Back to 2012 Tournaments 2012 NSSCContentsOverviewIn 2012, for the first time in the history of the event, teams competing in the National School SCRABBLE Championship (NSSC) will be seeded according to the past performance of their players at School SCRABBLE events and at events rated by the North American SCRABBLE Players Association (NASPA). Seeding means that pairings will be determined in the first round so as to avoid pairing two high-rated teams with each other, to reduce the chance that a strong team will be effectively eliminated from reaching the finals due to a bad first-round draw. Past performance of players is expressed as a rating, using a new rating system developed and implemented for this event by John Chew. For more information about the rating system, see Technical Details, below. Rated EventsThe following past events (listed in chronological order) have been used to compute ratings, in addition to all NASPA-rated tournaments since May 1, 2009. If you know of any other event that has taken place since the 2011 NSSC, and for which full rating information (pairings and game scores) is available, please email it to us as soon as possible so that we can try to add it to the system.
We are aware of the following additional events that may be added to the system, but have (as of April 11) not yet received complete information about the events from their organizers.
2012 NSSC SeedingAs explained below, the ratings used include a measurable amount of uncertainty, which depends on how active the players have been, and is shown after a ± sign. In the first round, corresponding teams in the top and bottom halves of the field will be paired with each other, avoiding pairing teams from the same school.
Technical DetailsSince school tournaments are still comparatively rare, the rating system chosen is based on the well-known “Glicko” system developed by Mark Glickman and used by the Australian Chess Federation, which continuously adjusts the volatility (multiplier) of a player's rating according to the number of games that they have played, and the volatility of their opponents' ratings. Player ratings have two parts: the rating value itself and a statistical deviation. For example, in one version of the rating list, Mack Meller was rated 1717±40, meaning that there was a ±40 point uncertainty in the value of his rating. The Glicko system includes a third term, which we have omitted for the sake of simplicity; it also increases the volatility with inactivity, but we have omitted that too because of the time limit on a player's school career. The Glicko system was modified by Taral Seierstrad to be more suitable for SCRABBLE tournaments, by scoring players not just on wins and losses but also based on the point spread of each game. So for example, winning by 200 points has twice the effect on your rating as winning by 100 points. We call this modified system the “Glixo” system. Finally, we developed the “Team Glixo” system where each team can have one or two players, and a team's effective rating against its opposition is the greater of (100 less than its highest member rating) and (its average member rating), and a team's effective deviation is the square root of the sum of the squares of its member deviations. Each team member's rating is updated individually against the same record of opponents. Here is sample Perl code for updating one player rating. $residualSum = 0; $weightedAchievementSum = 0; $b = 5; $tau = 90; $playerMu = $player->OldRating(); $playerSigma = $player->OldDeviation(); for $round ($first_round..$last_round) { $opponent = $players->{opponents}[$round] or next; $oppMu = $opponent->OldRating(); $oppSigma = $opponent->OldDeviation(); $spread = $player->Scores()[$round] - $opponent->Scores()[$round]; $achievement = $oppMu + $this->{'b'} * $spread; $rho = $b * $b * $tau * $tau + $oppSigma * $oppSigma; $residualSum += 1 / $rho; $weightedAchievementSum += $achievement / $rho; } $oldVariance = $playerSigma * $playerSigma; $varianceResidual = (1 / $oldVariance) + $residualSum; $variance = 1 / $varianceResidual; $newRating = $variance * ($playerMu/$oldVariance + $weightedAchievementSum); $newRating = 1 if $newRating < 1; $newDeviation = sqrt($variance); Because of the age of the participants, cross-tables and other details of the rating system are not currently being published online. If you have questions about the computation of a rating, please email us privately. ContactFor general information about the School SCRABBLE program and the National School SCRABBLE Championship, please email info@scrabble-assoc.com. For technical inquiries about this rating system, or to submit additional data, please email info@scrabbleplayers.org. |