Choosing a tournament format

You have a group of players and an amount of time. Which format actually fits, and what does each one promise?

Almost every argument about tournament format is really an argument about arithmetic that nobody has done. A knockout for sixteen is four rounds and fifteen matches; a round-robin for the same sixteen is fifteen rounds and a hundred and twenty. Those are not two versions of the same event. This guide gives the numbers first, because once you have them the choice usually makes itself, and then says what each format guarantees — which is the part that decides whether the result will be accepted.

The numbers, before anything else

Every format's cost is a formula, not a matter of taste. For n entrants: a single knockout is n−1 matches over log₂(n) rounds. A single round-robin is n(n−1)/2 matches over n−1 rounds, or n rounds if n is odd and somebody sits out each time. Double elimination is roughly 2n−2 matches, because almost everybody has to lose twice. Swiss is a number of rounds you choose — usually ⌈log₂(n)⌉ — times n/2 matches.

The gap between these grows fast and it grows in the direction people underestimate. At eight entrants a round-robin is twenty-eight matches against a knockout's seven: four times the work. At twenty-four entrants it is two hundred and seventy-six against twenty-three, which is twelve times. If you are choosing between a knockout and a round-robin for more than about ten entrants, you are not choosing a format, you are choosing whether the event takes an afternoon or a season.

What each format costs and guarantees, for n entrants
Format	Matches	Rounds	Guarantees	Everyone plays
Knockout	n−1	log₂(n)	One clear winner, fastest possible	One match minimum
Double elimination	≈2n−2	≈2·log₂(n)	Nobody out on one bad result	Two matches minimum
Round-robin	n(n−1)/2	n−1 (or n)	Everyone meets everyone	Every round
Swiss	rounds × n/2	You choose	Balanced opposition, no eliminations	Every round
Challenge ladder	As challenged	Ongoing	Nothing — it is voluntary	Only if they challenge
Box league	Per box, ~6–10	Ongoing cycle	A full card at your own level	Every round

One day, one winner: knockout or double elimination

If the event has to finish today and produce a winner, this is the choice. A knockout is the fastest structure that exists for the job and its weakness is exactly as bad as it sounds: half the field plays once and goes home, and the two best entrants can meet in round one and knock one of them out. Seeding mitigates that and does not fix it.

Double elimination roughly doubles the matches and buys one specific thing — no entrant goes out on a single bad result, which matters when people have travelled, when the draw is unseeded, or when one match is a poor sample of who is better. If your event has a lot of first-time entrants, or the matches are short, the extra half-day is usually worth it.

The awkwardness of double elimination is the end rather than the middle. The entrant who comes through the losers' bracket has one loss and the entrant waiting in the final has none, so a single deciding match is not symmetric. The two honest answers are a bracket reset — the loser of the first final has now lost twice, so a second final is played only if the unbeaten entrant loses — or to state in advance that the unbeaten entrant needs only to win once. Both are defensible; deciding which after the final has started is not.

A season, and fairness matters more than speed: round-robin

A round-robin is the only format that removes the draw as an excuse — everyone plays everyone, so the table at the end reflects the whole field rather than who you happened to be drawn against. That is why leagues use it and why it is the right answer whenever the competition runs over weeks rather than hours.

Its cost is quadratic and it is unforgiving about it. Beyond about twelve entrants a single round-robin stops fitting into a season, and the usual fix — splitting into groups and playing a knockout between the winners — is really a different format wearing a round-robin's clothes. Decide which one you are running before you publish the fixtures.

One detail is worth settling early because it is invisible until somebody complains. In a schedule with an odd number of entrants somebody sits out every round, and in a double round-robin the home and away split cannot come out exactly level over an odd number of rounds. Neither is a flaw in the schedule; both are arithmetic. Say so when you publish the fixtures, rather than being asked about it in week six.

Too many entrants for a round-robin, too few rounds for one: Swiss

Swiss exists for exactly the gap the two formats above leave. It keeps everybody playing for a fixed, small number of rounds — five rounds handles thirty-two entrants — and it decides pairings from results rather than from a draw, so after round two you are playing somebody who has done as well as you have. Nobody is eliminated and nobody spends the afternoon being beaten by the eventual winner's warm-up.

What it does not give you is certainty. Five rounds of Swiss will usually identify the strongest entrant and will occasionally not, and two entrants can finish on the same score having never played each other. That is the trade: you buy a manageable round count with a small amount of doubt, which is why chess and most large one-day events run it and why a cup final does not.

Swiss Pairing GeneratorPairs each round from the current standings, avoids rematches, and never gives the same player two byes.

No fixed end, and people arrange their own matches: ladders and boxes

Clubs where there is nobody to write a fixture list need a format that runs itself. A challenge ladder is the simplest: everybody has a rung, you challenge somebody above you, and a win moves you up. It needs no organiser at all beyond keeping the order. Its weakness is that it guarantees nothing — the player on rung thirty can go a whole season without a meaningful game, and often does.

A box league fixes that at the cost of a little organisation. Split the club into boxes of four or five who all play each other, then promote the top and relegate the bottom of each box at the end of the round. Everybody gets a full card of matches against players of their own standard. It is the format most racket clubs settle on once they pass about twenty-five members, and the reason is precisely the ladder's dead bottom half.

Both formats live or die on round length rather than on rules. A ladder with no fixed challenge round goes stale within a month because nobody wants to be the one who issues the first challenge; generate a round instead and the problem disappears. A box round that runs until every match is played never ends, because two people will always be unable to find a date. Fix the length in advance — three or four weeks for boxes of four — and start the next round on schedule with whatever is outstanding, applying the same rule to unplayed matches every time.

Whichever you choose, decide the tiebreaks first

This is the single most common failure and it has nothing to do with the format. Two entrants finish level, nobody wrote down what happens next, and the decision now looks like it was made to suit somebody. Publish the tiebreak order before the first match, in writing, with the entry details.

The order genuinely changes the result. Separating on goal difference first is what most domestic leagues do; separating on the tied teams' own matches first is what UEFA and most cup group stages do. The same results can put a different team top depending on which you picked, so picking afterwards is not a neutral act.

Head-to-head is also the rule most often applied wrongly, and the mistake is understandable. With two teams level it means the match between them. With three or more it does not mean anything so simple — there is no single match to look at, so the rule means a table built from only the matches those teams played against each other, in which a nine-nil win over the bottom side counts for nothing. Work that out on paper and it is easy to get wrong; it is worth having something derive it for you.

League Table GeneratorApplies either order and names the rule that separated each pair, so the table can be published with its own justification.

The bottom line

Do the arithmetic before the argument. If it must finish today, knockout — double elimination if one bad result should not end anybody's day. If it runs for weeks and fairness is the point, round-robin. If there are too many entrants for that and only one day, Swiss. If there is no organiser, a ladder for a small club and boxes for a large one. And whichever you pick, publish the tiebreak order before the first match.