GDC Austin: Wednesday Flash & Business

After all those writer sessions, I was leery of going to another one on Wednesday. In the event that I don't find a programming job at a game company by 2010, I'm seriously going to explore the Indie model.

We decided to hit up some of the Indie summit lectures and the Flash developement ones, and the non-summit sessions began on this day also so these notes contain some technical and business talks. Even though Dave Mark's presentation was technical and related to AI, it was also a very good design lecture--one of the best we've been at thus far.

Premium Flash Games by Daniel Cook

• traditional money in Flash comes from Advergames, working for hire for portals, or collecting ad revenue with Average Revenue Per User (ARPU) of $0.00035
• you could ask for money from the players instead: acquire customers; create value; charge money
• your game is valuable
• there are various payment systems available: Mochi coins, Kongregate; gamersafe
• Purchase Loop: Play Game -> Get Currency -> Buy Stuff -> Play Game
• areas to charge money:
• - Time poor: accelerators; selling currency; utility items
• - Money poor: advertising; dual currency
• - Status: aesthetic items; high visibility
• - More of the same: time gate; content gate; subscriptions
• how to generate value in game?
• most Flash games are low value
• make a hobby, not fast food
• use in-house metrics, not the portal ratings
• need at least 4/5 on the Fun Scale
• need to keep customers in order to retain revenue stream
• - build a brand
• - create a website
• you can actually release the same game multiple times, just change the first 10 mins or plot, or minor tweaks
• game is actually a service
• Facebook, portals, mobile devices are views on the game
• since it's a service you can't think "done!" and move on

Storytelling interview with Erik Wolpaw and Marc Laidlaw (Valve)

• game stories are very much integrated at Valve
• - e.g. the train mechanic was required by level design but was used again in the intro to Half-Life
• a lot of magic happens in level design; you can give all dialog to level designers but that is just to get them excited and imagine what is possible in the world (lore-wise)
• do anything to facilitate staying in the game
• - cutscenes are a no-no
• - no interrupts at all
• - what info is so important that you need to tell me right now?! (they made a joke about interrupting someone having sex to read them a short story)
• the broken panel hint/foreshadowing in Portal was seemingly accidental; it portrayed this "corrupt world" or fascade, but it was mainly to hint at the 2nd half of the game
• on the Bioshock audio tapes: if optional story elements are important enough, then why aren't they in the main part of the game?
• a lot of emotion comes form animators or coders
• storytelling is a team effort

Flash Multiplayer case study with Corey Bridges

• Flash is usually pre-loaded; this leads to large download times
• slow animations, and slower with more stuff on the screen
• Flash handles input slowly, esp. the mouse
• can be a memory hog
• usability challenges: controls, UI, authentication/login
• market challenge: do casual players even want synchronous multiplayer games?

MMO AI by Dave Mark

• goal: make PvE feel like PvP
• how do traditional NPCs make decisions?
• - "aggro" -> really the player is telling the AI when to attack
• - tank, healer, dps paradigm surfaces
• - removes enemy dynamicity and automony
• - to the players, fights become: "read the script... play your part"
• all the agents have the same model, same reaction
• we want a variety of reactions, similar to real life
• we don't need to know why, but differences do exist
• we don't want it to be completely random though
• use a simple formula which constructs weighted randoms to select a reaction
• for groups of NPCs: take a head count (allies vs enemies); judge the perceived strength of the opposition; HP; proximity to home/leader (morale); any other metric
• calculate individually the action of an NPC (fight, retreat, charge, flank, etc.)
• repeat periodically
• the result would be that the players seen one NPC run away; a few mins later they watch another one flee; then 2 more; and then the whole group breaks
• how to do target selection? (to break the tank, healer, dps paradigm)
• use a "Tactical Manager" entity which gives orders to NPCs
• TM might even have goals: "protect x" or "attack y"
• it can be reactive (reposition NPCs if some die), or proactive (charge the PCs)
• you can even have Strategy Manager to give orders to Tactical Managers
• influences maps: store relevant data in underlying grid; periodically update
• you can track PC locations and spawn ambush parties on roads or spawn rarer NPCs in low populated areas -> "strategic disposition"
• influence maps can be used by dynamic quest givers
• - game can keep track of mob locations and key locations or landmarks
• - "groups of orcs are east of X, north of Y" etc.
• you can have a much more "dynamic world" were certain NPC factions will more towards towns, away from towns, towards prey, or away from predators

Browser MMOGs with Nils-Holger Henning, Scott Kinzie and Samuel Loretan

• need to hook players in the first 30s
• need small Flash downloads, fast load times
• can't really go full screen
• more transactions come in through SMS (these games are primarily in Europe)
• more revenue comes in through credit card
• different games for different people