A fresh look at improving tourneys

This article is sort of a follow-up to my feature "Improving tournament formats", which was released in May 2012, and helped event organizers adapt to the GSL-format in groups in Counter-Strike starting with GameGune of that year.

So far the community-funded majors have all used the same format; four groups of four teams each played in GSL-format, followed up by a single elimination best-of-three bracket. Still, what could we do to improve on it in time for ESL One Cologne 2014?

EMS One Katowice was great, but could it have been even better?

Map pool and selection process

I think it's a given at this point that we should go back to the seven map pool most tournaments used in 2013 before Valve limited the maps played to the ones that ship with CS:GO. That's fine, and it makes sense for spectators, but then Valve needs to start adding more maps into the game.

We have some good ones available in de_cache and de_season, both of which have already been used extensively in tournament play. Some still also hold out hope for Colin "Brute" Volrath to release de_tuscan during this decade, and it should surely be included over a map if it does indeed surface at some point.

It's not realistic to expect teams, whether professional or not, to practice seven maps. It's not possible to be really good on seven maps at one point, certainly at least not as good as you would be playing a smaller number of maps. Since map selection needs an uneven number of maps, and seven has been proven to work before, we should stick with it.

If you're from the group of people who always say something along the lines of "professionals should play all the maps", please stop. Surely you would rather watch them be really good on seven, than average on nine. Let's work to find compromises that work both for the players and the viewers, there has to be some give and take.

Let's say we get Valve to add de_cache and de_season for now, and we have a seven map pool. If de_tuscan comes out at some point, we could remove the least played map, most likely de_train, and substitute de_tuscan in to keep the pool at seven going into the future. Another point to bring out is that we shouldn't always hold onto the old five maps - if one of them stops being popular, we should sub them out, like de_train.

This should promote more matches on other maps, and maybe half the group stage games, and every playoff series, wouldn't feature de_inferno anymore. It would remain popular, but at least there would be more alternatives. However, we need Valve to act on this first - not many tournament organizers will go rogue against the rest.

Map	DH Winter	EMS One Katowice	DH Summer
de_dust2	9/40 - 23% (3/20)	7/37 - 19% (5/20)	7/36 - 19% (4/20)
de_inferno	19/40 - 48% (12/20)	16/37 - 43% (11/20)	16/36 - 44% (9/20)
de_mirage	4/40 - 10% (2/20)	7/37 - 19% (2/20)	7/36 - 19% (5/20)
de_nuke	4/40 - 10% (2/20)	4/37 - 11% (2/20)	3/36 - 8% (0/20)
de_train	4/40 - 10% (1/20)	3/37 - 8% (0/20)	3/36 - 8% (2/20)

Numbers in parentheses are during group stage

Currently de_inferno dominates tournament play. A whopping 63% of all matches (e.g. a group game or a best-of-three series) featured it at DreamHack Winter, EMS One Katowice and DreamHack Summer (and it may have been third map in the series it wasn't played in). More than half of all group stage games, 32 of 60 to be more exact, were also played on de_inferno. Looking at the numbers above, the current system is clearly broken.

As for Valve's unwillingness to add more maps, how about listening to your customers? In a quick survey of Twitter replies, almost everyone wanted to see more maps being played. Who are the people voicing their opinion? Probably the same people who are funding this tournament. The developers shouldn't ignore them completely, especially when they have a valid argument to support their case.

People longing for a system where maps are pre-determined should give it a rest. One of the most fascinating things about Counter-Strike is the differences in teams' strategy, and that includes not playing a certain map. If maps were pre-determined, it'd suddenly favor teams who happen to be good on the ones played in semi-finals, and the grand final, and you couldn't have seven maps in the pool, because it would be too much for teams to prepare for properly.

de_inferno is played far too often in today's CS:GO 

However, we should also consider a subtle change in the map selection process. Currently, whether the pool is five or seven maps, teams take turns eliminating maps until three remain (some events have the second team remove two in the middle, but it's hardly relevant to the point here). They then pick their own maps, and the last map remaining will be the deciding third map, if necessary. It's a good system, but spectators are growing bored of the maps it produces.

What if instead of removing four maps off the bat, each team removed one, then picked their maps, then removed one more each, and the final map would be the third? That'd give teams a bigger incentive to practice more maps, because they could only ignore one map completely. It's not ideal for the teams, but it would be a good compromise between the players and viewers.

Current system	Proposed system
1. Team A removes a map	1.  Team A removes a map
2. Team B removes a map	2.  Team B removes a map
3.  Team A removes a map	3.  Team A picks first map
4.  Team B removes a map	4.  Team B picks second map
5.  Team A picks first map	5.  Team A removes a map
6.  Team B picks second map	6.  Team B removes a map
7. Remaining map is played third	7. Remaining map is played third

That would mean if Na`Vi refused to play de_nuke, an opposing team could force them to play a lesser played map, such as de_season. Similarly, if Titan would still ignore de_train, teams could force them to practice de_cache instead, for example. It would open things up, and certainly motivate underdogs to practice the less played maps, and force the top dogs to play them instead.

Another option to get more maps played, and one that would also work for the group stage, is a point system with eliminated maps. In that system teams would pick one map in the playoffs, two maps in the group stage, that they wish to eliminate against each opponent. The picking system is explained below:

Team A vs. Team B

1. Team A removes de_mirage
2. Team B removes de_train

Team A's map order: 1. de_nuke (7p), 2. de_inferno (6p), 3. de_train (5p), 4. de_dust2 (4p), 5. de_cache (3p), 6. de_season (2p), 7. de_mirage (1p)

Team B's map order: 1. de_nuke (7p), 2. de_cache (6p), 3. de_season (5p), 4. de_mirage (4p), 5. de_dust2 (3p), 6. de_inferno (2p),  7. de_train (1p)

First map in this series would be de_nuke with 14 points, second map would be de_cache with 9 points, and the deciding map would be de_inferno at 8 points.

This system would promote the newer maps, because underdogs could spend more time practicing them and sort of force them on the more established favorites.

The teams would then rank the remaining maps in the order they wish to play them. Organizers could force teams to submit their map order and the map removals for every possible opponent prior to the event starting, and we would already know all the maps for group stage games and best-of-three series in advance. That would be great for providing better coverage and analysis, and for betting.

An admin would then look at the points each map would receive, and the one most popular among both teams combined would be the one played. In reality this would all be figured out before first matches go live, thus removing possible map picking delays. It would also remove the randomness of a coin toss in having to start removing maps first. In a tie scenario in points among two maps, you would go with the lower seed's choice, giving them a small advantage since the higher seed is favored with everything else.

It would be a fairly complicated system and therefore not a popular choice from an event organizer's perspective, but it could be sorted out effectively without too much trouble, it would make the map selection a lot more interesting, I believe, and what's most important is that teams would more often meet on their best map - not the one that's neither team's best, nor their worst, thus producing better matches.

Na`Vi is a victim of best-of-one group stages, having gone out in both majors 

Group stage

I still think there's nothing better for group stage than the GSL-format, which was outlined here. We should hold onto it, but I also think we should try to force a tiny adjustment for major events, although it would require both Valve and event organizers to play along; a long shot in time for ESL One Cologne. Still, if DreamHack Winter gets picked for a second major this fall, we could make it happen in time.

Best-of-one group stages are fine for medium sized events such as DreamHack Summer or Copenhagen Games, which should give underdogs a chance at making a name for themselves by upsetting some of the bigger names and getting further in tournaments. However, when teams spend months practicing for a major, we should be playing best-of-three from the get-go.

Here's the problem; this idea requires either more time, or more computers and more matches played simultaneously. Still, there comes a point you need to trade a little bit of the entertainment factor to strengtten the integrity of the tournament. Besides, it's something many fans have been asking for, ditto for players, and the fans are the ones handing ESL and others the $250,000 prize purses, so they should be inclined to at least listen. This is how you would do it.

You spend the first day of a tournament playing four GSL-style groups in a best-of-three format, two groups at a time. If you started at 10am, and used the RC EMS One Finals format - where each match begins as soon as the previous one ends, thus eliminating breaks between the action - you should have it finished by 5pm at latest, most likely sooner.

The second pair of groups would begin at 5pm - with two matches slightly later (as they couldn't setup until the previous groups were fully done) if the first pair's elimination games went the distance - and would finish around midnight. Sound like a problem? Ideally you'd be done a little earlier, but remember these are conservative estimates, and in reality the matches could, and should finish sooner - and even the last major, EMS One Katowice, had the second pair of groups projected to finish around 11pm; hardly optimal.

No one complained about the late finishes at last majors 

If we polled all the players participating in these events, no way they'd pick finishing an hour earlier, or starting two hours later on the first day, over getting to determine their fate in the tournament in a best-of-three format. The majors should aim to have the best teams meet in the grand final, not favor lucky upsets in best-of-one games, which is what can still happen in smaller tournaments. The majors should be all about finding the best team in Counter-Strike.

Still, there is a second way to do it too, but it would bring on all the old issues of the round robin group stage; three-way ties, round differences made on different maps determining who goes through, etc. The only pro it has in comparison to the GSL-style best-of-three format is that teams would get to play more, and especially in lopsided series, such as the ones at Copenhagen Games' group stage, two maps is too much.

You could organize two giant groups of eight teams, where every participant plays each other once. That's seven games minimum for every participant, and the top four teams would advance to the playoffs. These massive groups would provide less variance in best-of-one than the current format, and best-of-ones have that aura of potential upsets coming in at any point, thus keeping fan interest alive.

It would be more fun to watch teams play against a wider variety of teams, and it would also make certain series more interesting - no one wants to watch two easy wins in a match dragged out to be a best-of-three, Copenhagen Games groups style. Teams must be very even for best-of-three groups to work out, though it would largely be the case at the majors. This large round robin style group is a better solution for smaller events - but it would require eight team groups; otherwise GSL-format is better.

The first approach seems more believable, but don't count out the second option, which is basically a much better version of the old Intel Extreme Masters system, that saw group winners proceed directly to semi-finals. We can't have that, but we can have a new, better, improved version. In addition, eight team groups could also be used to make seedings for playoffs, because the sample size would be reasonably large, almost like a regular season of a kind. Let's hope they go for one of these two.

Vox would love playing best-of-three series in groups, or at least more best-of-one games 

Seeding and playoffs

ESL's Michal "Carmac" Blicharz has made it clear he doesn't think double elimination is suitable for spectator sports, so we're going ahead under the assumption that double elimination is off the table, whether or not that's the case. It would also prolong the tournament quite a bit, and with group stage being played in best-of-three a higher focus, and re-seeding available to battle unlucky bracket draws, double elimination isn't necessary.

While it's hard to argue against double elimination being more fun to watch, it has no place in the biggest tournaments due to time constraints alone. I think tournaments such as SLTV and ESEA Finals, and Copenhagen Games should definitely use it; in fact, playing the lower bracket in best-of-one made the Danish event one of my favorites in 2013, while this year's event was very lackluster in comparison and not nearly as interesting as a spectator. Also, double elimination best-of-three remains by far the best format for finals events with eight or less teams.

Now that we've settled on single elimination best-of-three for our majors, how do we make it better? What is stopping from certain teams drawing match-ups, such as Titan playing NiP in the quarter-finals, or Virtus.pro facing Na`Vi in the same round, while Epsilon and Lemondogs get to come out of nowhere based on zero wins over good teams in the group stage to guarantee one of them a top four placing?

Re-seeding. That's not to say we should completely ignore group stage results, but they ultimately should not matter as much. Winning a single best-of-three series, let alone a best-of-one in our current format, in the least important part of the tournament should not be as important as good results for months leading up to the major; up to five months in the case of ESL One Cologne.

Teams should still get punished for poor play in the group stage, but really what group stage is for top teams, is a minefield they're hoping to get out of alive. It's not where you separate the best from the pretty good; it's just a way to eliminate half the field before getting to the playoffs. And that's how it should be, too, especially in the best-of-one setting, but also in best-of-threes.

Random draw was OK in 2010; it should not be in 2014 

I suggest we seed teams in pools of 1-4, 5-8, 9-12 and 13-16 based on their past accomplishments before the tournament. We don't do that based on their result at the last major, or whether they got in through an online qualifier or were directly invited. It should be based solely on how well they have done in the past X months of relevant LAN play, with online play a tiebreaker at most. We would do random group draws instead of seedings, because otherwise the first round of playoffs would see constant rematches.

We then give ESL TV the chance to do their random group draw, drawing viewership and getting the community hyped. Great, we now have reasonable groups, and they will be played out in a best-of-three format, which means even if group a group of death á la group A of EMS One Katowice, with Virtus.pro, Titan and HellRaisers, surfaces, the best should get out.

We also seed the sixteen squads one by one, for proper seedings, only we don't make this list public so teams do not get to prepare too much for each other. HLTV.org would be happy to help with this part. Now traditionally you'd now place the teams in the bracket with #1 playing #8, #4 playing #5, #3 against #6 and #2 versus #7. Only that would completely ignore the massive group stage we just played, and its potential upsets.

What I suggest instead, is subtle changes in teams' seedings based on their group stage result. Once you've got the teams sorted out based on their pre-tournament seedings, you then reward teams who actually came in underdogs and won the group by bumping them up, say two spots, and move down the teams who were supposed to win but failed, by a spot or two. If you want, you could also add a rule about not being able to play the same from your group in the first round of playoffs.

That would make the bracket stage more consistent, favor long term results and encourage teams to stick together and build some tournament history, give credit where it's due to teams who did well in the now best-of-three group stage, and finally, make even the bracket drawing process a lot more interesting. Who wouldn't want to see that happen live?

On the other hand, the eight team round robin group wouild require a different solution. It would be super important to do proper seeding from one through sixteen to draw those groups, and then you should simply use the final standings as seedings. Group A would have seed one, so therefore the first place finishers in A would be #1, first place in B #2, and so on. It would be a pretty good system as well, and would take group stage results into account in their entirety.

Lady luck shouldn't send Titan packing in quarter-finals 

Schedule

What does this do to our event's schedule then? Well, ESL One Cologne is a four day event, similar to EMS One Katowice, whose schedule you can see here. In our imaginary world we're finishing the group stage on day one, and what do you know, the rest of the schedule can be kept from the previous event. As long as the group stage is played out on day one, there need not be any chances.

However, for events not hosted by ESL that aren't as limited in terms of scheduling due to so many other games being played, limited tournament area due to events being held at exhibitions, for the most part, and stage schedules, we could even make this better. At four day events we could split the group stage into two days, and still play playoffs one game at a time on days three and four.

Imagine now groups A and B are played on day one, with C and D on day two. On day three we'd play all four quarter-finals one at a time, starting at 10am or 12am, with one game every three hours. In an ideal world we would even use the RC EMS One Finals format, with non-stop action by having matches go live as soon as possible. We'd finish before midnight, which would be perfectly acceptable for most events - remember, even EMS One Katowice saw players finish past 11pm, and DreamHack Winter's fourth quarter-final and grand final both ran well past midnight.

We could have either a GSL-style best-of-three group stage, or a massive eight team group, guaranteeing everyone a minimum of seven matches at the majors. We would have a truly packed and exciting final day, with two semi-finals earlier in the day, one at a time of course, and the big grand final on the main stage later on in the day with North America also able to watch. All of that with more maps being played, thanks to a new map picking process, regardless of which one we choose. How does that sound to you, as a spectator?

Could ESL One Cologne be even better than the previous majors? 

Now let's for a second imagine Valve adds both de_season and de_cache and green lights these changes, and ESL makes them happen in mid-August for ESL One Cologne 2014 as we watch sixteen of the world's best teams fight for $250,000 under better, tournament integrity preserving circumstances. 

Who is with me?

Follow HLTV.org's @lurppis_ on Twitter.