Postmortem: Quality improvements


Our last lesson was that things always take longer than you think, but that was not the only one we could have drawn. Another, even more fundamental lesson, it’s this:

Get caught up on details that are not crucial is all too easy. If you do that instead of working on the core of your game, they’re not a feature, they’re a liability.

By the end of that long process I had a implementation I was happy with for generating random stages from a series of parameters, with nice looking blending that still approximated the board game look I wanted. But did I need all that? Did it add to the game?

I wasn’t fully leaving aside the first lesson (it’s not a game unless there’s gameplay), but I was clearly focusing obsessively on something that wasn’t directly gameplay-related.

Here’s how that happens:

  • You pick a area of the game. For the sake of the argument, let’s look at the stages.
  • You settle on a level of detail you figure you can handle. A few tile sets, plus some transitions.
  • You start putting them together. Getting something thrown together is easy enough (terrain here, change there, the end).
  • You notice something that isn’t quite right. Say, the prairie is too flat with a single texture. You then spice it up with varying textures, which looks better.
  • This brings to your attention something else – for example, you then notice that you lack transitions. You implement them, but the results do not match the aesthetic you’re going for, which requires tweaking. Now you need to add some extra parameters to your algorithm, or at least indicate more information on your base data.

And so on. The more detailed one part of what the user sees is, the more detail the world around it will need to fit. If everything were to be painted in broad strokes, like a cartoon, the mind fills in the blanks, but the moment you raise the level of detail somewhere the demands your brain makes of the rest go up as well. This means that an increase on the level of detail in one area will propagate and increase the need expected detail in others as well.

Mind you, I didn’t detect this pattern right away – in my mind, getting the stage generation to at least that point was fundamental. I wouldn’t actually identify this problem until further down the road, once I could map its effects on the project’s timeline and budget.

Meanwhile, it was time to put my producer hat back on.

I’ve always hated that idea of thinking of people as ‘resources’, as if everybody had the same strengths and weaknesses, so after the bad experience of hiring a company which behaved less than professionally and shifted people around as if they were exchangeable cogs, I was a lot more inclined to deal with individuals.

And so after spending my time playing with procedural generation, I started looking for artists again. I had had problems getting the Mexican team to understand the particular ethnic background of Romania, so why not look for artists who were familiar with it?

I expected there had to be some good Romanian character modelers – after all, Ubisoft, Gameloft and others have studios in Bucharest. It didn’t take too long to find Ovidiu Voica, who was then a modeler at Ubisoft. He was excited about the idea of doing a medieval project set in Romania which didn’t hinge on the usual tropes, and while he was tied up with Ubisoft full time, he could contract for me on his spare time. I was somewhat concerned that his focus switch, combined with the total time he had available, would cause us to progress slowly. He acknowledged the possibility, and suggested a co-worker at Ubisoft called Daniel Matei might also be open for contracting – in this way, even if each model proceeded slowly, we could have more work done in parallel.

We went over the concept art that the other company had produced (a lot of which I haven’t shown here), and his assessment matched mine – that would not do. He recommended another acquaintance, and I asked for quotes and portfolios from everyone. I liked what I saw, so I agreed to their rates and we moved forward.

Remember that were were going for a skirmish/miniature game aesthetic. I first wanted them to create the sculpture for the main character, a female warrior that I wanted handled with restraint, avoiding the cheesecake that is all too prevalent in fantasy art. The image I got back matched perfectly the description and references I’d sent.


Now we were talking business.

Ovidiu got to work on it, while Matei awaited the creation of the next concept piece. Having seen Ovidiu’s previous work, I couldn’t wait to see his rendering of the heroine. Unfortunately communication with the concept artist was a bit slow and lossy – he did not speak English fluently, so I had to communicate through Ovidiu – but a couple of weeks later we had a second concept for Matei to work on.

Warewolf Final 01

Both of them did an excellent job on the sculptures, even if a spike on their obligations for their day job with Ubisoft caused them to take significantly longer in calendar time than originally estimated. The delays gave me no small amount of concern, but regular progress reports and samples convinced me to wait.

Ovidiu was upfront about the issues – another comforting change – and we discussed them over a conference call. Him and Matei also wanted to propose re-evaluating our arrangement. They had originally discussed an hourly fee based on a number of hours of their estimate, but that estimate had been blown. What they suggested was charging me a fixed fee per model, since they now had a better idea of how long each would take. I could see how this would work for both sides, reducing my risk while allowing them to bulk up their portfolio with a level of detail they were happy with. As long as the quality held, we would all benefit.

Looking at what they had created, I agreed.

Heroine01 Heroine02 Werewolf Final

In spite of the samples they’d sent me, a part of my brain hadn’t really expected the result to look this good, probably out of the repeated disappointments. The quality of the modeling and texturing were above what I had originally planned, they were an great representation of the concept and aligned perfectly with the story background I’d written.

There was an obvious red flag, however: the whole process had taken about 6 calendar weeks. If this was going to be the case, I would have to seriously trim down the planned roster to be able to get something out in a reasonable timeframe.

That was but a passing thought, filed away for future processing. While Ovidiu assured me that things were normal now, and that at most it would take them one calendar-month per character (which would still seriously cut into my timeline), I didn’t yet have enough information to judge, as I had never commissioned models like these before. I decided to make a decision after the next batch, which would give me a better idea of what to expect.

Meanwhile, I had to find an animator to skin the models so that we could pose them, which got me wondering how they would look animated. There was also another nagging question: how well would the simpler terrain hexes look against them?

I didn’t question if I needed models this detailed. Needing wasn’t an issue, now that I wanted them.

Leave a Reply