In the multi-player section, each committee members submits a list of ten games and can place an "*" next to five of them. Each "starred" nomination is worth two points, and each un-starred nomination is worth one point. The points are then totaled for each game. The "Top 10 and ties" comprise the list of finalists. In the unlikely event of a tie at the bottom of the ranking giving us a short list of more than fifteen games, the tie would be resolved by ignoring the "stars". We have never needed to use this procedure, but it is there if needed.
The 2-player category works in the same fashion, but because the field is smaller in this category, each committee members submits a list of five games and can place an "*" by two of them. The top five nominations make the list of finalists.
Each member ranks the list of finalists in order of preference.
Each committee member has one vote and a game needs an overall majority of votes cast in order to win. At the first stage of the count only the first preference of each voter is considered. If one game has an absolute majority of these votes, it is the winner. If not, all the games with no votes are eliminated, followed by the game with the fewest votes. The votes of this last game are reallocated using its supporters' preference lists. In all cases the vote goes to the highest game from that person's list that has not yet been eliminated. This process eliminates games one-at-a-time from the bottom of the rankings and reallocating its votes until one game has the required overall majority.
(So what is happening, in effect, is that whenever a voter's current choice is eliminated because it doesn't have enough support, they make a new choice from among those games left in contention. The use of preference lists means that one doesn't have to keep going back to the voters and asking them to pick again.)
When there is a tie for last place at some stage of the voting, the following tie-breakers are applied in this order:
It can be proven mathematically that there is no method which has all the properties that one would ideally prefer. All can lead to anomalous situations. The best one can do is select a method wherein these occur very rarely in practice. STV is one of the best in this regard. It also has the great virtue that you cannot manipulate the system by placing your candidate's principal rival artificially low on your preference list. With STV the winning chances of any candidate on your list are not altered by the order in which you place the ones below it. Neither points-based systems or Condorcet ones (which are the two main alternatives to STV) have this property.