Chess Caddy / Guides / How many rounds

How many rounds should a chess tournament have?

Work out the round count for your field size and time control — and see the reasoning, so you can defend the number when somebody asks.

Last updated: 25 July 2026 · Rules verified against the current FIDE and US Chess texts on 25 July 2026.

“How many rounds?” is the first question a new organiser asks, and most pages answer it with a table and nothing else. The number is easy: take the base-2 logarithm of your field and round up. What nobody explains is why, how much less it guarantees than people assume, and that the clock usually decides for you.

The calculator does both halves of the sum — the pairing mathematics and the timetable — and everything it does is explained underneath.

Rounds calculator

The rest of this guide is the working behind those numbers.

What US Chess Rule 27 actually says

There is a primary source for the round count, worth quoting in full because almost no guide does. US Chess rule 27, The Swiss System:

Read the second sentence carefully: it is stated the opposite way round from how everyone uses it. The rule caps the field given a number of rounds — R rounds is safe for up to 2^R players, so five covers 32, six covers 64, seven covers 128. Invert it and you get the familiar form: rounds = log₂(players), rounded up.

One attribution point: this is a US Chess rule. FIDE publishes no guidance on rounds versus field size — its Swiss chapter covers pairing mechanics, not event design. Pages crediting the rounds table to FIDE are guessing.

Why the answer is log₂ of your field

The derivation is short, and once you have seen it you will never need the table again.

Start with the defining property of a Swiss: players on the same score are paired against each other. So everyone still on a perfect score is paired inside that group. Every game there has a winner and a loser, so each game removes at least one player from the group — and a draw removes both, because neither is perfect any more.

That gives a hard ceiling. If B players are perfect going into a round, at most ⌈B/2⌉ of them are still perfect coming out. The group at least halves every round. Start with N players, all trivially perfect before round one, and after r rounds at most N/2^r can still be unbeaten and undrawn.

Now set that expression to one — the outcome Rule 27 asks for — and solve: N/2^r ≤ 1, so 2^r ≥ N, so r ≥ log₂(N). That is the whole thing.

It is also, not by coincidence, the number of rounds a knockout bracket needs for the same field. A knockout halves the field by throwing losers out; a Swiss halves the leading group by moving them down. Same arithmetic — but in a knockout half the entrants go home after one game.

What log₂ does not guarantee

This caveat gets dropped from every calculator page, and it decides how your event actually ends. log₂(N) rounds guarantees at most one perfect score. It does not guarantee a clear winner. Nothing stops two players finishing on 4½/5 with nobody on 5 — they drew with each other in round three and beat everybody else. No number of rounds prevents that, and US Chess says as much: the rule “cannot, however, guarantee a clear winner.”

So tiebreaks matter far more at club level than at elite level. Ties are the normal outcome of a short Swiss, not a failure. Decide your tiebreak order before round one and publish it — the Buchholz guide covers how the common ones behave. The alternative is to add a round, which is what the calculator's “one clear winner” setting does. That reduces ties; it cannot abolish them.

Rounds by field size, up to 200 players

Most published tables stop at 64 players, which is no help for a real open. “Minimum” is ⌈log₂N⌉; “recommended” bumps that to the next odd number the field can support, for the colour reasons below.

Note how flat that column is: doubling your entry list costs exactly one round. Anything from 17 to 32 players wants five, and 33 to 128 wants six or seven — which is why five rounds is the club default and weekend opens cluster on six and seven.

Too few rounds

Take 20 players in four rounds — a common shape, because four fits neatly into two evenings. The bound says at most 20/2⁴ = 20/16 = 1.25 players can still be perfect. That is above one, so the format has stopped doing its job: two players finishing 4–0 is not a fluke, it is permitted by the arithmetic, and it happens.

You get a shared first place settled by tiebreak arithmetic most of the room does not understand. Both leaders beat everyone put in front of them and neither played the other, so somebody leaves feeling the title was decided by a spreadsheet. A fifth round takes the bound to 20/32 = 0.625 and would have paired those two together — the difference between the event deciding itself and you deciding it afterwards.

Too many rounds

The failure at the other end is subtler. The rule you can never break is that two players must not meet twice. As the round count climbs relative to the field, the leaders have already played each other, so the engine reaches further down for a legal opponent. Pairings degrade: leaders float against players well below them, colours go out of balance, and eventually a rematch is unavoidable.

So the useful band runs from ⌈log₂N⌉ to about N/2: 4 to 8 rounds for 16 players, 7 to 50 for 100 — another way of saying that with a decent field the clock is your only real constraint.

Prefer an odd number of rounds

Given a choice between four rounds and five, run five. US Chess notes that events with an even number of rounds produce the largest colour disparities and are hardest to pair. With an odd count everyone can finish one colour out of balance, which is fine; an even-round event pushes players towards a ±2 imbalance the pairing rules must then work around.

Running five instead of four is usually the cheapest single improvement you can make to a club event. See the full Swiss guide for the rest of the setup decisions.

The clock math

This is what usually decides your round count. The standard estimate assumes a 60-move game, which makes the sum easy:

• Per-player minutes = base minutes + increment or delay in seconds. G/30 with a 5-second delay budgets 35 minutes per player.
• Worst-case game = twice that, because both clocks can run down. 70 minutes.
• Scheduled round = worst case + changeover. Fifteen minutes between rounds is realistic once you have collected result slips, verified them and posted the next pairings.

Five rounds, with four changeovers of 15 minutes between them:

Most games finish well before the worst case. You still cannot schedule on the average, because the last board to finish gates the next round. Time your rounds off the typical game and the slippage accumulates onto the final round, when everyone is tired and the venue wants its hall back.

Delay-based controls are also easier to schedule around: increment lets a player bank time and run a game far past its nominal length, while delay never adds time to the clock.

Rating bands: US Chess and FIDE

Your time control also decides how the event can be rated, and both federations classify it from the same total: base minutes + increment or delay in seconds.

The dual-rated band is why G/30 d5 is so popular in the United States: at a total of 35 it is the fastest control still earning a Regular rating, so a whole event fits in a day and counts. Several widely-copied pages give different ranges; they are reproducing an outdated archived US Chess page.

FIDE draws its lines in the same units but at different points: Blitz is 10 minutes or less, Rapid is over 10 and under 60, and classical is simply everything above that. FIDE's minimum rate of play for classical rating is not a flat 120 minutes either — it is a three-tier ladder based on who is playing: 120 minutes where a player rated 2400 or above is involved, 90 minutes where a player rated 1800 or above is involved, and 60 minutes where both players are under 1800.

Fitting rounds into your calendar

Four shapes cover almost every club event. Pick the shape first, then let it choose your control.

The weekly club championship rewards a long control most, because only one game has to fit in an evening. It is also where withdrawals hurt most: publish the round dates with the entry form and expect a few half-point byes.

If the calculator says the event will not fit, the two honest fixes are to drop a round or shorten the control. Starting late and hoping is not a third option.

When the answer is not a Swiss at all

Sometimes the right round count is the one a different format hands you for free.

• Six players or fewer: run a round robin. Five rounds, everyone plays everyone, no pairing algorithm to explain and no argument about strength of schedule.
• Seven to ten: a round robin if the calendar allows — six to nine rounds is a lot of evenings, but the result is unarguable. Above about twelve, players start dropping out.
• Round count approaching player count: you are running a round robin the hard way. Just run one, off a Berger table.
• A short, spectator-friendly event: a knockout finishes in ⌈log₂N⌉ rounds with only N−1 games — but the average entrant plays barely two of them and half the field is out after one.

For everything in between, the Swiss is the right tool and the round count is the number at the top of the calculator.

Frequently asked questions