Friday, July 06, 2012

Yucata Rank Changes

Last Saturday the people running the Yucata board game site completely revamped one of their two ranking systems. They kept the TrueSkill system where each game is rated individually in an Elo style. They massively changed the metagame ranking system which had flaws and was fairly open to abuse.

Before last weekend the way the metagame system worked is you earned or lost metagame points after winning or losing a game. The size of the change was based on the difference between your ranking and your opponent's ranking and completely ignored what game you were playing. In the old system if I played someone at a low rank I'd gain 1 point on a win and lose 18 on a loss. Gain 45 points to advance in rank. Lose 45 points to get demoted. Note that in order to stay even against the beginner ranks I'd need to pull off a 95% win rate. My record in A Few Acres of Snow is absurdly good. I'm 232-14. Which isn't good enough to stay even! I simply couldn't afford to keep playing my best game against low ranked people.

Now, you could say that it would all shake out in the long run. Those low ranked people who beat me would earn 20 points with a win and lose 1 with a loss. So they'd hit the 45 point threshold and get promoted pretty quickly, right? Then I'd need a lower win ratio to maintain parity against them. (The system is set up so you lose 80% of the points the other guy won so everyone should be upwardly mobile as a whole.) The problem is there were other restrictions on getting promoted. You had to win 3 different games, for example. That's not a big deal for anyone who plays a bunch of games. But some people, who didn't care about the metagame system, would just show up and play one game. A Few Acres of Snow in particular had a bunch of very good players who only played that game. So they'd just sit at very low rank and destroy other people's metagame ranking.

On the flip side there were also people who shot to the top of the metagame ranking by playing very few games. If you're really good at an abstract game you could start a new account, bash people in a single high skill game (while playing a few other games to hit the different win requirements) and rocket to the top. Again, this would all shake out in the long run except these people would quit once they hit the top and start another account. The top of the Six leaderboard, for example, was littered with accounts created by a guy named onel.

The consequences of this system for me were I stopped joining invites from anyone with a low ranking. I stopped creating open invitations myself for any game I wasn't _really_ confident in my chances of winning. (A Few Acres of Snow, Rapa Nui, and Stone Age mostly.) I stopped playing games I enjoy but which have large amounts of randomness like Can't Stop and Roll Through The Ages. I did this because I wanted to be at the top of the ranking charts and I had to choose between playing games I liked and games I was good at.

They haven't released a lot of details about the new system (making the case that it's 'fun' to explore) but they have made clear the basic change to the system. You earn ranking points not based on the difference between your rank and your opponent's rank anymore. Instead you earn ranking points towards promotion/demotion based on the difference between your TrueSkill and your opponent's TrueSkill in the specific game played. This has a few consequences:
  • You can't just play one game you're good at and scoop up tons of points. As you win a lot your TrueSkill will end up near the top of the list and you'll earn fewer and fewer ranking points per win. 
  • You can play a game you're bad at without too much fear. You're not risking much in terms of ranking points if you lose and you're due for a big payout if you win. 
  • Playing games where you've hit your TrueSkill equilibrium point is not a great idea for ranking points. If you keep playing other people who are at their equilibrium point you're likely to stay pretty even. (Since the loser only loses 80% of what the winner gains you do still rate to go up slowly.) Unfortunately if you often play against people who are below their equilibrium point you rate to lose ranking points. Conversely if you play against a lot of people above their equilibrium point you rate to gain points.
  • Since TrueSkill starts off at 0 and has almost everyone rate to be above 0 you can't expect to play people above their equilibrium point. You can expect to play people below their equilibrium point, and lots of them. In other words once you've played a game enough to stabilize your rating it's a bad idea to play it in terms of ranking up unless you're also getting better at the game yourself. Or you cherry-pick and only play against other people who have stabilized!
Personally I've played most of the game I like an awful lot. I'm finding it very difficult to actually rank up in the new system because they didn't reset TrueSkill at the same time. I'm debating how much I care about the new ranking system compared to how much I care about my existing TrueSkill values. I'm contemplating just starting a new account. The alternative, I guess, is to learn some of the games I haven't played yet and hope they're good enough to power me to the top before I stabilize in them. I certainly gained an awful lot from my first game of Alchemist earlier today!

1 comment:

Robb said...

Isn't the meta-ranking totally irrelevant, and exists only to encourage you to play new games? If you only want to play games you know, shouldn't you only care about TrueSkill? If you want to just increase levels, shouldn't you be playing "Progress Quest" or something?