Arena Style tournaments

This is a solvable problem actually. The past couple of tournaments I’ve run have been “Arena Style” aka players are continually paired as they finish.


  • winner might not be as clear (so don’t use for national championships)


  • you don’t sit around waiting for next round
  • play as little or as many games as you like

Players generally enjoyed the Arena tournament, but of course some missed the more predictable structure of having rounds.


That sounds like a good format for people who went to spend as much time as possible playing Go, but I actually like the breaks, chatting, watching casual games, eating and a drink after the last round of the day. I also like to know when the next game is due, to fit in the other stuff. I prefer a few intense games to continual play.


It’s the best for both. Players are able to play as much of as little as they wish. There is a playing area and a hang out area. Some folks only play two games the whole time.

It isn’t continual play, but rather continual pairing with the option to take as many ‘byes’ as you wish.


  • round 1: you play at the exact same time as everyone else
  • round 2: you finish your game early, take a break, get lunch, review your game with a pro or strong player, then get paired again
  • round 3: you get some coffee, then watch some games, then play your final game of the day.

When you play is up to you.


Some good points there. Might be worth a try, but have not heard of this format round here.

1 Like

How exactly are the pairings done? Do you have some special software for it?

I’ve been curious before whether it would be possible to run an arena-style tournament IRL, but I haven’t figured out the practical details myself yet.

1 Like

It’s relatively unknown, but I hope it can become more popular. Researching into the history of it, apparently it was first done in the US in San Francisco many decades ago. More recently I learned about it while playing in Chicago and they had picked it up from the club in Ann Arbor.

1 Like

A post was merged into an existing topic: Real life go tournaments: how do you experience them?

Thinking of the arena format, there is something else to consider: What is the difference between a go meeting and a tournament?
For example the fact that you select yourself your opponents or the quantity of games you play. I completly understand the friendly side but how relevant is it to be integrated in a global rating?

In the past there have been such IRL go tournaments in the Netherlands. AFAIU it is a bit challenging to come up with a fair performance comparison between players who play only few games and players who play many games in such a competition.
I just found that the go-club in Apeldoorn uses this format for their IRL club competition. They calculate performance like this: 500 + (#games) * (#wins - #losses).
That seems like a decent performance score formula to me for Arena style tournaments. It avoids participants camping at a good winrate but a very small number of games, while it also avoids that participants with a low winrate can get a high performance score just by playing an insane number of games.


Why the term “500+”? Just to avoid negative numbers?

What’s wrong with just “most wins” like lichess?

1 Like

That seems to be the case, yes.

The issue with that is that it perhaps favours players who play a lot a bit too much. For example a player with 11 wins out of 100 games would beat a player who wins 10 games out of 10, while the latter performance is arguably much more impressive.

Interestingly lichess also does a streak score where if you win two games in a row your next consecutive wins are worth double.

Normally wins are worth 2, so the 10 out of 10 player would get

2*2+8*4=36 points

While the 10 out of 100, assuming it was just randomly scattered wins would get

2*10=20 points

1 Like

That does look like an improvement, but still a player with 20 out of 100 will beat a player with 10 out of 10. I suppose it depends on what kind of participant behaviour you want to reward more. How much do you want to reward playing a lot vs playing well.

1 Like

Yeah it was the italics thing :slight_smile:

Probably when starting off small it’s best to reward people playing more than camping on a good number of wins.

For instance I wouldn’t say site participation in sitewide tournaments is the best it could be (the pre scheduled ones), so if OGS hosted an arena style format it’d probably be better to award people points for number of wins, in that at least other players are likely to have an opponent that way :slight_smile: While people who’ve achieved enough points/wins might be happy to stop playing.

Still the

seems pretty cool, and I’m sure there’s some other interesting scoring ideas too

1 Like

I would fear that it changes into a blitz tournament so adding a requirement on time min setting may attenuate this side effect.

I think Arena style tournaments (by design) favour players who can play many games in a short time period.

But if you’d use the Apeldoorn performance formula, you’d be rewarding players who can maintain a good winrate while playing many games. I suppose that when games become faster, the results of players become more random (i.e. deteriorate their performance score), so the optimum time control would be different for different players.

1 Like

To give some intuition about the behaviour of the Apeldoorn Arena Formula, here is a contour plot:


Horizontal axis is #games, vertical axis is #wins.

The curves connect points with an equal AAF score and the color of a curve (contour) marks the associated AAF score.
The straight diagonal green contour is at the neutral AAF score (coinciding with 50% winrate), the more blueish contours are higher AAF scores and the more reddish contours are lower AAF scores.

1 Like

and the grey is because you can’t have more wins than games played :slight_smile: