EPTTonight, I’ve implemented and released a superior method for seeding the losers bracket in double elimination tournaments.
Q: What was the problem with the old way of doing it?
A: The Problem:
In The Old Way of doing things, players were placed in the losers bracket in the order you saw them in the winners bracket. This meant players could face-off in the winners bracket, and then again in the losers bracket within 1-2 rounds.
For example:
1.) In round 1 of the winners bracket, Max loses against Player-X.
2.) Max moves to Round 1 of the losers bracket.
3.) Player-X moves to Round 2 of the winners bracket.
4.) Player-X loses his Round 2 match in the winners bracket.
5.) Meanwhile, Max wins against his opponent in round 1 of the losers bracket
6.) Max plays against Player-X again in Round 2 of the losers bracket.
This is the worst-case scenario, where there is only 1 round separating when these two players meet. There are numerous other instances where players could face-off against people twice in only a few rounds.
Q: Why is that unfair?
A: In the most extreme case, let’s imagine there was a player that could consistently beat all of the other players in a tournament except for one, and then that player experiences the scenario described above. If we were to do a Round-Robin or Swiss style tournament, then that player would definitely win. If the goal of the competition is to determine who the best player is, then the winner of this event could be unduly influenced by the random-seeding process.
Q: How did you fix this?
A: It depends on the tournament size. But in short, we ‘flipped’ the player order as they move into the losers bracket.
Here is the detailed, nitty-gritty answer:
There are 5 operations as we bring players from the winners bracket who have lost, into the losers bracket.
If there are 8 players moving into the losers bracket (0,1,2,3,4,5,6,7), then we have to place them into matches.
1.) (0,1,2,3,4,5,6,7) “Natural” ordering
2.) (7,6,5,4,3,2,1,0) “Reverse” ordering
3.) (4,5,6,7,0,1,2,3) “Half-Shift” ordering
4.) (3,2,1,0,7,6,5,4) “Reverse Half-Shift” ordering
5.) (1,0,3,2,5,4,7,6) “Pair-Flip” ordering
Some other terminology: “Bring-In Rounds” are rounds 1,2,4,6,8,10,12, etc in the losers bracket.
Now that we have some terminology, we can continue.
It used to be the case that for every bring-in round, we’d use Natural ordering.
Now here’s how the ordering goes for different sized tournaments
8-Man Tournament:
R1: Natural
R2: Reverse
R4: Normal
16-Man Tournament:
R1: Natural
R2: Reverse Half-Shift
R4: Reverse
R6: Normal
—5 Round replay-gap: WB-R1 → LB-R6
32-Man Tournament:
R1: Natural
R2: Reverse
R4: Half-Shift
R6: Normal
R8: Normal
—5 Round replay-gap: WB-R1 → LB-R6
64-Man Tournament:
R1: Natural
R2: Reverse
R4: Half-Shift
R6: Reverse
R8: Normal
R10: Normal
—7 Round replay-gap: WB-R1 → LB-R8
128-Man Tournament
R1: Normal
R2: Reverse
R4: Half-Shift
R6: Pair-Flip
R8: Pair-Flip
R10: Pair-Flip
R12: Normal
—9 Round replay-gap: WB-R1 → LB-R10
—6 Round replay-gap: WB-R3 → LB-R10
Here is an example of an improved, 16-man bracket:
You can see in round 2 of the losers bracket we shift the players up 2, then reverse (so instead of [juggernaut, xdarkkage, tornado, thrive], we have [xdarkkage, juggernaut, thrive, tornado]) and in round 4 we reverse the players (instead of [BYE, gokenkulegend], we have [gokenkulegend, BYE])
![[image loading]](https://s3.amazonaws.com/zeekmarketing/images/ImprovedDoubleElim.png)
Q: This post was way too long to read, I didn’t understand anything you just wrote. Can you summarize the conclusion?
A: Instead of it being possible to play someone in the losers bracket again within 1 or 2 rounds, with improved bracket seeding, we can now guarantee you wont encounter the same person twice within 5-9 rounds (the semifinals of the losers bracket).
(P.S. The figure 5-9 rounds does not include the possible 1-round rematch that occurs when the loser of the last winners-bracket match beats the winner of the losers bracket, and then plays the winner of the winners bracket in the finals. It also does not include the case when a loser of the winners bracket semi-finals re-encounters the loser of the winner-bracket finals)
If you have any feedback, or can come up with a more optimal solution, or want a more detailed explanation, then please let us know.
-Max
Z33K.com developer
Q: What was the problem with the old way of doing it?
A: The Problem:
In The Old Way of doing things, players were placed in the losers bracket in the order you saw them in the winners bracket. This meant players could face-off in the winners bracket, and then again in the losers bracket within 1-2 rounds.
For example:
1.) In round 1 of the winners bracket, Max loses against Player-X.
2.) Max moves to Round 1 of the losers bracket.
3.) Player-X moves to Round 2 of the winners bracket.
4.) Player-X loses his Round 2 match in the winners bracket.
5.) Meanwhile, Max wins against his opponent in round 1 of the losers bracket
6.) Max plays against Player-X again in Round 2 of the losers bracket.
This is the worst-case scenario, where there is only 1 round separating when these two players meet. There are numerous other instances where players could face-off against people twice in only a few rounds.
Q: Why is that unfair?
A: In the most extreme case, let’s imagine there was a player that could consistently beat all of the other players in a tournament except for one, and then that player experiences the scenario described above. If we were to do a Round-Robin or Swiss style tournament, then that player would definitely win. If the goal of the competition is to determine who the best player is, then the winner of this event could be unduly influenced by the random-seeding process.
Q: How did you fix this?
A: It depends on the tournament size. But in short, we ‘flipped’ the player order as they move into the losers bracket.
Here is the detailed, nitty-gritty answer:
There are 5 operations as we bring players from the winners bracket who have lost, into the losers bracket.
If there are 8 players moving into the losers bracket (0,1,2,3,4,5,6,7), then we have to place them into matches.
1.) (0,1,2,3,4,5,6,7) “Natural” ordering
2.) (7,6,5,4,3,2,1,0) “Reverse” ordering
3.) (4,5,6,7,0,1,2,3) “Half-Shift” ordering
4.) (3,2,1,0,7,6,5,4) “Reverse Half-Shift” ordering
5.) (1,0,3,2,5,4,7,6) “Pair-Flip” ordering
Some other terminology: “Bring-In Rounds” are rounds 1,2,4,6,8,10,12, etc in the losers bracket.
Now that we have some terminology, we can continue.
It used to be the case that for every bring-in round, we’d use Natural ordering.
Now here’s how the ordering goes for different sized tournaments
8-Man Tournament:
R1: Natural
R2: Reverse
R4: Normal
16-Man Tournament:
R1: Natural
R2: Reverse Half-Shift
R4: Reverse
R6: Normal
—5 Round replay-gap: WB-R1 → LB-R6
32-Man Tournament:
R1: Natural
R2: Reverse
R4: Half-Shift
R6: Normal
R8: Normal
—5 Round replay-gap: WB-R1 → LB-R6
64-Man Tournament:
R1: Natural
R2: Reverse
R4: Half-Shift
R6: Reverse
R8: Normal
R10: Normal
—7 Round replay-gap: WB-R1 → LB-R8
128-Man Tournament
R1: Normal
R2: Reverse
R4: Half-Shift
R6: Pair-Flip
R8: Pair-Flip
R10: Pair-Flip
R12: Normal
—9 Round replay-gap: WB-R1 → LB-R10
—6 Round replay-gap: WB-R3 → LB-R10
Here is an example of an improved, 16-man bracket:
You can see in round 2 of the losers bracket we shift the players up 2, then reverse (so instead of [juggernaut, xdarkkage, tornado, thrive], we have [xdarkkage, juggernaut, thrive, tornado]) and in round 4 we reverse the players (instead of [BYE, gokenkulegend], we have [gokenkulegend, BYE])
Q: This post was way too long to read, I didn’t understand anything you just wrote. Can you summarize the conclusion?
A: Instead of it being possible to play someone in the losers bracket again within 1 or 2 rounds, with improved bracket seeding, we can now guarantee you wont encounter the same person twice within 5-9 rounds (the semifinals of the losers bracket).
(P.S. The figure 5-9 rounds does not include the possible 1-round rematch that occurs when the loser of the last winners-bracket match beats the winner of the losers bracket, and then plays the winner of the winners bracket in the finals. It also does not include the case when a loser of the winners bracket semi-finals re-encounters the loser of the winner-bracket finals)
If you have any feedback, or can come up with a more optimal solution, or want a more detailed explanation, then please let us know.
-Max
Z33K.com developer
Here's the link to the OP: z33k.com/starcraft2/forums/4/topics/319
I would encourage people to use this method (or equivalent) over the current standards.
For instance, the 16, 32, and 128 double-elimination brackets generated by challonge could be improved if they used this method. (I didn't check brackets over 128, but it looks like their 64-man brackets use an equivalent method).
The minimum rematch distance for 16 and 32 man is 3 rounds with challonge, but with this method it's 5 rounds. Likewise, the minimum rematch distance for 128 man is either 4 or 7 rounds with challonge, but with this method it's either 6 or 9 rounds.
I'm not posting to endorse z33k or advertise, but to introduce a more fair method for conducting double elimination tournaments
- Max
z33k developer
