Schizoid Post-Mortem

Today's Gamasutra cover feature is our Schizoid post-mortem.

Many more things than 5 went both right and wrong on the project - though I do believe that single-player issues (the fact we had no online co-op in the trial *and* that most people don't seem to play co-op games) were the biggest "went wrongs" and made the others somewhat immaterial.

Some more "What Went Rights":

The macro game design - the way we do checkpoints and gold stars, is a "What Went Right" - we've gotten a lot of kudos for this.  In our first round of kleenex testing, before we had checkpoints, we discovered that players were cheesing their way through the levels, dying repeatedly and using their shields to win.  They weren't using teamwork.  We didn't want to just take the shields out, because that meant when one player died the other was defenseless, and partly because using the shields that way was fun, in moderation.  So we put in the checkpoints.  Predictably, we hated playing some of the levels we'd already beat over again.  But some of the levels - the ones we hadn't mastered yet - we didn't mind playing again.  That's where the "gold star" system came from, so once you've mastered a level you don't play it again.  It resulted in two ways to make progress as you play - you can get to the checkpoint, or you can gold star a level and make it easier to get to the checkpoint the next time, and people love it. 

The network programming was a big what went right.  The bifurcated nature of the game lent itself perfectly to an asynchronous peer-to-peer network model, where each machine keeps track of the enemies that the player cares most about.  Combined with Microsoft's tools for monitoring bandwidth and their network playtesting lab where they caught some bandwidth issues we missed, we ended up with a really solid network experience.  I play it with my brother on the east coast all the time and it's excellent, and we've played it with people in Europe and it's been playable.

Another what went right:  almost no crunching!  Without having a publisher breathing down our neck we didn't feel too much pressure to work overtime.  And although it's possible that if I (the bottleneck programming resource) had worked more overtime we could have gotten the game done sooner, I believe if you want to keep making games on into your old age you have to learn to pace yourself.  Also, we want Torpex Games to be the kind of studio that doesn't crunch people, and part of that is setting an example.  I don't want to be the kind of boss who says, "There's no mandatory overtime", but then works all night and looks at you funny when you leave at 6 pm.

I also - and I've said this before - was really impressed by the way Richard and Skaff approached level design.  They took a small number of enemies and parameters and found all kinds of unique combinations, situations, and experiences.  Sometimes a level idea would come out of just a name, sometimes it would come out of "What happens if we pair these two kinds of enemies together", sometimes they were based on a particular challenge, sometimes they were variations on a theme.  I think if we'd employed their techniques to make the encounters for Spider-Man 2 it would have been a much richer game.

And some more "What Went Wrongs":

Is Schizoid an action game or a strategy game?  Although it looks like an action game, and reflexes definitely help on some levels, if you're losing a level, it usually means your approach is wrong and you need to re-think it.  We didn't do a good job of communicating this, and we probably lost some sales to people who threw up their hands and said, "That's impossible!" when really all they needed was a different approach.  At Torpex we love games that make you think, so in the future we'll probably make our games look less like action games...

The game was too hard.  We did do a fair amount of focus testing - around a hundred kleenex testers played the game - and it was clear that it was difficult.  In particular, the "Skulks" - the star-shaped enemies - are more maneuverable than you, and people aren't used to seeing that in a videogame.  They had to be more maneuverable than you to force you to work together - but did they have to be *that much* more maneuverable?  We experimented with toning down the skulks and found the game was less fun for us.  Combine that with the thought that XBLA gamers are some of the hardest of the hard core - N+ and Ikaruga are both very difficult games but hot sellers - and we decided to leave the difficulty where it is.  (I hinted at that with an earlier blog post, saying something like, "This game may not be for everybody, but if there's something in there you don't like, it's in there because we liked it.")  Looking at the leaderboards shows this was a mistake - people are giving up much earlier in the game than we expected.  We should have made it marginally less fun for us to make it more accessible to more people.  In the future, we'll take the words of Richard Rouse III to heart:  "Your Game Is Too Hard."  I'm sure it cost us some sales, but not near as many sales as the single-player problem cost us. 

As much as I like the level design, some of the reviewers have dinged us for repetition, and there a few levels that some people just plain don't like.  (One of my favorites, "Breeding Pits", has shown up on more than one person's "Least Favorite" list - ok, so it's really random.  But it's exciting as hell!)  We liked all the levels and although we cut a few it was painful to even cut those.  What we should have done is have kleenex testers rate each individual level as they played them, and then cut the lowest 10-20%.

Speaking of kleenex testers, although we had about a hundred kleenex testers, we could have used more.  (And maybe that would have finally convinced us to lower the difficulty.)  An advantage guys who make PC games had over us is they can distribute the game to their friends, whereas we had to bring people in to play on devkits.  But we did have a PC version that we used for development - we could have made it robust enough to distribute to friends, although the gamepad is still an issue, and they'd have to be friends who have another friend handy, as there's no online play in the PC version.  And do we get them gamepads?  I guess we'd have to.

The distributed development thing did have the expected communication issues - often Bill and I would play post office between James and the designers.  Jake Simpson has some good ideas for improvement here.  One thing he doesn't explicitly mention is that at the beginning of a project, Linden Labs will usually fly everyone in to the same place to work together on the project for a while before they split up back to their home towns.  That seems like a really cool idea;  it would have been great to get James and Brian and the designers to actually meet and work together in person for a while to kick things off.

Not viral enough!  With a competitive multiplayer game, you want to keep finding more people to play it with, which means more sales for the developer.  With *Schizoid*, you're particularly motivated to find one really good player and always play with them.  We should have had achievements or even a leaderboard that would encourage people to play with newbies;  the idea being you'd get points for helping someone else get points.  This would help introduce newbies to the game and give people a reason to recommend it to more than one friend.

Comments

[Darius K.]

Here's hoping you have some more balance testing metrics in the next game :)

Illuminating post mortem, though.

[Ryan W]

Wow, going through that post mortem again led me onto guy kawasaki and what an inspiring man that guy is so now I am doubly thankful for you and the post mort :)

Check out http://blog.guykawasaki.com/2006/06/the_art_of_the_.html if you haven't already