Double elimination tournaments

Let me start by saying that I really like the concept of a double elimination tournament. So I might be biased in my analysis.

Second, this post deals specifically with potential use of double elimination tournaments in settings of EGC/LSG side event.

What is wrong with the current approach

Current approach has two separate parts — qualification and finals. Qualification consists of several groups playing round robin system (8 groups of 6 people). First two of each group get to the finals, which is a simple single elimination (16 players in our study case). The total number of rounds is 5 + 4.

There are several related problems which stem from this concept. First, not all games are important. Some people leave halfway during the eliminations or just decide to resign the remaining games, as they have no chance of advancing to the finals anymore. This can influence who advances to the finals and has a negative impact on the tournament atmosphere. Second, the final only determines the winner reliably. Plus it just sucks that by blundering once, you get eliminated.

Why double elimination

Double elimination eliminates the unnecessary games. Every game matters. If someone decides not to participate anymore, his opponent gets a free win, but it doesn’t harm anyone but the one who quit.

You are free to lose any single game and can still win the tournament.

Double elimination is also much more accurate in determining the second to fourth places, which are available without any extra playoffs, with single playoff necessary to determine fifth and sixth place.

So where’s the catch?

Perhaps the biggest disadvantage is that double elimination of (up to) 64 players takes 12 rounds. That is considerably more than 9, and so the necessary extra time needs to be reserved. The reward is elimination of redundant games and much fairer results.

Another disadvantage might be that quarter of the participants only get to play two games. On the other hand, they can go to the beach and have fun instead of having to play.


Double elimination is particularly suitable for faster tournaments, where you can finish a round in under half an hour (and the whole 64-player tournament under 6 hours). The slightly asynchronous nature of double elimination allows for certain brackets to develop faster. On the other hand, there can appear bottlenecks when someone doesn’t show up. This can be taken care of by giving a default win to bottlenecker’s opponent, hence speeding up the tournament even more.

I would like to try double elimination for 9×9, 13×13, and blitz tournaments at LSG 2010. I’m all ears for your opinions on this idea.

By the way, have you already registered for LSG 2010?

7 thoughts on “Double elimination tournaments

  1. OK, I see how it works with players of equal strength. But what about players with strength in some range? In theory it doesn’t matter how we arrange games in one round (so we can pair players by str in every bracket), but question is for example how long “weakest” player can avoid game with someone from top half? :) Do you know something about it?

  2. Good point. Perhaps I wasn’t clear — I wanted to apply it in handicap tournaments, where rank difference is effectively elliminated (or at least greatly reduced). I don’t think double elimination is suitable for go tournaments with even games and high skill spread.

  3. Well… for handicap games it looks like a great idea.
    I like one thing about it – it somehow removes one problem with sandbaggers: influence of them on other players. Still it’s a problem if sanbugger wins all games – maybe it’s possible to create macmahon-like version of double elimination with starting points or something like topgroup at least?

  4. I think that double elimination tournaments aren’t good to go.

    The general idea behind the double elimination is that you play lower number of games totally in the whole tournament compared to other systems. That;s why it is a good system in a competitions where the playing field is a rare resource and you can’t play many games at once (like table-tennis, volleyball, table football).
    In go you can play many games at the same time so rather minimizing the amount of rounds is a better idea.

    Of course all your arguments apply, and when the tournament is very quick, no-serious and people in general tend to quit their interest in the middle of the tournament it might be a good choice.

Comments are closed.