Update to Double Elimination Tournament

The current implementation of double elimination tournaments can be summarized as follows:

Note that in the “final”, if the player from the the first tier loses, they will be eliminated, even though they have lost only one game over the whole tournament.

Proposed change: replay the final game in such a situation, so that the tournament follows the concept of “a player is eliminated if and only if they have lost twice”.

This has been discussed and proposed as a change several times in the forums before:

A possible reason against adopting this change earlier may have been related to concerns about implementing and dealing with the possibility of replaying the final. However, recent changes to properly accommodate jigo have now already introduced the possibility of replaying the final game, so the feasibility of making this change seems to have been demonstrated.

Please share your thoughts, concerns, objections, etc. regarding this proposal.

  • No Grand Final on Loser Bracket Win (current behavior)
  • Grand Final on Loser Bracket Win (proposed behavior)
0 voters

I basically agree with the proposal for the reasons given. I just feel it’s important to set out an alternative rationale.
To the extent that Go is a non-random full information game then it’s reasonable to expect that if two players play then even if there is a relatively small difference in strength the same player should generally win. Therefore if two players have played and there is a result then it’s a bit pointless to make them play again since we can expect the result to be the same.
So the extra final is pointless and if it didn’t happen the winner of the tournament would be the same in neatly all cases.

Since tournaments are there to determine the correct winner, the extra game doesn’t help in this aim.

I realise that this doesn’t apply to handicap tournaments but really those are just elaborate and effortful coin tossing substitutes!


FTR, I think this hits the nail in the head for why the current behaviour works as it does. In fact, switching to the proposed behaviour involves deleting the special case for whether there are just two players left.

I just had the idea of checking Wikipedia’s entry on double elimination tournaments. It seems from the article that it’s reasonable to organize either with or without the grand final, but it seems more typical to use the proposed behaviour. From Conducting the tournament:

The final round of a double-elimination tournament is usually set up to be a possible two games,[2] with the second referred to as the “if game”.


Note that it is possible for the players in the final to have played together earlier in the tournament.

The proposed implementation turns such cases into like a “best of out three” situation, while the current implementation ends the tournament after these two players are tied 1-1 head-to-head.


Here’s another thought: The finalist from the second tier (“Losers bracket”) had to win more games to get to the finale than the winner of the first tier.

For example, in a tournament with 64 players and no draws, the second tier winner plays 3 more games than the first tier winner (one of them being a loss).

One could argue, if this player survived that many games and they win in the first finale, then they should be awarded the win without a second finale, to compensate for the difference in number of games.


I just did some searching the web, to see if I can find any examples (besides OGS) of doing double-elimination without an “if-game”.

These all describe an if-game:

This describes the if-game as optional, depending on format, but doesn’t give examples of tournaments skipping it:

I didn’t find any sources that described an example of a double-elimination tournament without an if-game. (To be clear, this was just a cursory, non-exhaustive search…)

It kind feels to me like OGS is standing alone here. (FWIW, I had voted against the proposal early on, but reading Wikipedia’s entry changed my mind and I switched my vote.)

I’m curious if anyone, especially perhaps one of those that disagrees with the proposal, is aware of other tournament venues (besides OGS) that skip the if-game?

This is true; the asymmetry is inherent to double-elimination tournaments, though. Skipping the if-game doesn’t remove the asymmetry, and it also seems like skipping the if-game might be unfair compensation. Also, OGS has lots of tournament formats that don’t have this asymmetry, for tournament organizers that want to avoid it.

Are there other reasons people would like to avoid the if-game?


The vote seems pretty strongly in favour of the change…

@anoek, do you have any concerns?

@_KoBa and @PaladinGo, since you both voted against (but didn’t comment), I’m curious if either of you has some other concern not mentioned above that you’d like to share?


For example, in a tournament with 64 players and no draws, the second tier
winner plays 3 more games than the first tier winner (one of them being a loss).

Is that just for the
increase the number of rounds until there are 2 players left, by having a significant fraction
of the rounds ignore a significant fraction of the players who are still in the tournament

I get that, for 64 players and no draws run in the obvious way,

there are at most 2 byes total in the entire tournament
the 2-bye cases are ​ ​ ​ 1 bye was already given, there are
exactly 3 players left, and each of those 3 already lost a game


(In particular, one could have the rule that in those cases, the second bye does not
go to the player who got the first bye, so that no one will get more than one bye.)

Not really a consern even, just my gut-feeling saying that it might not be very fun to get paired against the same person for two consecutive games. Espesially if the first game is lopsided, having a rematch immediately after resignation can feel pretty exhausting.

But yeah i also think it would make more sense to keep the tournament running until all but one person has lost 2 games ^^


I assume anulled wins still count towards the tournament, right? If so, could a player who lost the final from upper bracket, and feels confident that they have no chance in the grand final, just resign move 1, which would not impact ratings, but would effectively resign the tournament? Of course, it would be great if they were up for playing the final game, even if only to learn from a stronger player, but if not, I can’t think of how resigning move 1 in this specific situation would negatively impact other players


Note that the 1st-tier player has “transitively beaten” the 2nd-tier player earlier in the tournament (either beat them directly, or beat another opponent who had already beaten them, or some longer chain…). As such, if the 2nd-tier player wins the first final, then the two of them are fairly likely to be well-matched.


Sorry but I don’t understand your question. What do you mean by


Ah yeah thats also very true ^^

I dont play on live/blitz tournaments very often so im just speculating here ^^

For corr tournaments it should be ok to have another final game if the two finalists are both sitting at one loss. It would make the tourney last bit longer, but thats pretty much the point with corr tourneys xD



I mean, for what sort of pairings do you get that
in a tournament with 64 players and no draws, the second tier winner
plays 3 more games than the first tier winner (one of them being a loss)

I get

 after this     the winners bracket      the losers bracket
many rounds    has this many players    has this many players

     0                  64                        0
     1                  32                       32
     2                  16                       32
     3                   8                       24
     4                   4                       16
     5                   2                       10
     6                   1                        6
     7                   1                        3
     8                0 or 1                   1 or 3

, ​ for which
the second tier winner plays 3 more games than the first tier winner
can’t happen, because the tournament has at most 2 byes total.

(I am guessing that the one you are thinking of is one where a significant fraction of the rounds
have a significant fraction of the players who are still in the tournament not play that round.)

1 Like

As per my understanding, I agree with your table except for row 8. By this description

…after round 7 the condition

… is not met, and so I believe that there are two byes in round 8 - one for the first tier winner, and one for a player from the second tier. So the table would continue like this:

 5                   2                       10
 6                   1                        6
 7                   1                        3
 8                   1                        2
 9                   1                        1

… and then the finale takes place in the tenth round.

Ah, so not as many byes as I thought you were expecting.

However, by omitting those two byes, one gets true double-elimination in
at most the number of rounds that your table would get by allowing the
winners-bracket player to lose the tournament with a single game-loss.

(Omitting those two byes makes it so that row 9 will be ​ 1,0 ​ or ​ 0,2 .)

1 Like

Your solution is tidy, and does take fewer rounds.

However, it doesn’t seem to fit the definition of a double-elimination tournament. All the descriptions I found keep the two tiers separate, until the final 1-2 game series between the winners of each tier.

There is asymmetry in double-elimination, for better or worse.


I didn’t vote, but I can imagine people being happier in theory with double elimination having the conditional extra final, at least from some previous threads and discussion.

In practice who knows, you can get fairly one sided sitewide tournaments like a 1 kyu with everyone else a ddk. Sometimes you just get people timing out or forgetting about rounds, and all sorts of funny things in practice.

For instance the first game the winner of the winner bracket played was in the final here, the rest the opponents just timed out on move 1:

So that could be true most cases say, or it could be that you won because a person you played dropped out or timed out (maybe they had a bye and forgot when the next round starts) and you didn’t actually beat someone who beat the person coming up from the loser bracket.

So there’s still some chance for weird mismatches.

Not an argument against per se.

I can agree that if double elimination is to mean you lose twice to be knocked out then sure there should be a second final.

If the “double elimination” was just seen as a way to pick finalists, like doing a Swiss or round robin followed by a final game between the top two players then one wouldn’t need it - though maybe another name would be better suited to it.


The vote was pretty strongly in favour, and the resistance wasn’t strongly-held, so I landed this over the weekend. It was just now deployed, so any double-elimination tournaments that hit that scenario going forward should have an if-game.

Here’s a screenshot of how it looks (from my testing):