CS2's Swiss System: A Broken Format?
People think the Swiss system is what's ruining the CS RMRs. Let's talk about it.
Preamble
Be sure to follow me on Twitter to flame me give feedback and engage in discussions!
The European RMRs have concluded (and as of writing this, the Americas have yet to begin). 32 teams walk into those hotels and studios with one goal, and only 17 walked away triumphant…except “only 17” is an oxymoron of sorts. Suffice to say it is difficult to sort such a large number of teams, many of whom are not well known or significantly better than the other, into a perfect system where the best rise to the top - and Valve settled on the Swiss System for their contemporary take at Major qualification.
We’ve had a few different kinds of formats used across the years for this all-important stage leading up to the Major. For instance, the StarLadder Berlin 2019 EU Minor was a 2x4 GSL group stage, which came from a closed qualifier that utilised a simple double elimination bracket (which came from the open qualifiers that we know and love). Before that, we have had 2x8 double elimination groups in qualifiers for Cologne 2015 Europe, and 4x4 GSL groups for the Offline Qualifier for Cluj-Napoca 2015 - and even round robin in the dark days of the Road to Rio in 2020. (I will not be talking about round robin.) There are no doubt more tournaments I did not mention, but these are just examples of different formats we have seen.
Many improvements have been made for the current state of Major qualification. Seeding is now done using the Valve Ranking at certain stages, and invites to closed qualifiers speed up the process and allow teams with existing presence to skip the hellish open qualifiers. Yet this is not enough for many people - their gripe is with the format to begin with, known as the Swiss system. I’m sure those of you who read articles know what the Swiss system is, but it would be untoward of me to not try to explain it.
As per the Valve rulebook, a sixteen team group is formed by spreading the qualification slots between either one or two groups (Americas has one, EU has two) depending on the number of slots the region has. The groups are seeded such that the average strength of the groups is equal (essentially alternating in ranking). Within the groups themselves, the first matchups are as above. Subsequent matchups of teams with the same Swiss score are done using Buchholz to match the top and bottom teams first, and original seeding second (more about that later).
There’s a fancy calculator that lets you figure out the next round’s matches, but each team essentially has no guarantee of who they could possibly face round to round. This is contrasted by a GSL group, where you know the three other teams, your next matchup and the possible subsequent matches you have to play.
Now that I’ve gone over the droll details, the prominent question is “why” - why talk about it? Well, people have been complaining about formats for ages. I myself wrote about the perfect Major format over a year ago (looking back on it…that is one complicated group stage) where I replaced the Swiss system with a group stage. However, the Majors are not the RMR. There are more teams and weaker teams at this qualification stage, and different decisions should be made.
Most people have three problems with the Swiss format; the matchups are bad, the format is “too random” in results, and a general distaste for the lack of predictability and consistency compared to a standard group format. There is an unseen and unsolvable fourth problem which I will keep secret until the end, because it ruins the entire premise of the discussion.
I’ll be using EU RMR Group A as a data set to demonstrate some talking points - let’s get into the problems.
The Matchups Suck…?
When somebody talks about matchups, what they are talking about is seeding in disguise - since that decides the matchups, at the end of the day. If a team had a seemingly unfair route in the Swiss system it would be because their seeding did not reflect their perceived strength as a team from the outside perspective.
Here are the teams in order of Valve Ranking for the aforementioned RMR EU A. Apart from the glaring issue of Falcons being 16th (due to Valve ranking updating in a strange order and the seeding being chosen at an odd time), everything else seems fairly reasonable.
I propose a fair question - sure, there is a ranking difference that leads to a direct linear order between AMKAL and fnatic. But can you meaningfully tell me that AMKAL at #6 is so much better than fnatic at #15 to the point where that seeding should stick throughout the entire system? Furthermore, would it be reasonable to split these teams into GSL groups knowing that the order is somewhat flimsy? Let’s find out.
By splitting the teams into pots (each row of four is a pot) and making four groups (Group A: 1 8 9 16, Group B: 2 7 10 15, Group C: 3 6 11 14, Group D: 4 5 12 13), we get the following GSL groups:
Group A
FaZe 3DMAX 9Pandas Falcons
Group B
VP BB SAW fnatic
Group C
NaVi AMKAL KOI ENTERPRISE
Group D
G2 EF ITB NIP
The interesting thing is that many of these matchups in the groups actually happened in the Swiss system anyway. In fact, in each of these groups, one of the teams actually played another in the same “group” during the actual tournament, and in most of them this is overwhelmingly the case (take Group A, where many of the matches did occur).
The issue with seeding the teams into GSL groups versus Swiss matchups is that no room for error is given. Too much reliance is placed on the initial seeding. In the actual Major this would be fine - you would hope most of the teams are established enough that an accurate order can be formed and groups would thus be fairer. But in the RMR teams such as SAW, KOI, 9Pandas and 3DMax cannot really be distinguished between each other based on previous performances in my opinion. Any differences in ranking is, at least to my eyes, ephemeral and not of a high enough modality.
This is why we have the Buchholz seeding. People rarely understand it, and it is certainly confusing on the surface - but in actuality it is simply a performance ranking based on the teams you have played in the tournament up to this point. If you played teams that are now in the 0-2 bracket, the Buchholz score basically says “you’ve been farming trash teams, you can’t be given a high seed”. If you’ve played teams that are winning, you are granted a higher seed due to your “tougher” opposition beforehand.
Previously when the same Buchholz score had ties broken by a busted seeding, you got the scuffed matchups that people complain about.
This was a high matchup in the Paris RMR. Naturally, people wondered why in the world B8 and fnatic were gifted this. The reason was that the seeding was based on prior Major performance instead of a ranking, and fnatic had the higher seed due to their quarterfinal appearance. The Buchholz score in round 3 was tied for FaZe and fnatic, so the seeding kicked in and the story writes itself. This is no longer possible now that the seeding used is a ranking, even if Valve’s ranking is imperfect. We already see that the consistency of matchups is better with a half-decent ranking, and I expect it to get better as Valve continue to apply it.
So the complaints about a bad matchup now are purely based on whether you agree with the concept of Buchholz scaling the strength of teams during a tournament. In my opinion, as stated earlier, the high difficulty in accurately ranking 16 teams in a RMR which will no doubt have new teams and less frequently seen teams participating pushes me in favour of using Buchholz rather than fixed GSL groups where a group’s strength can fluctuate.
Too Random!
Let’s distil this quickly before people get distracted. The vague complaint that Swiss systems are “random” stems from a few places; the matchups (discussed above) and the results. Certainly, the fact that the matchups are not fixed in nature can lead to the perception of randomness - that is merely a feature of the system. However, I believe the true problem is the Bo1s that are used in the format that make it “more random”.
It’s no secret that a Bo1 is less consistent than a Bo3 when it comes to determining the best team out of the two competitors. The veto is simplified, limiting experienced teams with diverse map pools, and there is no chance for a team to reset after a bad map. That’s not to say that the better team is not the victor of a Bo1, but that the result should be taken with a smaller degree of confidence.
The Swiss system used in the RMRs (and Majors, and CQs) treats a Bo1 victory or loss equal to a Bo3 victory or loss. A team can be 2-2 having won 2 Bo1s and lost 2 Bo3s, or having won 2 Bo3s and lost 2 Bo1s. Anyone with a brain can tell you the difference between the two. This is the real flaw of the exact system used in the Majors.
And yet, would it be better if the same kind of format (only Bo3 for elimination/qualification games) was translated to a GSL group? In that case, one Bo1 and one Bo3 is all that separates a team from elimination or qualification. At least the Swiss system, in this one to one comparison, gives teams more chances to recover from poor Bo1 performances.
If you have to stick with the same base rules for the kind of series played that Valve mandates right now, it seems more unfair to return to a format where the Bo1 matches actually mean more, rather than less. Especially if your distaste of the Swiss system is based on the inaccuracy of the results.
I Just Prefer Groups.
You can’t argue with the platitude itself as stated above, so let’s try and break it down. Apart from the earlier reasons of matchups and results which I have separated already, the other effects of fixed groups are that you know your opposition and thus can more effectively prepare for them, and the possible matchups are known beforehand, allowing better stories to be crafted for the broadcast and viewers.
This is just prima facie and is never going to be remedied between formats. I can understand the value of it - I also believe that there is merit to knowing your opponent as soon as the round is over, thus giving every team one night to frantically study their newest opponent. If this is your primary reason for choosing GSL groups then I will let you pass without interrogation.
It’s just more difficult to craft storylines for a Swiss system on the fly. In a group you can build up beef between teams, talk about how a certain IGL will counter a team if they both win their games, and so on. The Swiss storylines mostly happen by accident, and there is a magical element to that as well. To each their own in this aspect.
The Unsolvable Unseen Dilemma
If you’re sharp you already know what I am about to say. After all, it’s a product of the real world, and real problems are the worst because you can’t always reason your way out of them like some sort of chess puzzle.
The Major is run by a tournament organiser who has to keep in mind their bottom line and schedule. Currently, one RMR group of this size takes four days, with five rounds (day one containing both opening Bo1 rounds). With an average of 8.5 maps played consecutively (and on the first three days, in parallel on both streams) any format that is discussed needs to fit within the constraints of the actual schedule. My “apples to apples” format comparison which includes Bo1s in GSL does not apply itself to the real world.
This is where the undeniable advantage of GSL groups exists. Each group of 4 teams includes 5 matches - the two openers, an upper and lower bracket for qualification and elimination, and a consolidation final. That’s 20 matches overall - so 60 maps (counting each Bo3 as 3 maps). The current Swiss format contains 33 matches, but 20 of them are Bo1s, which makes 59 maps. For essentially the same overall length, you could have every game be a Bo3; that is hard to argue against when you realise how difficult it would be to run each match in Swiss as a Bo3.
In MR12 it is reasonable to run 4 Bo3s in a day (BLAST’s groups do this already!), and each round has a maximum of eight matches, so the two streams could handle one round per day. This means you would have to extend every 16 team Swiss group schedule by one day, and make each day a little longer. Not impossible, but which TO would accept this deal when the only thing you gain is the good will of a few diehard fans, and nothing else?
The important part is that the schedule is not mandated by Valve at all, at least in their rulebook. It is the TO who cuts the costs to be able to run the event. ESL has infinite money and could easily run the extended schedule if they so desired, but PGL claim to be profitable. As such there is no future where they willingly change this format to be more “robust” - it will have to be Valve forcing Bo3s in every match to change this.
I earnestly believe that a full Bo3 Swiss system is better than a full Bo3 GSL group stage of four groups for the RMR. But if the discussion is between the latter and the current system with Bo1s, I could be convinced despite my preference for the dynamic seeding of the Swiss system. Begging Valve to change the format is the best alternative…but I’m not sure I want to see the cost cutting measures PGL would take to run the ensuing format.
Conclusion
The current Swiss system has solved many of the problems that people have been complaining about, though they will continue to do so not knowing that they have been solved. The seeding makes a lot more sense, and the matchups as well as the results are relatively stable.
However, the inclusion of MR12 Bo1s is a sore spot for the format and is an undeniable edge towards having GSL groups instead. Less chances but more robust and reliable matches is definitely worth consideration. The ideal solution of Bo3s in Swiss is a scheduling and financial cost to the organiser that they will not shoulder unless their hand is forced.
Ironically, the most realistic way to get a change in format is to ask for Bo3s instead of Bo1s for Swiss. From Valve’s perspective very little about the format they have devised has to change; just one number in their document would reduce the audience’s doubt in the format (to the anguish of PGL’s CEO).
However, I still believe the current format is sufficient to decide, on average, the best teams who should go to the Major. The system affords a lot of chances to offset the perceived randomness of Bo1s, and while some teams may face harder or easier runs, the overall experience is straightforward. For the Major, I have had stronger words to describe the expectations for a format. For the RMR, where you have to sift through a lot of teams? I prefer it.
Afterword
Thank you for reading my article all the way to the end! I enjoy thought pieces and discussion provoking stuff like this - though sometimes it takes me a while to put it in writing. I’ve been steadily increasing in subscriptions, so thank you to everyone who wants to see more - I’ll continue with this little passion project of mine.
I’ve been uploading some gaming content pretty much every day on YouTube, so if you are bored check it out below. See you next month, hopefully with less procrastination!