What seeding is for
A seed is a ranking you give each entry before the bracket starts. Seed 1 is the entry you expect to do best, seed 2 the next, and so on. In an elimination bracket the seeds decide where each entry is placed, and a standard bracket places them so the strongest entries are kept apart for as long as possible: in a sixteen-entry bracket, seeds 1 and 2 can only meet in the final, and seeds 1 to 4 can only meet from the semi-finals onward.
Seeding is not a reward for the favourites. It is a way of making the later rounds meaningful. Without it, a random draw can put the two strongest teams against each other in round one, so one of them leaves after a single match and the final is played between teams that would not have reached it otherwise. Everyone in the bracket is better served by a draw where results in the early rounds go roughly as expected and the close matches come at the end.
Where seeds matter in Clutchat
Every entry in a Clutchat bracket has a seed, and the seed is used in more places than the draw. When a round has a time limit set with Round time limit in /tournament-admin manage, any match still unreported when the time runs out is advanced automatically in favour of the higher-seeded side, which is the side with the lower seed number. For team events, a team takes the best seed among its players. So seeding affects not just who meets whom but who goes through when a match is not played, and it is worth getting right.
What to seed from
You need some evidence of strength that is fair, recent and visible to everyone. In a community that already uses Clutchat, you usually have several sources:
- Duel ratings. For 1v1 events, the rating players earn from duels is the most direct measure you have. It only moves when a result is confirmed, so it is hard to inflate.
- Clan Elo. For clan events, each clan's Elo from clan fights is a ready-made ranking. Every clan starts at 1,000 and moves up or down with settled fights. See clans.
- Ratings on public profiles. Each player's public profile on the website shows a rating per game, built from results and placed on a tier (see how rankings work), which helps when players come from different servers.
- Past results. How entries finished in your last few tournaments, and the server leaderboard (
/leaderboard), which ranks members by tournament results and clans by Elo. - In-game rank. Useful as a tie-breaker or for new players with no history, but self-reported and not always current.
A simple, defensible method
- Pick one primary measure and announce it before registration closes: "Seeds are by duel rating at the time registration closes." Announcing it in advance means nobody can accuse you of choosing the method to suit a result.
- Take a snapshot of that measure at the stated time, and keep it. Ratings move, and you want to be able to show what you used.
- Sort the field. Highest rating becomes seed 1. Break ties with a second measure you also announced, such as past finishes.
- Handle entries with no history. Place unrated entries after all rated ones, ordered by in-game rank if you have it, or at random if not.
- Publish the seed list before the bracket starts, so mistakes are caught before they matter.
Seeding teams
Teams are harder because their members may have very different ratings. Three common approaches, from simplest to fairest:
- Best player. The team takes the rating of its strongest member. Simple, and it matches how Clutchat picks a team's seed from its players, but it overrates a team with one star and four newcomers.
- Average of the starting roster. Fairer, as long as you average only the players who will actually start, not substitutes.
- Team results. For established clans, clan Elo already describes the team as a unit and is usually better than any average of individuals.
If teams are formed from solo signups by /tournament-admin start, you can use manual mode to balance teams by rating, in which case seeding matters less because the teams should be roughly even by construction.
Byes and odd numbers
When the field is not a power of two, for example thirteen entries in a sixteen-place bracket, some entries get a bye into round two. Byes conventionally go to the top seeds, which is the most visible consequence of seeding and the one players will check first. If your field shrinks at check-in (see check-in that stops no-shows), seed after check-in closes, not before.
When not to seed
Seeding is not always worth it. For a casual, fun-first event, a fully random draw is perfectly fair and saves an argument about methods. If you have no reliable data, a random draw is more honest than seeds made up from reputation. And in round robin, every entry plays every other entry, so seeding has little effect on the outcome. Swiss pairs entries with similar records after each round, so seeding only shapes the first round.
Common mistakes
- Changing the method after seeing who registered.
- Seeding from ratings that are weeks out of date.
- Giving a friend's team a higher seed because "they are good really".
- Seeding before check-in and leaving a no-show as seed 1.
- Not publishing the seed list, so mistakes are only found mid-event.
For the rest of the tournament night, see the sixteen-team bracket guide, and for how ratings are calculated in the first place, see rankings.