Randomising a Tournament Draw
A blind draw can put your two best entrants in round one. Seeding exists to stop that.
A tournament draw is where randomness and spectacle openly conflict: an unbiased draw can put the two best entrants in the first round, and seeding deliberately reduces randomness to protect the later rounds from that outcome.
A blind draw can put your two strongest entrants in the first round, and about 38% of draws pair some two of the top four. Seeding exists precisely to prevent that, at the cost of being less random.
Neither is wrong. The choice depends entirely on whether you have ranking information you actually trust.
What a blind draw actually risks
In a sixteen-slot bracket drawn at random, the two strongest entrants meet in the first round about one time in fifteen. Some two of the top four meet in the first round considerably more often — about 38% of draws, which is 1 − (14/15)(12/14)(10/13).
When that happens the tournament has spent its best match before anyone has warmed up, and the final is between an entrant who beat a favourite and one who did not have to. That is not unfair in any procedural sense and it produces a worse event.
The other risk is byes. In an unseeded draw the byes fall at random, so an entrant who did not need one gets a free pass to the second round while a strong entrant plays an extra match.
A 16-slot unseeded bracket
- Top two meet in round one: about 1 draw in 15.
- Some two of the top four meet in round one: about 38% of draws.
- With 4 byes, each entrant has a 25% chance of one regardless of strength.
What seeding costs
Seeding is not a tweak to the draw — it removes the randomness almost entirely. The same seeding order produces the same bracket every time, so the draw stops being a draw and becomes an announcement of a structure you determined.
It also concentrates ranking error. A strong entrant placed at seed twelve is guaranteed to meet seed five in round one, where an unseeded draw would only have done that by chance. A badly seeded bracket is genuinely worse than an unseeded one.
The test is whether your ranking is based on something real. Previous results in the same competition, a rating system, a league table — those support seeding. An organiser's impression of who is good does not, and seeding on it makes the impression decisive.
The middle options
Pot seeding is the common compromise and it is used by most large tournaments. Divide the field into pots by strength — top four in pot one, next four in pot two, and so on — then draw at random within the constraint that each group takes one entrant from each pot.
That keeps the strongest entrants apart without determining the whole bracket, so the draw remains a genuine event with a real outcome. It also degrades gracefully when the ranking is imperfect: a mis-ranked entrant lands in the wrong pot rather than in a specific guaranteed fixture.
The other middle option is seeding only the byes. Give the free passes to the top entrants and draw everything else blind, which addresses the most concrete unfairness in an unseeded bracket while keeping the pairings open.
- Full seeding: strongest apart, no randomness, concentrates ranking error.
- Pot seeding: strongest apart, genuine draw, degrades gracefully. The usual compromise.
- Byes only: fixes the most concrete unfairness, leaves pairings blind.
- Fully blind: maximum randomness, risks spending the best match in round one.
Doing the draw where people can see it
A tournament draw is an event in its own right, and the same principles apply as to any public draw: announce the method before drawing, do it visibly, and do not re-run it.
The specific failure to avoid is drawing privately and announcing the bracket. It is not that anyone assumes manipulation — it is that a bracket with no visible draw has no answer if somebody later suspects one, and generating that answer afterwards is impossible.
If the draw is seeded, say so and publish the seeding order. A seeded bracket presented as a random draw is the thing that actually causes trouble, and it costs nothing to describe accurately.
Frequently asked questions
How often does a blind draw pair two favourites early?
In a 16-slot bracket, the top two meet in round one about one draw in fifteen, and two of the top four about one in five.
Should I seed?
Only if your ranking is based on something real — previous results, a rating system, a league table. Seeding on an impression makes the impression decisive.
What does seeding cost?
Almost all the randomness. The same seeding order produces the same bracket every time, so the draw becomes an announcement.
Is a badly seeded bracket worse than an unseeded one?
Yes. It concentrates the ranking error — a mis-seeded entrant is guaranteed a hard early match rather than merely risking one.
What is pot seeding?
Divide the field into strength pots and draw at random subject to each group taking one from each pot. Keeps favourites apart while leaving the draw genuine.
Can I seed only the byes?
Yes, and it is a good minimal option. It fixes the most concrete unfairness in an unseeded bracket while leaving the pairings blind.
Should the draw be done publicly?
Yes. A bracket with no visible draw has no answer if anyone later suspects one, and that answer cannot be generated afterwards.
Do I have to say whether it was seeded?
Yes. A seeded bracket presented as a random draw is what actually causes trouble, and describing it accurately costs nothing.
Tools for this job
Tournament Bracket Generator
Draw a single-elimination bracket with byes — random draw or classic seeding.
Round-Robin Schedule Generator
Everyone plays everyone once, in the fewest rounds — by the circle method.
Balanced Team Splitter
Rate each player, then snake-draft into teams with matched total strength.
Background reading
Bracket and Fixture Scheduling Maths
Byes, seeding, and the circle method that fits a full round-robin into n-1 rounds.
Constraint Satisfaction in Fair Draws
What happens when the rules and the randomness disagree — and when there is no answer.
How Weighted Random Selection Works
Cumulative sums, one float, and why duplicating entries is the wrong way to do it.