TruSkill: Rumored To Be Working

[admin_old]

There were several things about the 360 launch that had me excited but never really panned out. One of the most exciting things was TruSkill. Basically, it was supposed to be Halo 2’s matchmaking, but for every game on the 360. The problem is that I have never seen any evidence of it working, especially [...]

More... (http://www.8bitbrigade.com/truskill-rumored-to-be-working)

[Slitt Vicious]

Would someone be willing to post this article from OXM?

[Duster72]

Gears of War, Saints Row, Rockstar Games Presents Table Tennis — if you’ve played any of these multiplayer online games on Xbox Live, you’ve earned yourself a TruSkill ranking. The TruSkill system is designed to give you a fair match when you play ranked matches against other gamers on Live; it makes sure that you’re playing against people or teams of a similar skill level by taking notice of just how good you are in the matches that you play. It’s visible in games like Ghost Recon Advanced Warfighter 2, where your TruSkill ranking is represented by actual named ranks, but in most of the other games, it’s something you never see. And because you never see it, a lot of gamers assume the word “TruSkill” is just marketing speak, up there with “blast processing” and “64-bit Jaguar.” So…whatever TruSkill is, at least we know it really exists. But does it really work?

Every brutal slaying in Gears of War affects someone's TruSkill ranking.

“The strategy for the Xbox 360 launch was to build a multifaceted matchmaking solution that addressed both the social and competitive aspects of online gaming,” explains David Shaw, Program Manager for Gaming and Community with Xbox Live. “To that goal we ended up focusing on two parallel solutions: a social matchmaking solution — player match — which based its matchmaking on how players interacted with each other in gaming sessions over time. The other — ranked match — was a competitive matchmaking solution that centered on accurately identifying a user’s skill level in a particular game quickly, and then getting individuals of similar skills into game sessions.”

The ranking system was developed by the Microsoft Research team in Cambridge in the UK. They based it on a system developed for working out the skill levels of chess players called the Elo rating system, though Shaw points out that it was necessary to make a few changes since Elo is “only accurate for determining ranking for two-player games.” One of the other problems is simply that player rankings in the Elo rating system aren’t regarded as valid until they’ve played more than a certain number of games — say, 20.

So, TruSkill is what’s known as a “Bayesian based skill rating system” — it works on probability, basically. Or, as Ralf Herbrich and Thore Graepel point out in their technical report on the system, it “assumes a factorizing Gaussian prior distribution” and uses “an online learning scheme referred to as Gaussian density filtering.” Well, duh.

The system, Shaw explains, is “baked into” Xbox Live statistics, session, and matchmaking application programming interfaces — basically, a library of code that tells Live how to operate at particular times. This makes it very easy to implement for any developers who want to. They don’t actually have to worry about coding it into the game; they just have to ensure that any games labeled as “ranked” post results at the end of each match. Live maintains all the leaderboards, works out what to do with the results, and then makes all kinds of magic calculations to figure out whether you should have leveled up in your ranking or not. “The 365 algorithm has already been developed,” notes Saints Row multiplayer designer Anoop Shekar, “so all you have to do is implement it.”

Saints Row

Saints Row uses TruSkill to calculate individual rankings in 12-player bloodbaths.

Not that it’s forced on developers. “TruSkill is an integrated part of our matchmaking system that is available for games to create better experiences,” explains Shaw. “The depth to which titles take the technology is up to them.” Adds Shekar: “It’s probably one of the fairest systems to use for ranking people. As it’s a modified version of the chess ranking system, it has stood the test of time. Once we learned of the system we were pretty much on board with it. I don’t remember considering another method.”

The results are simple enough, and they benefit “those who want to get into a trusted game with people that will give them a run for their money, but not blow their doors off,” Shaw explains to us.

Is there room for improvement in TruSkill? “The only thing that comes to mind is that the magnitude of a victory doesn’t matter — a win’s a win,” says Shekar thoughtfully. “That’s a legacy from the Elo system, and makes sense for chess since there’s no way to fairly determine how badly someone was beaten in a game of chess. But in a deathmatch, you can!”

Fortunately, it’s not a closed project for Microsoft — Shaw tells us that the research team in Cambridge is still hard at work on it, thinking up ways to make it even easier for developers to use the system, and “continuing to assess how TruSkill and ranked matches can be used.” Cryptic, but such is the intricate nature of algorithmic research.

Until then, Shaw defi nitely feels the current system works. “Because Xbox Live and multiplayer gaming are such staples of the Xbox 360 experience, we wanted to make sure our matchmaking system was fair and accurate, and based on the more than 2 billion hours our gamers have spent on Xbox Live to date, we are confi dent that TruSkill is the best skill-system for the console. It has been battle tested by our Xbox Live community and by game developers and the response has been positive.” Shekar agrees: “Overall, the TruSkill system lets gamers play with the right people so they’re not winning or losing too easily. Everyone gets the right level of challenge.”