EPTMLG's bracket sucks. OSL prelims suck. GSL needs to take more time for each season, yo. MSL groups are dumb. NASL wastes time. All complaints we've heard before, right? Every tournament on the face of the earth has problems - and everybody thinks they know what to do to fix it.
Well, I'm not a professional tournament organizer (though I have run tournaments before, for battlereports.com, SC2GG, and the Hillsdale College Chess Club). What I am is someone who analyzes things - all sorts of things - for fun, so I'm here to introduce you to all the things you have to decide when running a tournament. We're going to start by examining what's common to every tournament ever - or as mathematicians and logicians say, let's define our terms.
Any tournament or even as a whole we are going to call a League. (Major League Gaming, OnGameNet Starleague, IGN ProLeague.)
Every league consists of one to several stages.
Each stage is characterized by using a specific format. A format is a method for deciding which players move on. Common formats are single elimination, where winners move on; double elimination, which is like single elim but losers get a second chance; round robin, in which all players play each other; and Swiss, where you play someone with similar results as often as possible.
Due to the size of most tournaments, most early stages are often broken down into group play, where a limited number of players advance to the next stage.
Results of any given stage's groups are usually used for seeding, where some players are given a chance to skip preliminary stages (moving "up" in the league); and relegation, where players are required to start over, or move to a less-prestigious ranking (moving "down" in the league).
Any given stage can be considered either a playoff stage or a scheduled stage. A good rule of thumb is if losing a single match in a stage eliminates a player, it is a playoff stage; if elimination is only done on some total of results, it is a scheduled stage. (Exception: Double-elimination has to be considered a playoff stage, even though it takes two games to get eliminated.)
Many stages will have multiple rounds: in some formats each round will bring eliminations while in others the eliminations may be postponed.
For an example, let's evaluate the Bacchus 2010 OSL's format.
OSL Preliminary: A single-elimination playoff stage in twenty-five groups, open to any licensed progamers (except seeds from previous tournaments).
OSL Round of 36: A single-elimination playoff stage in twelve groups. 24 winners from prelims advanced to here; 12 group losers from the previous tournament were seeded in a second round. 12 winners advanced to groups, while the remaining 24 players were relegated to prelims.
OSL groups: A round-robin scheduled stage in four groups. The top two players in each group advance, while the losers are relegated to the Round of 36 (or in this case the ODT).
OSL bracket: A single elimination playoff stage of eight players. The top four players are seeded into the next OSL groups, while the other four are relegated to the Round of 36 (or ODT).
It is also instructive to look at the elimination ratios at each stage. In general it is obvious that the higher the proportion of eliminations, the relatively better are the remaining players. (This is the fundamental reason preliminaries eliminate players so fast: the average professional player simply isn't a very good player by the standards of absolute excellence). At the same time, it is best to give as many very good players as possible to best chances of continuing. This calls for balance.
Preliminaries: 132:25, or approximately 5:1
Round of 36: 3:1
Group Stage (16): 2:1
Bracket Stage/Final Playoff: 8:1
In general, 3:1 seems to be an ideal elimination ratio when moving from one stage to another, until we reach a point at which the number is small enough to require a change to 2:1. Anything less than 2:1 would be essentially meaningless. (Consider the apparent waste of time of the MLG group stage, which is 1:1, for example, as it eliminates no one and is used for seeding only. The only thing to be said for it is that every game "matters" as it means less of a gauntlet later.)
To support my 3:1 contention I would like to compare the playoff pictures of the Big Four in American sports. The NFL has two stages, a schedule open to its 32 teams after which the top 12 are taken to the next stage. This ratio is 8:3, or not quite 3:1. Normally there are approximately that many teams that "deserve" to progress further (although the selection process sometimes forces selection of a bad team instead of a good one). The NFL playoffs are probably the most-watched four weeks of television in the American year. MLB has a schedule open to its 30 teams, from which 8 progress to the playoff stage. This ratio is 15:4, or almost 4:1, and is mandated by the logistics of series play. Almost always the 8 teams selected are good. An expansion to 10 teams is plausible in terms of quality but would add two weeks at least to an already over-extended schedule.
On the other hand, the NHL has 30 teams and sends 16 to the playoff stage: as a result, not only is the playoff stage extended into June (which is silly, when you think about it), the first rounds are almost always bad. Similarly, the NBA has 30 teams and sends 16 to the playoffs: not only are the resulting games bad, and playoffs over-extended, teams make the playoffs with losing records, every year.
Coming back to Starcraft, the Shinhan Proleague last year had 12 teams in the schedule, and advanced 6 to the playoffs. The last few were not very good. This year, with 10 teams on the schedule, the league has 6 advancing: seed six, and maybe five as well, is probably going to be taken by a team with a losing record. Which is still ridiculous.
Finally, we have to consider the issue of timing. It is evident that prestige increases with uniqueness. The Proleague has grown in importance as its length has increased and the schedule has stabilized to a yearly one. MLG holds events every two months or so, within the structure of a yearly season. In contrast, the Korean individual leagues are a little odd. The MSL and OSL seasons come in, on average, at just under 3 a year, and occasionally that feels rushed. The GSL, in ten months, has held 7 different seasons - 3 Opens, 3 Code S/A tournaments, and the GSL Super Tournament. That's ridiculous, and doesn't include the team leagues, which were even more farcical, so far.
From all of this, I am going to establish a somewhat arbitrary rule of thumb: play as few seasons as reasonably possible per year, and keep a regular schedule to it. (GOM hasn't learned: they're trying to cram another tournament into a single month, and run a 12-team league in about three months, including playoffs. I've found the switch amazing, considering these are the same people who ran 3 GOM Classic seasons in 2 years. But I digress.)
Last, we need to consider who gets to play. Even tournaments with "Open" qualifiers usually set restrictions - B+ or Masters or similar. Leagues with open qualifiers also almost always have a maximum bracket size, for obvious reasons. TLPD gives 100 Korean players with an ELO rating; there are 256 ELO-rated "International" players, which includes a number of Koreans. For the sake of argument I would say that an ideal individual league pool can't start with more than about 150 (compare the OSL prelim). The GSL has 64 players at any given time: the prelims have included another 20-30 on various teams. MLG of course starts with its huge 256-man open bracket and 16 pooled players for a total of 272 - but we all know that a lot of the open players are, how shall we say, bad.
To me, the GSL is actually leading innovation right now - except in their "we're a minor league trying to get attention by constant hype" scheduling - and their most important innovation in ESPORTS is the introduction to the Starcraft community of an actual minor league. Further, the GSL is the only league I know of that actively practices relegation: players are given the chance to move up and down as they or others improve, and even to move in from outside the tournament. The whole thing is illustrated here:
![[image loading]](http://wiki.teamliquid.net/starcraft/images2/2/26/GSL_Sponsorship_League_format_2011_March.png)
So what I am going to do is take the GSL as a model and, throwing in all the things I've been talking about for pages, improve on it as I design my own ideal league.
The GSL Code S begins with a scheduled stage. This is excellent - and it should be expanded. I would like to put 48 players in the top group. I would have eight groups of six each. These groups would play round-robin: the top two in each group advanced to a playoff, the middle two retain status. The bottom finisher in the group is transferred to the minor league, which is formatted almost exactly the same way. The 5th place plays a minor league group's 2nd finisher, while each #1 in the minor league group advances to the major league. In addition, the eight group #1s are put in a minor league championship playoff. The sixteen players at the bottom of the minor league are relegated as seeds to a tournament open to qualified players (on the Korean scene, this would likely be players from qualified teams: in the US my hypothetical league might have a similar relationship with teams, ideally). Matches would be Bo3. Placement within groups would be on record, with game score as a tie-breaker.
After that, the 16-man major league playoff would be standard. For the playoff, the bracket would be done so A#1 plays B#2, B#1 plays C#2, etc. Bracket is designed so no players from the same group can meet before the final:
A1
B2|
|---|
C1| |
D2 |
|---|
E1 | |
F2| | |
|---| |
G1| |
H2 |
|_____
|
F1 |
G2| |
|---| |
D1| | |
E2 | |
|---|
B1 |
C2| |
|---|
H1|
A2
Playoff games would be all Bo5 with a Bo7 final.
The next season groups would be re-seeded by overall records as follows: #1 finishers stay in the group. Ties here are decided by playoff results. Players advancing from the relegation matches are seeded by record in reverse, with any advancing minor leaguers automatically getting lower seeds than returning major leaguers. #2 group finishers are then seeded in so #16 (overall) is in the #1 group, #15 in the #2, etc. Automatic minor league qualifiers are seeded as the 33-40 seeds. So the regrouping for the next season would be:
1 2 3 4 5 6 7 8
16 15 14 13 12 11 10 9
17 18 19 20 21 22 23 24
32 31 30 29 28 27 26 25
33* 34* 35* 36* 37* 38* 39* 40*
[Minor League Advancing]-----------[Major League Returning]
* Minor League #1 Advancing
Minor league reseeding would be similar. Demoted major league players receive the 1-8 seeds, while losers of the relegation matches would get the 9-16. Winners of the qualifier (16) would get the minor league's 32-48 seeds.
Any players resigning from the league would be replaced as follows:
Major league player: The top minor league player remaining is promoted, seeding adjusted accordingly. Repeat as necessary.
Minor league player: Minor league seeding is adjusted, extra spot is added to the qualifiers.
The scheduling could work as follows:
3 matches a day of Major league, 3 of Minor (cast at a less prime-time position?), 4 days a week. (Thursday, Friday, Saturday, Sunday?) This would take 10 weeks to work through the round-robin stage. (Alternatively you could move to 6 and 6 and take five weeks, or play every day at 3 and 3 and take three weeks (20 days) total. You can come up with your own system. Given the tendency for replay-casting and online leagues that SC2 has given us, the game scheduling almost wouldn't matter, though of course we all want live casts eventually.)
The minor league 8-man playoff would take 3 weeks. The final would be the week of the major league semis, but obviously happen on a different day (or as a preview?)
At some point, I would say the weekend after the major league final, the relegation matches would be played, probably over the course of two days. Simply Bo5/Bo7 matches of #5 major seeds vs #2 minor seeds.
So: 10 weeks schedule + 4 weeks playoff + 1 week relegation is a 15 week season, or about 3 1/2 months. The schedule could actually be extended, especially to work around other leagues. I would be planning to run this (hypothetical) league in a twice-yearly season.
Money: Obviously this league requires a decent commitment from players, and envisions a scenario with around a hundred players with stable careers. I believe the esports scene requires team leagues for stability, so I'd assume that most players would be getting the majority of money from a team salary. At the same time, in addition to the obvious playoff pay-outs, I think it would be ideal for an extensive league like this to pay per game, $50-$100 per win (or similar), with a possible small payout for a loss, for time spent etc.
League Format Thoughts: The basic concept can be used with smaller numbers of players as need requires. 48 per division (major/minor) is a handy number to work with, as I'll explain. (24 is also possible for 48 total.) 36, 18, or 12 could work but would require either smaller groups or staggered (think NFL; if you don't know what that is, think OSL Ro36 and you'll get the basic idea) playoffs, or both. Any division size has to be divisible by 6 in order to get both the 3-tiered playoff/retention/relegation structure and a plausible playoff. Any division size not divisible by 8 will result in a staggered playoff.
This could also be used - although without the minor league qualifier, probably - a team league format, in which case you're looking at probably around 24 teams, 36 max. Without a qualifier, it would be possible to have a two-tiered minor division, so for instance a 12 team major and 8 team minor setup.
