GDC Europe: Obsidian's Five Hard Lessons Of RPG Design

In a talk at GDC Europe J.E. Sawyer, project director for Obsidian Entertainment (Fallout New Vegas) discussed the "challenges that the RPG industry has faced in adapting from its pen and paper roots." His concerns were "ways in which we've succeeded" and "ways in which we fail when we don't take player experience into account." In his view, "RPG developers have often misstepped in understanding why things are not fun for people." Often, players and reviewers will say that a gameplay mechanic is "pretty good for an RPG. This is a backhanded compliment." Describing it as frustrating, he is perhaps more frustrated with developers than players: "developers think it's okay to have crummy mechanics," he said. "I think that if you are going to have the player engage in something that's not conversation and story, it should be fun, and I get really frustrated by this." "We repeat errors over and over again... Because it's a part of what an RPG is," he said. "We ignore established mechanics from other genres... We don't look and see how they did it." He also pointed out that frequently, games offer gameplay "that really makes players do degenerate tactics... They're working around our dumb design" to get the results they want, such as repeatedly reloading to pass random checks. Another problem is that "we listen to the vocal players, who in many cases are wrong-headed." "Fundamentally, we don't consider the player's experience to be more important than the ideas we've had or the expectations" for the genre. Instead, he says, "We have to think about how a person engages the story and goes through it." Five Hard Lessons Sawyer outlined five hard lessons that he's learned over the years: Mechanical chaos is frustrating. RPGs often rely on random number generators, "in part because that is the only way to simulate things in a tabletop environment." However, he said, "In some cases, where you can reload, mechanical chaos is pointless." It also can be frustrating either way. What you can perceive is the most important thing. Games "often focus on statistics, but we often can't perceive the effects in games." Small stat upgrades don't mean anything to players at all when they can't see the effect. Conversely, he said he's "implemented broken things in games but players don't notice it," because there's no external statistic reflection. Strategic failures are the biggest disappointing failures for players. When building a character or a party, "you're making long-term decisions," said Sawyer, "but many RPGs effectively punish you for making bad choices." The idea of player vs. character is a false dichotomy. Developers with a traditional tabletop background expect players to be roleplaying when they play games. However, he said, "it will be the player doing the action... ultimately games are about the players trying to accomplish a goal." There is a definite question of "how much are we asking the player, and how much are we asking the character." Good gameplay is better than whatever your ideas or whatever the player's expectations are. Simple and understandable: don't follow genre conventions simply because they exist. Beyond that, "attempting to execute something because you think it's a good idea or players insist it's a good idea doesn't always result in something good." Breaking It Down 1.) Mechanical chaos -- "randomization as a means to resolve a gameplay conflict" -- is "very frustrating to players," said Sawyer. Contemporary games which offer FPS-like interfaces still rely on randomized accuracy, which drives players nuts. His own company's Alpha Protocol is one example of this. "No actual human being likes this! You really struggle to get to the point of competence in the game," said Sawyer. When it comes to randomized lockpicking/hacking/speech/crafting etc., "All it causes is this: 'Yay! I'm gonna reload the game!' There's nothing to prevent me from reloading. Any of these checks where there's something important on the line... It just results in degenerate gameplay behavior." In short, with this sort of gameplay, gamers have bad experiences "not because they did anything wrong, but the game capriciously decided you fail." Mass Effect changed its combat from 1 to 2: "Most of the weapons feel a lot better, and what they did was make it feel like a more traditional shooter in many ways," said Sawyer. 
 2.) What you perceive in a game is ultimately what matters the most -- Mass Effect had tons of weapons but they were barely differentiated. They had incremental stat differences only. "What's the chance that a 5 percent difference is going to make you take the enemy down in one fewer hit? If it takes me four shots, but the fourth shot killed him a little more, that doesn't mean anything to me," Sawyer said. In Alpha Protocol, "The player could get abilities to upgrade their stealth but often they couldn't see the effects in the game," he said. It was widely considered to have a broken stealth system. "It was a cool idea but certain aspects of it didn't feel good because it didn't feedback to the player." In Dungeon Siege III, Obsidian changed the game so that the AI-controlled companion characters do 25 percent damage and take 25 percent damage. It's an improvement over the characters in Fallout New Vegas because "they're still there and doing something, but you don't have to babysit them and the player still feels like they're doing something." "Players did not react negatively to it," he said. 3.) Strategic failures feel really bad -- In an extreme example, he mentioned that The Bard's Tale, a 1980s classic, required you to have a bard in your party to progress past a certain point -- something that was not telegraphed by anything but the game's title. More relevantly, Icewind Dale and Temple of Elemental Evil required the player to create entire parties at the adventure's outset. "The games were tuned for D&D veterans. There are tons of ways you can make strategic errors. There are tons of ways you can make bad parties. What happens is 20 to 30 hours into the game, you can't go any further." "Yes, the player made the error but we placed a high demand on them," Sawyer said. In Fallout 1 and 3, specializing in "big guns" was not that useful, as there were few such weapons and they didn't show up early in the game -- neither of which the player could know at the point of character creation. "In Fallout New Vegas, we got rid of the big guns skill and pushed those guns into other gun categories." "We kept the idea, we wanted the experience, but we didn't want them to have to deal with the weird system," he said. "I don't see a compelling reason to not" let players re-spec characters that aren't suited to the gameplay design in an RPG, he also added. 4.) Player vs character is a false dichotomy -- "In every game you are expected as a player to use the resources available to you. A player very consciously makes decisions on how to build their character, so really it's about what do you ask the player to do over the course of a game," he said. "You have to be cognizant of what you're demanding of people." Some games expect the player to manage too many options at once, and often developers argue that this is "dumbing down" the game to reduce them. However, he said, "This isn't about whether an RPG gamer can play twitch gameplay, it's about if a player is asked to manage a lot of stuff you shouldn't ask them to." "Mental awareness and their ability to engage what's in the game," is something developers need to better pay attention to, said Sawyer. 5.) Summing up the importance of strong gameplay, he said, "The idea is that good gameplay ultimately is what you want to create, because that will produce good experiences for people. If you create a game with bad gameplay that was the result of your idea, and it met player expectations, it's still bad and frustrating." Never create gameplay mechanics simply because that's "just the way that RPGs are," he cautioned. "If we ignore the lessons that those games [in other genres] teach us then we're really limiting our audience's ability to have fun," he concluded.