MIT Tennis Graph

The MIT Tennis Ladder is an unofficial tennis ladder maintained by Mika Latva-Kokko. To see the current ladder, read the rules, and join go to the MIT Tennis Ladder webpage.

This graph summarizes the results from the MIT Tennis Ladder webpage. Each arc indicates the result of a tennis match. The length of an arc shows the margin of victory. We found this graph useful when deciding who to challenge at our level. Hint: Use the arrow keys rather than the mouse to pan over the graph.

Any more questions? Have a look at the FAQ.

This graph is auto-generated every day. It was last updated Wed Nov 2 18:00:08 EST 2005.

Graph

Frequently Asked Questions

How does this help me?
Is there anything wrong with the current ladder?
Why is my rank so low?
How did you guys do it?
Are the edge lengths (rank differences) really the margin of victory?
Why do you guys suck so much?
Why are forfeits not shown?
Why am I not shown?
Is there any room for improvement?
How do I report errors/questions?

  1. How does this help me?

    It is based purely on the games played. Not on your self-assessment or forfeits. Here you can compare yourself and find people on the same level as yourself. 6-0/6-0 matches are rarely as entertaining as more equal matches - this illustration will help you seek out the latter.

  2. Is there anything wrong with the current ladder?

    Mika is a cool guy, doing cool stuff. There is nothing wrong with the tennis ladder per se. We did this because we wanted more equal games and to that end we wanted a system that would stabilize faster. Having too low or too high a position on the ladder can easily happen and is not fun because either you have to play people far below your level or you get beaten severely in every single game.

    There are some quite common scenarios where a beginner can climb very fast and skilled players can get stuck in a low spot. In the current ladder the order in which matches are played has an enormous impact. We feel that since this is a very short "season", the order should only have an impact when we need to break ties (in favor of the most recent match, of course). The graph presented here retains history and will therefore look the same regardless of the order in which the matches are played.

  3. Why is my rank so low?

    Because you have not played challenging people

  4. How did you guys do it?

    Perl, dot, topological sort, Simplex, Rank Assignment Problem, Margin of Victory Weights, also tried Google PageRank, which did not work as intended.

  5. Are the edge lengths (rank differences) really the margin of victory?

    No, but they are /minimum/ the margin of victory scalar.

  6. Why do you guys suck so much?

    Because we spend our time in front of computers rather than on the tennis court.

  7. Why are forfeits not shown?

    We feel they skew da shit - maybe we will provide both graphs later.

  8. Why am I not shown?

    You haven't played any games yet!

  9. Is there any room for improvement?

    Yes, very much so. Right now the graph suffers from some dot idiosyncrasies. The edge lengths are the result of a Simplex pass attempting to minimize edge lengths subject to the minimum rank difference, which is obtained as a linear function on the margin of victory fraction. It would be reasonable to consider other formulations of the optimization problem to, say, maximize predictability of future outcomes, or to choose between a conservative, an optimistic or an average-case player assessment in the face of incomplete information. Right now, all you have to do to get a high/low rank is to play a high/low ranked player and loose/win.

    Second, the graph could be collapsed to a total order. This would also require something a tad more clever than relying on dot's internal Simplex formulation. Every node could conceivably have a confidence index computed as the fixed point of some connectedness function over the entire graph. This index could be used to break ties in favor of the highest confidence index.

    Just a few loose ideas...let us know if you want a more structured briefing. Or if you have any insights.

  10. How do I report errors/questions?

    Write an email to tennisgraph [at] stefansen [dot] dk


Christian Stefansen, Peter Pietzuch August 2005