduce this problem: less art assets mean less objects for players to develop wacky theories about.

We can see a clear connection between logical puzzle games that value graphical prettiness more than usability, and bad chart performance. We see the opposite effect when usability is considered first. It is clear that puzzle games that value usability over art outperform on the marketplace. Sudoku is a good example because there so many Sudoku apps (Appannie gives 823 results).

The players top favorite Sudoku game out of those 823 competitors is the example to the right. Their secret? They have the largest buttons possible.

The GDC talk “How to make your player feel smart” by Randy Smith (available for free here) gives advice about puzzles in games. He references the book “The Design of Everyday Things” and further emphasises the following principle:

Look at the example on the right. Though you might not be sure of the rules or the objective at an initial glance everyone will know how a car moves and player will try to interact with the element in that manner. Logical puzzle games that stick to this principle are clear favorites among the players. For example, see Traffic Jam & Cross Fingers

I will hold off with the remaining slides of the presentation. They rely heavily on statistics and I don’t want to bloat this post with pie charts. If you like that sort of stuff send me an email ( and I can provide you with the whole presentation but you will have to draw some conclusions by yourself. I will publish the remaining data with descriptions sometime in the future.

No tags Hide

In 2011 we made loads of prototype games; some small, some funny, some that sucked, almost all ugly (except the lucky few that receive Tian’s touch :D). Here’s a visual tour of 20 of the 24 games that we made in five months. Together they paint the picture of what Voxel Agent games look like when they’re born – a mish-mash of squares, circles and terrible colour schemes!


, , , , , , , , , Hide