Tar-Palantir wrote:
nullstar: Don't you think the scores should be normalised for each poster?
They will be normalized for each voter (assuming they vote for a full 25 -- vote for less, and your pie is smaller by (games_voted_for/25)^2....which reminds me:
About the Scoring:
The scoring mechanism was modified slightly from the one used for the first top 25. This post will explain how the whole mechanism works. For the most part, it is irrelevant to how you vote individually, but some people like to know how everything works. (If any of you read Cryptonomicon, this system should look somewhat familiar.) If it seems overly complicated, don't worry about it -- several people have explained good strategies for ranking (or not) the games on your list:
- People vote for x number of games, 25 in our case, assigning a point value to each entry on that list.
- Each person's total point value is scaled to 1; the point values they assigned are scaled accordingly. In this way, each person can proportionately split the points they assign to particular games however they like, but the sum of those votes becomes the same for each voter.
NOTE 1: In cases where a person voted for fewer than the 25 games, the overall weight of their voting sum is reduced from 1 to (x/25)^2, where x is the number of games they voted for. Thus, someone voting for 20 games only has 64% the voting power of those voting for a full 25. Someone voting for only 12 games gets a value of around 23%. This is one of the two major differences from the scoring mechanism used the first year, which relied on the linear reduction of x/25, rather than the exponential reduction used this year.
NOTE 2: People voting for more than 25 games do not have the overall weight of their vote reduced; however, their vote is still scaled to 1, so they may be themselves possibly reducing the value of the weight they assign to the individual games listed on their ballot; certainly, the average value per individual game will drop.
- Two subscores are generated from the votes above:
- Subscore 1, percentage of voters, is determined by summing the scaled weights of all the voters. (For the 2003 vote, for instance, we had 39 voters, but several did not vote for a full 25, so the scaled number of voters was ~ 36.31) This number is then divided by the number of people voting for each game on the list, resulting in a scaled percentage of how many voters voted for each game. Games that made the list this year typically received votes from 25%-50% of the scaled number of voters.
- Subscore 2, importance to voters, is determined by summing the scaled points people assigned to each game receiving votes and finding the game with the maximum value. <i>That</i> value is then scaled to 1, and the lesser values are all scaled accordingly. This was the second change from last year's scoring and was implemented to keep the subscores between 0 and 1.
- The final score is determined by multiplying 2*Subscore 1 and adding it to Subscore 2, resulting in a value from 0 to 3. This value is then scaled to a 10 point scale (by multiplying the interim value by 10/3.)
- The list of games that received votes is sorted by final score, and the top 25 determined based on that. The point value of the 25th game is used to determine the honorable mention list. Any game that has a score of at least half that of the 25th game is given honorable mention. If/as group consensus develops on top 25 lists, the number of games on the honorable mention list will shrink. If/as consensus is lost on the top 25 list, the number of games receiving honorable mention will grow.