Deathmark Postmortem: a Game Where Everyone Dies

Check out and play our game here


Our team is composed of Tyler Stokes, Nick Babcock, Ian Propper, Tommy Seidel, and Mark Mevorah. We were tasked with creating an original game that would be publicly showcased at the end of the semester.

Initial Phase and Brainstorming

Our team of five formed soon after the final week of individual prototypes. The week before the final project was assigned we convened in a meeting to discuss our future game. To get a head start, we decided to build off one of our prototypes, as that would allow us to take advantage of work already completed. One of our teammates, Ian Propper, by far had the most impressive prototype. He was able to utilize a package created by Dastardly Banana Productions, LLC for Unity called FPS Constructor, which provides the foundation for creating a first person shooter game. From this foundation, he added the unique concept that upon dying, the player is teleported to the location of their most recent victim.

All of us agreed that our game should be based on Ian’s prototype. We came up with ideas for improvement, such as an AI, a checkpoint system, and additional levels for the player. Positions and responsibilities were soon assigned, but later many jobs were merged in an attempt to facilitate progress in the few weeks that were allotted for us to finish the project.

Our first deliverable was to brainstorm a title and logline for our game, and we settled on the following:

Deathmark: A fast-paced first-person-shooter and puzzler game. Players make a deal with a Shinigami, a death god, to become a shade; a soulless creature who, upon dying, teleports to the body of its most recent victim.

Mixed Feelings

While having the third party package was initially a productivity boon, there were growing pains. The entire package was written Unityscript. There is nothing wrong with Unityscript by itself as a language, maybe outside of verbosity. However, in the Unity editor, Unityscript is referred to as javascript, but this is a severe misnomer. Unityscript is a classical inheritance language with other Object Oriented features such as type annotations, whereas javascript is prototypal. Newcomers, not knowing the difference would find seemingly conflicting information on the internet. All members of the group came from a C# background with past prototypes, and dealing with Unityscript was cumbersome at best. If the languages that Unity accepted (C#, Unityscript, and Boo) could better interoperate with each other, this may have not have been as large of an issue. Unfortunately, there is only a very narrow and inefficient method to achieve communication across languages, so we stuck with Unityscript. Briefly, an idea was tendered of rewriting the foundation in C#, but this was decided against because of time.

We decided to store our code in git hosted on a private Github repository. As discovered in previous Unity projects, this doesn’t come without difficulties, as Unity extensively utilizes binary files, which are impossible to merge given two different files. To reduce merge conflicts we adopted an impromptu naming convention where we would store binary files under our individual names. However, due to several members and their inexperience with git or situations outside their control, issues with git arose. Thankfully, the team’s git expert, Nick Babcock, shouldered the burden of straightening out issues, and in the end very little work or productivity was lost. Using source control, overall was beneficial, but occasionally it was taxing on enthusiasm and time.

Github wasn’t without its problem either. Tommy created a giant cityscape in the 3D modeling program Maya that ended up being over 100MB in size. He committed the cityscape, but then realized that 100MB might be excessive and so he promptly deleted the file and committed again. Then when he tried to sync to Github the sync was denied because Github sensed a file over 100MB and Github limits repository to 1GB in size with a max of 100MB for an individual file. Thus, even though the 100MB file wasn’t in the repository at the time, git saved the file so that it could recover the file in the future if needed. Thankfully, git provides a mechanism to rewrite history, so that it could appear as if the large file was never in the repository in the first place.

Some of the design decisions we made had mixed reactions. For instance, the notion of teleporting to the location of the last person you killed can be an unnerving concept. We mitigated this by applying a toon shader to all objects, which makes the game look cartoonish and more approachable. Another good example is the aspect of ammo. There were multiple requests to remove the ammo, but we kept it in game for a couple reasons. The first reason is that having a mechanical gun without ammo would be foreign to most people. A laser would fix this issue, but it was decided that would go against the feel of the game. Another reason for ammo is that to remove it from the game wouldn’t be worth the effort, and so we kept ammo around. Ammo also presented a unique obstacle to players as they had to plan their efforts in advance so they didn’t have to reload while chaining teleports in mid-air.

Room for Improvement

The development of the AI didn’t go as planned. Originally, there were turrets that simply turned to follow the player and shot back when fired upon. We wanted to replace the turrets and their AI mechanism with a Unity AI package called RAIN. RAIN advertises their GUI based approach to creating a sophisticated AI; however, there was a steep learning curve. Many of the tutorials and the videos simply didn’t work because RAIN had recently updated, and didn’t retain backwards compatibility. We had Mark work full time on creating a level that used RAIN, while the rest of us focused on polishing the existing enemies. We ended up replacing the box model of the original turrets with a much sleeker version that Tommy constructed in Maya. Mark was able to implement a level using several soldiers in RAIN that were able to patrol an area and attack the player if the player encroached. Since Mark worked in near isolation on his level, when trying to combine his level into the rest of the world, numerous problems arose that prevented us from successfully incorporating it. Thus, the fruit of Mark’s labor can’t be seen in the main game and can only be seen in the repository.

If it hadn’t of been prescribed and required for us, our group wouldn’t have utilized the burndown chart. It is, in our opinion, that a burndown chart doesn’t accurately represent the fluidity that is game development, especially a game that is developed in a short amount of time by a small and close-knit team. Any question of progress was better answered in the next group meeting than through the burndown chart. Thus, the burndown chart was more of a indicator to the instructor of our progress, and even then it was an extremely rough estimate. Updating the burndown chart wasn’t ever a priority of the group. A couple hours before each class period, a team member had to remind everyone to fill in their portion of the burndown chart. In the end, the chart was a chore for our team.

Weaving a plot into our game was always an afterthought. The logline allowed for a number of intricate narratives, but unfortunately, due to a lack of time and skills these narratives were left by the wayside. If we knew how to create cut scenes, we would have created the scene where the player meets the death god and gains the ability along a pistol. Given the short period of time to implement the game, we decided that focusing on gameplay would increase the value of the game more than a cut scene.

By the end of development, the project turned into somewhat of a monstrosity. Maya and Blender had to be installed so that objects could be rendered. To view the data logged on the server, one had to dump the database into a specific location so that Unity could read it. Not to mention, at one point in the project a dependency on DirectX 9 was introduced, which we believe is because of the shaders we were using. All of this complexity negatively impacted build times and the ability of the team members to build the game. Only a couple members could successfully build the game so that it could be deployed. This took valuable time away from development. If we had known about the complexity we would encounter, we may have better prepared for it.

What Worked

Github issues. Not only did Github host our code, but also our asynchronous communication. Anyone could create an ‘issue’, a discussion would ensue, and after discussion was resolved the issue was closed. For instance, one of our earliest issue was determine whether our game title should be one or two words. In total a hundred issues were raised.

Collaboration workflow. Unlike previous projects, which were individual or partner based, having five people all working on the same code proves problematic logistically. It is paramount that everyone knows how the code is changing and to make sure that no changes corrupt the code that would then spread to the rest of the team. These problems were solved using what is known as Github Forks and Pull Requests. By having each team member fork the master repository, any syncing of a member’s code would only effect their remote repository and not the master. Then if the member wanted their changes propagated main master branch, a pull request was sent to the master repository. Another team member would then review the code and make a decision on it. This way, every single line of code had at least two pair of eyes review it.

Data logging was surprisingly easy. Our team decided we wanted to log the user id, the level, x, y, z coordinates, and whether their death was a deathmark death. Since our game is played through a browser, it was trivial to host our application on a server and then have Unity execute external code that would relay the information to the database on the server. This data could then be dumped into Unity’s editor and then display the deaths as spheres, so that we could visually identify where the majority of the deaths were occurring.

Playtesting was key in our development direction. We first playtested amongst ourselves, as we covered a wide array of skills when it came to first person shooter games, from Ian who has computer gaming equipment to Nick, who had never played a first person shooter on the computer before. This worked for a while, but we soon fell set in our ways, and it took a class playtest for us to really recognize what players wanted with our game. Whenever we had a playtest in class, we would pay attention to the user, see what their suggestions were, and then create a Github issue, so that the team could discuss. Some of our core functionalities came from these suggestions. For instance, allowing the player to commit suicide by a button was immediately dismissed in our team, but playtests revealed that players consistently utilized the feature when it was implemented, so we decided to keep it.

The team was very effective at communication and each member was enthusiastic. A team member could text the group that they were working on the project, and within the hour, all of us would have convened to their location. While remote work was theoretically possible, nothing beats close contact so that ideas can be bounced around and problems solved on spot. This was especially true when two people were editing the same binary file, as they had to communicate their changes to each other.


In conclusion, we are satisfied with the output of our game. We believe we captured an original idea, teleporting to your latest victim upon death, and created a prototype that displayed and stayed true to the idea. Our team operated effectively and there wasn’t a problem that we couldn’t solve. We believe that our prototype is viable for further development, and with enough polish could become a successful indie title.