Gooseberry Rating System version 1.0

My proposed rating system is a way to provide Stellar Crisis players with a numerical rating which represents how good they are at the game. While some rating systems (e.g. the Bridier system) work very well for grudges, my system works very well cutthroat games, blood games, and alliance games. It also is savvy to subtle issues such as ruins, 2 vs. 1 games, surrenders, etc.

OVERVIEW

Here's how it works, in a nutshell:

For starters, let me explain the secret to conceptualizing how ratings work. Think of rating points like money, and SC players are gamblers who wager on games of SC. When you win a game, you take the points away from your opponents; you get richer, they get poorer. However, like all shrewd gamblers, odds are demanded by the underdogs whenever they challenge stronger opponents.

The rules for ratings points demand that money is not created out of thin air, or evaporated. It only changes hands. (The one exception is when new accounts are created.)

At the start of a game, the system looks at the player's ratings, and figures out who the weak and strong players are in the group. Each player then must wager a certain amount of rating points into two pools: the win-pool, and the nuke-pool. Imagine that in the beginning of the game, the players drop their money into two jars. The amount of points wagered is calculated by a forumula (described below), and it ensures that the top-rated players stake the most points.

Every time somebody gets nuked, the nuker gets a slice of the nuke-pool. (He reaches his grubby hand into the jar and grabs a wad o' cash.) By the end of the game, the nuke pool will be much smaller than it was at the start.

At the end of the game, the leftover remainder of of the nuke-pool gets added into the win-pool, and the winners share in the pool equally. The winning players, generally, get back what they wagered plus a fraction of the points lost by the losing players.

GOOSESCORE FORMALLY EXPLAINED:

  1. Definitions
    M = Median value of ratings in the game
    K = Arbitrary value assigned to represent the volatility of ratings
    P = The number of players in the game
    N = The "nuke pool" (points set aside for nukes)
    W = The "win pool" (total number of points for the winners)
    B = The "bounty" (how much you get for a nuke)
  2. Starting Ratings
    Each player has a rating which starts at 1000. These are floating-point numbers, although for display purposes (and bragging rights) we drop all fractions.
  3. Initial Calculations for a Game
    1. When a game is conducted, the system calculates the median rating (M) of the players in the game. E.g., if the players are rated 500, 550, 1500, 2000, and 2200 the median rating is 1500. If there is an even number of players we average the two middle ratings.
    2. Each player risks a certain number of rating points to be divided between nuke-pool (N) and the win-pool (W). The amount contributed to the pools, however, is not the same for all players:
      1. A value of K is assigned to the rating system to reflect the degree with which ratings change during games. I suggest a K value of 16.
      2. Each player contributes a different amount of points to the two pools, depending on their rating as compared to the median rating.
      3. If the player's rating is equal to the median rating, the contribution is K to each pool.
      4. If a player's rating R is X points lower than the median rating, the contribution is K - X/2K, although can never be less than 2.
      5. If a player's rating R is X points higher than the median rating, the contribution is K + X/2K, although can never be higher than 2K.
    3. A bounty (B) is set for all nukes during the game. This is simply N/P or the nuke-pool divided by the number of players.
  4. Nukes
    If you nuke a player, your rating increases by B on the spot. N then is decreased to N-B to reflect your withdrawal from that pool.
  5. Ruins & Surrender
    If a player goes into ruins, or if he/she surrenders, the value of N is decreased by B, and the value of W is increased by B. In other words, the points set aside for his nuke are no longer available except to the winners of the game. [Note: I don't agree with the "taking over the remains of ___'s empire" modification, so I don't feel that it should be given any special bonus other than a nice 50/50 builder.]
  6. Game Completion
    Upon completion of the game, all points in the remaining nuke-pool is transfered to the win-pool, and all surviving players divide the win-pool evenly among themselves.

    COMMENT

    This might sound complicated, but it would be fairly simple to program. Keep in mind: you don't need to understand the inner machinery of the system to get a feel for how it works. In general, when you play whimps you stake lots of your points for only a few of theirs. When you play stronger players, you have more to gain. All nukes are equal, however, otherwise everybody would gangbang the best players to get their rewarding nuke-points, which I don't think is fair.

    This system gives great benefits for the players who enter blood games and come out on top--and why shouldn't it? It also highly penalizes gang-banging, as the win pool must be split so many different ways.

    Hopefully, the system will match up the beginners with their ilk and give the good players a forum to only play other good players. The raw count of wins, as you know, is a meaningless indicator of skill, as is the win/loss ratio. Go to any chess server and you can see: the best players (highest ratings) often have win/loss ratios under 50%!! Why? Because they like to play people better than they are--and that's how they got to be so good. The people who always prey on the newbies find that after a while, they can only gain a point or two per game and stand to lose a landslide if they slip up.

    IDEAS FOR CHANGES:

    1. The K value of 16 is abitrary. It could be higher or lower depending on taste. The higher it is, the quicker a good player can achieve a respectable rating. However, the higher it is, the more a slump can crush a good player's rating.
    2. To avoid the problem of old players with new accounts running amok, as well as the problem of new players taking many many games to establish their true rating, I propose that a certain number of games goes into making a player's preliminary (or provisional) rating. This system might work like this:
      If you have less than 10 wins, you have a provisional rating and not a real rating. You are technically not a rated player yet.

      When you play, you stand to gain or lose points with a 2K value; that is, your "swings" will be more pronounced than with an average player. However, rated players participating cannot gain or lose points from you. I.e., if they nuke you, you lose the points, but they do not gain points. If you are nuked, your points are then removed from the win pool (or perhaps never go in there to begin with?).

    3. Since I enjoy the strategy that lies behind ruins so much, I would love to see the Ruins rule changed to read "If a player goes into ruins they are returned the rating points that they risked to the nuke pool" instead of merely transfering it to the W-pool. This would make ruins something to strive for, as it's not as bad as totally getting blown away. This is highly contraversial so I include it only here, as an aside.

    Comments? gooseberry@2020tech.com
    HONK HONK!
    Gooseberry 10-28-99