Another of the nice side benefits that we get at Ensemble as being part of the Microsoft Game Studios is that from time to time we get looped in on internal post mortems from other MGS internal or published games.  Today I got a chance to read a pretty in-depth and detailed post mortem, and it was quite useful to compare some of those problems to our internal procedures.  For obvious reasons I don't want to go into any particulars here, but it was a useful reminder that so many issues can be difficult to see when you are actively in the midst of them. 

At Ensemble we're pretty religious about doing a comprehensive post mortem after every project that we ship.  Typically each of the departments just captures a torrent of commentary of what worked (and didn't) on the project, then the producer assembles that data into something more coherent.  By isolating the top trends and having a little separation from the intensity of shipping the product, we can often come up with a lot of very practical changes for our next iteration. 

Of course, sometimes we fall back into bad habits, but overall I think the process of continual self-improvement is a very healthy one.  No one can avoid making mistakes.  But a smart organization at least tries to avoid continuing to make the same mistakes over and over again -- that frees up time for us to make all new mistakes!

Over the weekend I figured out how to rip the soundtrack audio from GTA San Andreas, as I did for Vice City.  Yes, I am obssessed, why do you ask? 

Interestingly the complexity of the radio structure for San Andreas is significantly greater.  In Vice City there was basically one big loop which included all the transitions and advertisements baked in.  San Andreas, on the other hand, dynamically assembles the audio stream from audio clips that represent possible variations on intro and outro segments, topical news bits, and game-world state like the weather or time of day. 

In making my uber-completist version, I actually made two entire variants on the soundtrack, to cover the possible combinations of the intro / outro variations, as well as to split up the massive number of extra transitions and radio spots.  All those little details really did make the radio experience in San Andreas a lot nicer -- but they make my little ripping project more complex...

I picked up Chromehounds this week and I've been playing with some of the other folks from QT3.  You can read some of our exploits here.  It's quite an interesting little game, but full of strangely mismatched quality levels.  One moment, you are marveling at the amazing graphics, the next you are cursing the fact that there is no way to tell friend from foe, even on the radar.

It's a very hardcore game, the kind I almost thought no one made anymore.  You can configure your "hound" (ie, mech) with many, many, many parts, juggle numerous stats, play with tons of weapon configurations, and actually place your components in 3d (which matters for true locational damage).  It is crazy detailed, and that's definitely a lot of the fun.

I haven't played around too much with the single player campaign, but it really seems that multiplayer is where the action is at.  Good strategy and coordination seems critical, and the persistent world campaign is interesting.  Plus it's just a game best enjoyed with one's comrades.

We've been doing a lot of talking lately around here about the use of a "perfect scene" in our development process.  Back in AOM we used a lot of conceptualized screenshots for prototyping things like god powers and our water effects, but really didn't get a perfect scene, as in, a representative high water mark for the graphics in the game, until fairly late in the process. 

One of the approaches we are taking to move that earlier in the developmental process this time is working more towards hand-drawn (or painted) representational sketches.  That is to say, concept art not of units or environments but of what a hypothetical screenshot might be.  This approach doesn't always capture the details of representation and rendering perfectly, but can really help for composition and structure.

I view the role of concept art "screenshots" as much like other forms of a concept art... a modeler needs a direction to go in that a traditional concept drawing provides, and doing that at the higher whole-rendering concept level can provide inspiration and guidance for both artists and programmers working on how the world looks.

I was excited to see EA officially announcing a huge swath of games today.  EA, despite all their troubles, has an excellent sense of production quality and brand building that always makes their games competitive.  However, once I got to the list I was disappointed to see that a large chunk of them were for mobile phones only. 

I'm cool with good games for portable gaming devices, like the DS or PSP.  Heck, I was just playing Final Fantasy IV this evening, and spent a fair amount of time providing design feedback for the Age of Empires  game on the DS.  But mobile phone games, on the other hand, are quite frustrating to me.  The devices just aren't suited for them.  More importantly, I just find the cases where I have my phone but not my DS, and want to play anything more sophisticated than Minesweeper, are basically nonexistant.

Clearly there is an audience out there, presumably a lot of people who have phones as part of their normal lives and don't have a handheld gaming system.  But it's a segment that just doesn't reach me at all.  There's an Age of Empires game for mobile phones too... but I've never had the opportunity to check it out.

Games in their early stages, like the one I'm working on, sometimes develop wacky idiosyncracies during development.  Here's an amusing example... for many months now whenever someone would take a screenshot from within the editor, one of our units would appear prominently in the final image.  This alone might have been amusing, but as it turns out, the unit would always appear stripped of any techs the player had researched.  Due to how we represent the player's military units, that means he wouldn't have even his baseline civ-specific armor so he would appear, well, mostly naked.  Fortunately his modesty would be preserved by a pair of underwear that was painted into the base mesh.

Thus was the legend of "Underwear Man" born, making his ubiquitous presence in every screenshot we would make for internal distribution.  He was, as you might imagine, the subject of much amusement.  But all good things must come to an end, and today I finally fixed the bug, banishing Underwear Man forever.

I think my favorite bug of this form was back on AOM.  For a few weeks, at some random point in the game whenever you would train a new unit, all your existing units would identify it as an enemy and immediately turn on their new comrade, murdering him as soon as he exited the barracks.  There was something just so very wrong and yet amusing about watching your units act in such a twisted manner...

Ah, it was nice to have an afternoon clear from meetings and get some coding done again.  We've been integrating some large code changes recently and the general chaos level was a little higher than usual, but it was still a quite productive day.

Not that meetings aren't productive in their own way, of course... but having a good chunk of time to just work through bugs is always enjoyable.

As a certified insane GTA addict and general game soundtrack enthusiast, I of course have all the GTA soundtracks (and listen to them fairly often).  Since there was no soundtrack ever released for GTA 3, the only way to get the songs was directly off of the disc.  This also had a nice side benefit of getting access to all the little DJ interludes, and the entire contents of the talk radio station.

After poking around a little bit on how to do the same for Liberty City Stories (since there is no soundtrack one can purchase), it occurred to me that I'd never gone through the "complete" soundtrack ripping process on the other PS2 generation GTA games (Vice City and San Andreas).  A few useful GameFAQs and some music software purchases later, I've deconstructed my Vice City disc into a truly "complete" soundtrack edition, with all the radio content in the game as well little bonuses like the intro music.  In the process I learned a great deal about WAV editing and MP3 conversion, something I'd been meaning to play around with more for a while.

It is a little unfortunate that I had to grovel around with crazy editing tools to get this though.  Although the GTA games have generally set a very high bar for soundtrack releases (especially compared to almost any other Western game), there is still a lot of awesome content there with no official way to purchase it.  Compare this to the numerous Final Fantasy soundtracks and remixes... heck, I even have several books of FF sheet music!  I certainly don't feel any ethical remorse, given that I own 3 copies each of Vice City and San Andreas, and 2 copies each of Liberty City Stories and GTA 3 (what can I say, I'm a completionist).

Thanks to the wonder of the iPod, I can now drive around listening to my hand-assembled uber version of the Vice City soundtrack.  A pointless exercise in geekery?  Absolutely, but it still puts a stupid grin on my face...

I found this note, about Valve patching HL2: Ep1, rather interesting.  Obviously patches for games are nothing new, even episodic ones.  But what was intrigued by was this line:

Based on the statistics gathered from people playing through Episode One...

I really love the idea of using real time reported data to make patch modifications.  The core idea is nothing too new; we do tons of data mining on the game results from Age of Mythology and Age of Empires 3 to decide how to make balance adjustments.  But being able to reliably gather that data for difficulty adjustments from a single player game is a powerful evolutionary step.

One of the internal tools that was part of the package mentioned in my posting a week or so back as being part of the shared studio technologies is an application of this concept for use during development.  Instrument your game properly, then let the data from playtesters roll in -- far more efficient and precise than the traditional mechanism of collecting feedback. 

The flip side of this is kind of interesting too -- we have tools to run combat simulations in our RTS games, and tools to run AI games, but generally haven't been able to actually use vast number of AI oriented games to generate high level rolled-up balance data.  I suspect that will be coming in the future, and maybe even some day we will be able to use those kinds of automated tools to proactively gather AI "data" about single player campaign difficulties...

My copy of the Final Fantasy XII import soundtrack arrived today, along with a huge pile of game soundtracks I'd been meaning to buy for a while... Metroid, Parasite Eve, Katamari, etc.  Getting the upcoming FF soundtrack on import, then listening the heck out of it while waiting for the actual game to arrive on US shores has become a bit of a ritual for me.  The music and mood of those games is such a huge contributor to the overall experience, and instead of "pre-listening" ruining it, I find it draws me in much more. 

Yeah, you know you're a huge videogame geek when you not only buy game soundtracks, you import them for games not yet released.  I'm at peace with my fate...