Esports tournament formats explained, with the maths

Not from what sounds fairest. From how many entrants you can actually fill, and how many hours you have. Everything else follows, and almost every amateur event that overruns did so because the format was chosen before those two numbers were.

So each format below leads with its match count as a formula, because that is the number that turns into hours. Multiply it by your per-match time — including setup, draft and the gap between rounds, not just the game clock — and you have your evening.

Everyone plays until they lose once. With n entrants it takes exactly n − 1 matches, because every match eliminates exactly one entrant and you need to eliminate all but the winner. A 16-team bracket is 15 matches, always.

Byes and awkward entrant counts

Brackets want powers of two. With 12 entrants, four of them get a first-round bye: 16 − 12 = 4 byes, and the byes go to the top four seeds. It is still 11 matches — the formula does not care — but it does mean four people sit out round one, which is worth saying in the rules before somebody complains.

Seeding when you have no ranking data

Amateur events rarely have real seeding data, and inventing it is worse than not having it. Random draw is honest and defensible. If you do seed, seed on something checkable — previous results in your own series — and say so.

What it is bad at

One bad match ends a night. Half your entrants play exactly one game and go home, which is fine for a free 32-player bracket and poor value for a paid one. If people paid to enter, consider giving them more than eleven minutes of tournament.

You are out after two losses. That takes 2n − 2 matches, or 2n − 1 if you use a grand final reset — roughly double single elimination, for the same entrant count.

That doubling is the whole decision. A 16-team single-elimination CS2 event at 40 minutes a match is about 10 hours of match time before parallelism; double elimination is 20. On four servers running concurrently that is the difference between an evening and a weekend.

The grand final reset

The team coming from the losers bracket has one loss; the winners-bracket team has none. If the losers-bracket team wins the grand final, both have one loss, so a second set is played. Decide whether you are running a reset before the event, and put it in the rules — this is the single most argued-about clause in bracket play.

Why the FGC uses it anyway

Because 1v1 fighting-game sets are short. First-to-two at three minutes a game is under ten minutes, so 2n − 2 is affordable in a way it is not for a 40-minute MOBA. Format choice is downstream of match length, not of fairness preferences.

Everyone plays a fixed number of rounds, and each round pairs entrants on similar records. Nobody is eliminated. You need ⌈log₂ n⌉ rounds to produce a clear winner — 4 rounds for 16 entrants, 5 for 32, 6 for 64 — and each round is n/2 matches.

So 32 entrants is 5 rounds of 16 matches: 80 matches total, but only 5 rounds deep. Compared to single elimination it costs far more matches and far less time, because every round runs in parallel and everyone plays every round.

Pairing without software

Sort by record, pair top half against bottom half within each score group, and avoid rematches by sliding one entrant down if a pair has already met. It is genuinely doable on paper for 16 entrants and genuinely painful above 32 — that is the point where Swiss stops being a manual format.

Tiebreakers

Swiss produces ties constantly, so you must state a tiebreaker in advance. In descending order of how easy they are to compute by hand: head-to-head result, then game or round difference, then strength of schedule. Pick two, write them down, stop there.

Everyone plays everyone: n(n − 1)/2 matches. It is the fairest format there is, and it dies above about eight entrants because it grows quadratically.

Which is why you almost never see it standalone. It is excellent for a 4–6 team league played across weeks, and for group stages — which is the next format.

The format most amateur events should be running, and the one they run least. Split entrants into small round-robin groups, take the top one or two from each into a single-elimination playoff.

Sixteen teams in four groups of four is 4 × 6 = 24 group matches, then an 8-team playoff at 7 matches: 31 matches. More than single elimination’s 15, but every team is guaranteed three games instead of one, and all four groups run in parallel so the wall-clock is three group rounds plus three playoff rounds — six rounds deep.

For a paid event this is usually the right answer. Nobody pays an entry fee to be knocked out in eleven minutes, and "guaranteed three matches" is a real thing to put on a listing.

You cannot bracket 25 squads. Battle royale events run a fixed number of matches and score placement plus kills, with the standings after the last match deciding it. The match count is your choice rather than a formula, and the honest floor is higher than most organizers use — three matches is mostly variance.

• Battle royale points systems

A different shape entirely: the same event, same slot, every week, with results accumulating across weeks rather than within a night. Each week is usually a small single-elimination bracket or a points event, and the season standings are the actual product.

• How to run a weekly scrim ladder

Cross-reference how many entrants you can fill against how long you have. Assume matches run in parallel where the format allows it.

Worth being direct: Fragnet does not generate brackets. There is no bracket, no seeding, no advancement and no results entry. If a bracket is the core of what you need, use a bracket tool — that is what they are for, and several of them are free.

What Fragnet covers is everything around it: the listing people find you through, the approval queue instead of DMs, the waitlist that refills a dropped slot, the entry pass, the check-in, and the room credentials that unlock on a timer. Most organizers running bracket events use both, because the two barely overlap.

• Compare platforms
• For organizers