The Swiss Defense

A few days ago, the format for the Kiev Major was announced as a Swiss group stage, followed by a Single Elimination bracket (given the time frame of the main event, the Single Elim part was pretty much a given).

Whilst most of the feedback I saw on the matter was positive for the format change, there were many people who didn’t truly understand the format (this is the first significant Dota 2 event to have Swiss in it) — and some who thought it didn’t solve the problems with the previous (Boston) format.

CS:GO has been using Swiss for a few events, but like the Kiev Major format, it’s not true ‘vanilla’ Swiss — it’s modified Swiss. Let’s discuss the similarities and differences:

  • Teams play a series of rounds, each playing a single match (bo3) against an opponent who is semi-randomly drawn. “Semi-randomly” here is because their opponent can’t be someone they’ve previously faced, and it’s someone with an identical score (wins to losses) as them.
  • In normal Swiss every competitor plays the same number of rounds, here you stop once you have 3 wins or 3 losses. This turns out to represent only a single match difference (all teams play 3 or 4 matches), but prevents there being a ‘finals’ played where one team would go to 4–0 and another would drop to 3–1.
  • In normal Swiss you’d generally use tiebreakers such as Opponent Match Win %, or Game Win % for teams with tied raw scores (matches won in this case), but I would guess that the Single Elimination bracket will be done also as a random draw for the given places (so if you end as 3–0 you could be paired against either of the two teams that end 0–3).
Possible/likely playoff bracket

Now let’s look at some ways in which this is better and worse than Boston Format.

  • [+] A single upset doesn’t mean you’re guaranteed to play a group-winner in the ro8. Here only the top 2 seeds (3–0’s) get split to potentially meet in grand-finals, so if you drop a game you can still secure a 3–1 spot which allows you to pass to the semi-finals without running into a top seed.
  • [+] Group imbalance. Nobody can complain about a ‘group of death’, and a ‘draw of death’ is way more unlikely.
  • [+] Top 4 clashes. Since there were 4 GSL groups, there’s a 1/3 chance that the truly top 2 teams meet in the round of 4, instead of the Grand Finals. No longer possible here since there’s one Swiss group.
  • [+] Fewer rematches. Whilst the EG/coL rematch was interesting from a storylines perspective, part of groups is seeing teams perform against a multitude of opponents and rematches severely limit that. 3 of the 4 groups at Boston had match #5’s being rematches (and all 3 were different results to their previous meetings).
  • [+] More games. Every team played 2 or 3 matches in the GSL groups, now they’re playing 3 or 4. We get slightly more data points to establish seedings moving into the bracket stage, but not too many.
  • [-] Asymmetrical bracket. In the round of 8, the likely match-ups will be two 3–0 vs 2–2’s, a 3–1 vs a 2–2 and a 3–1 vs a 3–1. This is because there are 2 teams that end 3–0, but 3 that end 3–1.
  • [=] Intentional feeding. A good team could intentionally lose their matches to ensure a 1–3 or 0–3 record and then have just a single upset to make in order to be in a good position (bracket-wise) to make it to the finals. This is very risky.
  • [?] Preparation. Teams have no idea who they are playing in the Swiss, so they can’t prepare specifically for those teams. In GSL you know your 3 possible opponents once the groups are drawn (once all teams arrive at the Major) but here you know your next match possibly only an hour or so before your match. This isn’t strictly better or worse than before, it’s just different (prioritizes general preparation over team-specific preparation).

A similar issue to the aforementioned ‘intentional feeding’ issue is where a single upset can disrupt the balance/fairness of the bracket. Ad Finem for example were seed #3 in their group and in the ro16 they upset a seed #2 team (Newbee) before being paired against against a seed #4 team (LGD.FY) who had upset a seed #1 team (LGD). This put Ad Finem in a favourable position despite only beating #2 and #4 seeded teams. One way to solve both these issues is to redraw each round of the bracket. This means that each round of the single elimination playoff is structured whereby the highest uneliminated seed is paired against the lowest uneliminated seed, the 2nd highest uneliminated seed is paired against the 2nd lowest uneliminated seed, etc. This devalues upsets in the playoffs substantially, and rewards a high Swiss group-stage seed.

All in all, even as we expect the Swiss group-stage to go the format seems way more fair for the teams than the GSL Boston group-stage, rewarding better performance, limiting the difference in average opponent strengths and differentiating better between placings. It’s not perfect, but it’s a big step in the right direction.