So, while this might be more than amusing for Ross in Friends to utter as his friends try and move his couch up the stairs, it’s less amusing when it’s applied to video game development.
While it’s been a while since we updated here, it’s time to do another glimpse into AA Game Development.
In the interim, we’ve got full Android compatibility, we’ve pretty much completed our layouts for all the screens (or thought we had), we’ve sorted layout for mobile phones vs PC or Tablets, and we’ve been QA-ing and fixing bugs as they’ve come up.
We are content complete with Visualizers – including 6 more for post initial release – and and also all FX, Riff, Beat and Instrument cards. 3300+ in all. That’s enough for release.
We are at the point where we can start generating themes, but…
We’ve started talking about writing some of the tutorial, defined what it looks like and generally got our heads straight with what needs to happen.
But.
It’s become clear that our initial concept of 1vs1 games, where the game generates a music genre and the two players are then trying to create a track – with the card deck they currently have – that matches that genre, then judging it via pre-created meta data for that specific genre (see one of our other posts for information on that) is just not going to work.
Why? Because music is extremely hard to mathematically or programmatically quantify, it turns out. The meta data we wanted to generate – which, we have to admit, not being musicians, we weren’t entirely sure what that would be – turns out to be extremely hard to create, particularly for those who are musicians.
We talked to several musical engineers, and we got lots of responses like “I’ll know a key signature when I hear it”. When questioned, “Well, how do you work it out?” we’d hear “I just know.” Which is great for that person, but sucks a little when you are trying to teach a computer how to determine it. Add to that, a key signature can be different based on who is hearing it (apparently, they can be quite flexible in terms of what they actually are), and you’ve got a recipe for problems.
Part of the problem was finding meta data that would generate sufficiently different values per genre, so we could actually differentiate between genres. Turns out meta data can be quite ‘noisy’ and therefore hard to pin down and get specific.
Another part of the problem is that music is subjective. Enjoyment of it is subjective. And even differentiating between “this is good” and “this is bad” is hard, and computers cannot do that. It would be entirely possible for one player to create a discordant cacophony with Lassens Loop, and still win the game, because their tone deaf track contained that which the music analyzer would be looking for, and therefore marked higher. When that happens to a player, they’d be upset and annoyed, and justifiably so, and would almost certainly complain, publicly.
The other problem is that the game providing genres for players to try and create a track for tends to mean that people will only ever create tracks for those genres. They won’t be rewarded for doing weird and wonderful things and creating strange and edifying music.
So what to do?
Well, the solution is The Wisdom of Crowds.
So, what we are doing is ripping out the whole 1vs1 mechanism, and replacing it with Daily, Weekly and Monthly challenges. The idea being that we can generate challenges that vary day to day or week to week. Players can create a track and submit it for that challenge. They can only submit one track per challenge but can come back again and again, to edit it and change it. The same rules about card usage counts being decremented by one (regardless of how many times they are used in one track) apply, so we do get cards removed over time.
The challenge will last as long as it does, and once it’s done, then the challenge sticks around, but in a different mode. This one is “vote on it”. There are two different modes of voting – random voting and looking at the leaderboard of votes. Both award virtual currency when players do it, to encourage them to do so, although random voting will pay more than leaderboard voting (where you can see the leaderboard and select tracks from there, as opposed to random voting, where a track is randomly selected from the challenge for you to vote on.)
The challenges themselves can be anything, and we are only limited by our imagination. “Best 30 second drum solo”, “Best minute long trumpet based track”, “Best dueling piano track”, etc.
Those who are voted on receive virtual currency too, plus boasting rights and achievements.
We are elbow deep in now revising code to reflect this new approach, which we feel is much more public and social, and therefore better for the game.
Still adds 3 months of development though:( Oh well, making games is HARD.