Post-mortem: end-of-2017 marathon

MrBond's picture

Hello everybody,

2017 is behind us, and it will not be missed.  We took the opportunity to send off the year on a positive note with our end-of-2017 marathon, a 16-hour stream event benefitting One Gamer Fund. We handily exceeded our goal by 50%, raising $750 in total. Thank you to everyone that participated and watched!

As with our marathon for Extra Life 2017, what follows is my postmortem of the event. Beyond a few cursory "soft" topics, I will focus on the software tech side of things, as that is where a comfortable majority of my time was spent.

First, a word on the restrictions we faced. Despite having more time (three weeks versus two) to prepare, a lot was done in a hurry. Again, we pulled off a minor miracle; and again, this miracle was based more on hard work than luck. This event was further proof that we can pull together to perform as a team. Now, on to some specific points.


Restricting the total event time to 16 hours (versus the previous event's 24 hours) was an excellent change. Unless and until we can recruit more participants and / or volunteers, future events will likely not exceed 16 hours total time - long enough to be worthwhile, short enough to not be unhealthy for everyone involved. As none of us are true "early morning people" (I think), starting at a more reasonable mid-ish morning and running late seemed to work better, as well.


Given our available participants, and games played, we had ample activity going on throughout. Like before, we scheduled out all games back-to-back, with no breaks for rest or even food. With two events' worth of experience, that will be an important change for the next, at the very least to acknowledge basic physical requirements, and account for any freak technical problems. Speaking of...

Now, the "fun" part - the software tech stack

This event was a major test of some custom-written software to handle dynamic data (like donation totals, schedule, music...) and composing our stream layouts. We received a lot of feedback - things looked sharp and appeared to run very smoothly - thank you for that! Under the hood, there is still a lot of work to do. Plans are actively forming for dozens of fixes and new features, a few major redesigns, and a whole mess of streamlining.

From a deep-and-dirty technical standpoint, it's all a mess - sloppy code, hacky workarounds, and very user-unfriendly interface. The important target - less manual intervention during the event itself - was certainly hit, though at significant cost to future-us. The system is workable for now, but can become so much more.

Scheduling was a major functional point. Instead of manually hacking up a text file, we managed it straight from a web interface using the Tiltify schedule information as a base. No more restarting server processes manually when things didn't go quite as planned; we even had to substitute a few games that were not initially scheduled, and that process worked well.

Receiving and displaying donation data drove most of the current "fully automated" portion of the tech. We received donations, updated our totals, and propagated to whatever layout(s) were active; not too much unlike our previous implementation, though we had more incoming data to parse and process.

Beyond those two specific cases, a lot of internal architecture is now vetted, though not implemented in its final form. A lot centers around receiving, processing, and transporting data, and doing so in secure, repeatable, flexible ways. I've always found systems interaction to be fascinating, and having this as a practical and useful application is exciting.

On the visual front, we're lagging, though not by too much. My philosophy (for general development, and here) is to "make things work first, make them pretty later". Thankfully, I delegated much of the "pretty" portion, so everyone could be spared at least some affront to the eyeballs. I anticipate future events will only become more visually attractive, assuming better minds (i.e. not mine) are put to the task.

Overall, I am very happy with the "test".  In many respects, the tech stack will evolve to interact with many outside sources, as well as drive not just our future marathons, but our individual streams, as well. In fact, if you are a lovely viewer of Tormod's stream, you may see some recognizable things in action...and those who partake of Quest for Semi-Glory can expect similar things in the near future, as well.

That's all for now.