Knockout or League? How the Structure Decides Who Wins
By the BrainSnail editorial team. How these articles are written and checked, and how to tell us when one is wrong.
The same set of competitors produces different winners depending on how the competition is organised. Each format trades accuracy against length and drama, and none of them reliably identifies the strongest entrant.
The main structures
Competitions are built from a small number of arrangements. A single elimination bracket pairs competitors and removes the loser of each match, halving the field each round, which is the shortest format and the most sensitive to a single bad performance. A round robin has everyone play everyone, which uses far more matches and produces a far more reliable ranking. A group stage followed by a knockout combines the two, using round robins to reduce the field and elimination to produce drama. A double elimination bracket gives each competitor a second chance. A ladder or league runs continuously over a season. Swiss pairing matches competitors with similar records each round without eliminating anyone, which is standard in chess.
What each format costs and buys
The trade-offs are consistent and quantifiable:
- •Single elimination needs the fewest matches and produces the least reliable ranking
- •Round robin produces the most reliable ranking and grows quadratically in matches
- •Knockout guarantees a decisive finish, which suits broadcasting and ticketing
- •Round robin can end in ties requiring arbitrary tiebreakers
- •Swiss pairing keeps everybody playing while converging on a ranking efficiently
- •Seeding reduces the chance of the strongest competitors meeting early, which protects the favourites
How often the best entrant wins
The question can be answered quantitatively by modelling each match as having a probability determined by the competitors' strengths. Under reasonable assumptions, a single elimination tournament of sixteen competitors identifies the strongest one well under half the time, and the figure falls as the field grows and as the strength differences narrow. Round robin does substantially better and still fails regularly. Best of several matches per tie improves reliability considerably at proportional cost in time, which is why formats in sports where upsets are common use longer ties in later rounds. The general result is that no practical format reliably identifies the strongest competitor, and the more evenly matched the field, the less any format can. Organisers have generally accepted that and optimised for drama instead.
The bracket itself
Drawing up a knockout bracket involves choices that shape the competition before anything is played. Seeding places the strongest competitors in different sections so they meet as late as possible, which increases the chance that the final is between the two best and reduces the chance of an early upset removing a favourite, and seeding is therefore a deliberate thumb on the scale justified by wanting the later rounds to matter. A field that is not a power of two requires byes, which advance some competitors without playing and are conventionally given to the highest seeds, adding a further advantage. Random draws after each round rather than a fixed bracket change the incentives entirely, since nobody can see their path. And regional protection, keeping competitors from the same country apart early, is common and cuts across seeding.
The incentives a format creates
Structures produce behaviour that their designers did not intend, and several notorious cases are instructive. A group stage where the final matches are played at different times allows the later teams to know exactly what result suits them, which produced an infamous match in 1982 where both sides stopped competing after an early goal gave both what they needed, and simultaneous final fixtures became standard afterwards. Formats where finishing lower in a group gives an easier knockout draw have produced teams visibly trying to lose, including a badminton case at the 2012 Olympics that ended in disqualifications. Seeding based on rankings creates incentives to manipulate those rankings. And any format where a team is already qualified produces matches nobody is trying to win.
The takeaway
Single elimination needs fewest matches and is least reliable, round robin is most reliable and grows quadratically, and hybrids trade between them. Modelling shows that a sixteen-competitor knockout identifies the strongest entrant well under half the time. Formats generate unintended incentives, which is why final group matches are now played simultaneously.