TD Blog 1: Woodworking

My goal as a new tournament director is to add cool features to the typical tournament:

1. Target Points - an estimate of how many points, based on ratings, each player should earn in their group.
2. Top Performers - who outperformed their Target Points total and by how much?
3. Awards for various performances such as Biggest Upset, Most Elegant Checkmate and Best Game.

My first tournament effort, Woodworking, has 24 players in 4 groups. I made a note of everyone's rating the day the tournament started, and I used those rankings and the matchups in each group to estimate the predicted number of points each person should earn.

I enjoy thinking about chess tournaments in terms of expected performance vs. actual performance. In Group 3, we have amd1508 nearing his target score of 7.13 by earning 7 points with two games left to complete. In the same group, GrzegorzL has 4 points compared to a target of 4.52 with three results as yet unkown. Whose performance will end up better? Whose will be best? We'll soon find out.

I also plan to analyze at least one game each round, and post my thoughts plus some interesting positions. An early candidate for Most Elegant Checkmate is a 22 move win by cameLiToe. The graphic below shows the "mate in 1" position, black to move.

I hope everyone is enjoying the tourney, and I will post at least one more blog when Round 1 ends with my awards and analysis, together with a recap of everyone's performance vs. their targets. If you have any suggestions about what blog posts and information could make this tournament more fun for the players, post them in the comment section below.

Good luck at the boards!

Details of the "Target Points" Analysis

While the Chess.com ratings are Glicko, not Elo, we can use the simple Elo comparison formula to estimate the "point percentage" expected between any two players:

          PTS for Player 1 = 1 / ( 1 + 10^( (R2 - R1) / 400)

where R1 and R2 are the Elo ratings of Player 1 and Player 2 respectively. Two main differences between Glicko and Elo is that Elo assumes a normal distribution which in turn underestimates the lower rated player's chances. More modern systems like Glicko use an assumed logistic distribution, and Glicko uses a further "Rating Deviation" term which is large for ratings that are less certain (new players, players who have had long layoffs, and players who ratings have changed dramatically recently, etc). The main difficulty with using Glicko Ratings for this blog is that looking up each player's RD requires 3 clicks and then manual entry. And treating the ratings as Elo is quite accurate enough for our purposes.

Here is a sample table from the spreadsheet I'm using:

The Est Pts (Estimated Points) column is simply the sum of the row multiplied by two, and it sums to 30, which makes sense because there are 30 games per group per round. Looking at my own UnionStationFan row, I will be happy with 4 Pts, and thrilled with 5 or more as it would represent an overachievement based upon the rankings of my group members.