EPTHomeStory Cup III was a milestone tournament in the ongoing “David and Goliath struggle” between western players and South Korea when HuK and Naniwa took first and second with wins over two time GSL Champion MC just a few days after HuK had captured the DreamHack crown in a field that included Bomber, July, MC, and Moon. It was the first glimmer of hope that the “Korean invasion” may not shut western players out of their own tournaments forever. In this respect, HomeStory Cup IV in January of 2012 stood in stark contrast. With several more representatives in attendance this time around, South Korean dominance was on full display. Korean players took six of final eight slots and nearly shut the westerners out in matches for the entire weekend. The only exceptions: Polish stalwart Nerchio winning his group with consecutive victories over HerO [2-1] and viOLet [2-0], and Delphi beating MarineKing [2-1] in one of the opening matches of the tournament.
Wait, Delphi who?
My point exactly. Delphi is a German Zerg, active in the online scene in Europe but yet (then, and still) to make his mark internationally.[1] This competition had a global player pool and audience but took place in Germany, which means a big chance for a local player like Delphi who wouldn’t have the chance to compete on a major stage month in and month out. A strong showing could be the kind of breakout performance that gets him invitations to future tournaments, access to sponsorships, and opportunities to compete around Europe or in North America. Unfortunately, he drew the only first round group with two Koreans, pitting him against MarineKing and viOLet along with compatriot HasuObs. Despite these odds, Delphi managed to win his match against MarineKing as well as his match against HasuObs. This outstanding performance was still insufficient, and Delphi was eliminated. Why? Look at how the group finished:
Group B MRO GSOAfter defeating Delphi, viOLet lost his series against MarineKing, tying all three players. Since each match was a BO3, HomeStory Cup broke ties by scoring each individual game. Delphi’s win over MarineKing went to a third map, while both MarineKing and viOLet swept the two series they each won. MarineKing advanced as first in the group thanks to the extra game win, viOLet in second without having an extra loss, and Delphi went home.
1.MarineKing 2-1 [5-3]
2.viOLet 2-1 [4-2]
3.
4.HasuObs 0-3 [0-6]
Realistically, Delphi’s chances in a sudden death would not have been great. I doubt anyone would argue that point. He had much less experience under pressure, and no longer had the advantage of being unknown or underestimated. But he earned his chance to prove us all wrong — that, no one can deny. Win or lose, he deserved to have a couple extra games penciled into the broadcast schedule, the asterisk on the results table, and the extra attention. Instead, his tournament was over. It ended with him on the sidelines, watching himself be eliminated in a match between two other people.[2]
Most tournaments in the international Starcraft scene rely on a round robin system in some capacity. Any format has its pros and cons, and one downside to round robin is the potential for ties. In an elimination tournament, there are no ties — two players face off, one advances, no room for disputes. When results of multiple games against multiple opponents are taken into account, however, two or more players may finish with the same score. This becomes more likely as group sizes increase and player skill converges. Since round robin is frequently used in groups of four and always used in the earlier stage of a tournament when the skill levels are most variable, ties do not occur often. When they do occur, few notice — less people watch the beginnings of a tournament and those that do watch the best players, the ones least likely to tie. Because of this, tie resolutions have stayed under the radar and the use of rules which award advancement to one player or another by default has quietly become standard without public outcry from players or fans. Despite no official collaboration on rules across the various tournament organizations, not a single major tournament has taken a stand against this practice and used playoff matches as a first resort. A possible exception is
GSL, which has steadily moved to only elimination formats in their various leagues, last using a round robin phase at the end of 2011.[3] Instead, tournaments rely on one, some, or all of the following criteria:- MRO (match results overall): match by match performance against other players in the group.
- MRH (match results head to head): your performance against another specific player or subset of players.
- GSO (game score or map score overall): since individual matches are generally comprised of a series of games, this is your game by game performance against all other players in the group.
- GSH (game score head to head): your game by game performance against another specific player or subset of players.
One can argue merits of all approaches — putting priority on head to head criteria (MRH/GSH) means an advancing player earned his spot with a victory over another, while putting priority on overall results (MRO/GSO) means an advancing player earned his spot by demonstrating a good performance over multiple trials. Giving weight to match results (MRO/MRH) puts significance in winning matches regardless of the details, while looking to game score (GSO/GSH) makes every single map in a series count. In theory, this seems like a simple difference of opinion with little relevance on the quality of a tournament or the results of the players. In practice, this isn’t the whole story. The competitors in these tournaments are real people, players struggling to get noticed or stay relevant in a scene that is both fiercely competitive and completely unrewarding of all but the crème de la crème. These players work hard to hone their abilities and deserve better than being told “sorry, not quite good enough” — especially, but not just, when prize money hangs in the balance.
The most common round robin implementation, as mentioned above, uses groups of four with two players advancing. If two players advance and two players are eliminated, there are three different ways a tie can occur and one is not likely to cause controversy.[4] As an example, here is a group from the third round of
DreamHack 2012 Stockholm Open:Group E MRO GSOIf the tournament uses group results to seed the next round, as DreamHack did, then MorroW's ranking above Protosser carries some value.[5] Being paired against a player with a weaker performance is certainly an advantage, but a better seed is the opposite side of the spectrum from outright elimination. If tiebreaker rulings must be used, using them like this to establish pairings is tolerable.
1.
2.
3.
4.
There are two other possible ties in four man groups. Both involve three of the four players in a rocks-paper-scissors relationship, contesting one or both of the advancing spots. Look at this group from
Assembly 2012 Winter:Group F MRO GSOTaeJa takes first place without contention, but we still have three equally matched players vying for the second spot. Not only did elfi, Jinro, and Adelscott split matches with one another, each match had the same result [2-1]. The deciding factor proved to be the series between elfi and TaeJa — while TaeJa swept both Jinro and Adelscott [2-0], he lost the second game in his win [2-1] over elfi. This earned elfi the second advancing spot.
1.
2.
3.
4.
The more common possibility[6] is when the three players tie for first place rather than last such as in this group, also from the third stage of the same DreamHack event:
Group D MRO GSOLiveZerg beat Cytoplasm, TitaN beat LiveZerg, and Cytoplasm beat TitaN. After eliminating Traveller there are still three players when only two can advance, so DreamHack used map score to break the tie. The key game was in the series between TitaN and LiveZerg — LiveZerg had managed to win the second map, giving him an extra win and first place while the extra loss eliminated TitaN.
1.
2.
3.
4.
In these situations, one or two players is simply eliminated on the spot, despite a marginal difference in performance compared to one that advances. The logic behind the decision, while known ahead of time, still leaves a bitter taste in the mouth of the player and his fans. Regardless of which method is used to break the tie, the player is eliminated based on old information — neither the player, his opponent, or the viewers knew explicitly during the match that this would, not could, mean elimination. Sometimes the information is so indirect and complex that explanations and deliberations are needed to sort everything out. All the tension and excitement that make these contests intense for the players and entertaining for the spectators is thrown away so tournaments can proceed with flow charts rather than competition.
Using larger group sizes not only do we see ties more often, but they can become more complicated as well. They cease to be an occasional annoyance and almost become expected. Let’s take a look at what happened in a few tournaments which used this type of format, starting with the
GSL 2011 Blizzard Cup:Group A MRO | Group B MROThe rules stated that the top three players would advance to a single elimination bracket, with the first player receiving a bye. In Group A, this gives us two two-way ties worth consideration — DongRaeGu and MVP both advance, but being awarded first place carries quite a bit more value than just preferential seeding. MC and Stephano, on the other hand, simply want to stay alive. GSL deferred to the head to head results, giving DongRaeGu first place due to his win over MVP and MC third place due to his win over Stephano. At first blush this seems perfectly reasonable, but is it? Due to the larger group, the results of the players tell more of a story. MC and Stephano both lost to MVP while beating HerO, but Stephano won against DongRaeGu when MC couldn’t. Stephano will not proceed despite earning a win over not just an advancing player, but a player granted a spot in the semifinals. MC, on the other hand, advances with only losses against two players left in the tournament and no results against the other three. Along those lines, DongRaeGu had indeed beaten MVP but took a loss to Stephano, a player who fared poorly and whose performance was deemed too weak to continue. Should these results grant him such a valuable placing without question?
1.
2.
3.
4.
5.
There was also a tie in Group B. While all three players would advance, first place would get the bye into the semifinals and second place would get a better seed in the bracket. MMA beat Leenock, Leenock beat Polt, and Polt beat MMA. Since each match was a single game, GSL did not have the option of using map score and instead went to a playoff. MMA beat both players for first place, and Leenock beat Polt to take second.[7]
IEM Season VI World Championship had six players per group and plenty of chaos. Out of four groups total, two had ties:Group A MRO GSO | Group B MRO GSOIn Group A, viOLet, JYP, and Socke have a three way tie contesting the two spots to advance below MMA. This tournament used overall map score as a tiebreaker, so despite the rocks-paper-scissor relationship between the three, Socke was sent home by a razor-thin margin due to his game losses in matches he won against both KiLLeR and Strelok. Group B gave us a four way tie and is perhaps the best demonstration of how foolish breaking ties without extra games can be. With three [2-0] wins and no shutout losses MC took first place with elfi close behind in second. HasuObs and Nerchio had an identical map score, so HasuObs's match victory from their meeting gave him third. Look at the disparity here — despite very similar showings of 3-2, Nerchio is eliminated in 13th-16th place and wins $700 while MC is given a bye into the top 8, guaranteed $2,000, and is just three matches away from the gold.[8]
1.
2.
3.
4.
5.
6.
In the structure of
MLG 2012 Winter Championship no group participant is eliminated outright, but each rung above sixth place until second corresponds to a one round bye in the championship bracket.[9] While being penalized a round or two is not nearly as cruel as elimination, a single bye in a competitive bracket has a significant effect on a player's potential in a tournament. In four six man groups there were five ties!Group A MRO GSO | Group B MRO GSOThere’s a three way tie for 3/4/5 in Group A, a two way tie for 2/3 and 5/6 in Group B, a three way tie for 4/5/6 in Group C, and a three way tie for 2/3/4 in Group D. Thirteen of twenty four players had their eventual placing, prize money, and ever valuable circuit points strongly affected by the nuance of the rule-set. That's more than half. At what point does it become ridiculous?
1.
2.
3.
4.
5.
6.
---------------------------------------------------
Group C MRO GSO | Group D MRO GSO
1.
2.
3.
4.
5.
6.
Let’s shift our focus to the first
MLG Spring Arena. MLG used a novel system, a single group round robin into a knockout bracket for its eight man invitational. Not only does an eight man group mean many more matches, an invitational based on merit suggests that the players are closely matched in skill level. This shows in the pool results:Group MRO GSO
1.
2.
3.
4.
5.
6.
7.
8.
MarineKing undoubtedly marked first place as his own by finishing with six wins when no other player even managed five, while HuK clearly took last place with only a single win. GanZi, viOLet, and MC all finished next in a three way tie with Heart, DongRaeGu, and PartinG in another tie below them. MLG rules allowed for six players to pass into the next round — after giving MarineKing first place and HuK eighth, this left six players and only five spots. GanZi, viOLet, and MC were guaranteed safe passage, but MLG rules also granted a first round bye to the top two finishers, as well as choice of first round match to the third place finisher and choice of second round match to the first place finisher. This means MLG would use their tiebreaker criteria to award one player a bye (2nd), award a second player first round match choice (3rd), and eliminate a third (7th). All in a tournament with only eight players total! This is a lot of factors left up to the stroke of a pen rather than the click of a mouse.
Take a step back to realize exactly what’s at stake for each group of players. Second place in the pool would receive an automatic bye into the semifinals for this tournament. This guarantees them at least $1,450 with a better chance at winning first or second prize worth $5,000 or $3,000 respectively. Furthermore, the semifinalists in the tournament receive a travel-paid invitation to participate in
Spring Arena 2 as well as the Spring Championship, both of which have larger prizes and more prestige. While finishing 3rd and 4th in the group stage does not preclude a player from the semifinals, he will have to win one more match against one of the best players in the world to make it there. The player finishing third place gets the added bonus of making his choice of opponent between the players finishing fifth and sixth, theoretically increasing his chances of winning the critical first round.[10]Things are much less sunny for the trio of Heart, DongRaeGu, and PartinG. The player slotted in seventh place is automatically relegated to the qualifiers for the next Spring Arena — a tough online tournament with little exposure and very difficult to pass through.[11] While the fifth and sixth players are still in danger of relegation, they will receive one more shot — a single match to claim the majority of the bounty at stake in the entire tournament, circumvent the qualifiers, and be within arm’s reach of the championship trophy.
How did this play out? Had MLG used GSO as the primary tiebreaker, four of the six players would be resolved. They didn’t. Their priority went MRH, then GSH, then GSO, and then, only then, playoff matches. So you can follow along, here is the match information:
Tie for 2/3/4GanZi’s wins over both viOLet and MC secured him second place. viOLet beat MC in their game, ranking him third and MC fourth — all according to MRH. As it happens, the rankings come out the same regardless of which of the three criterion is used first — GSO is listed in the first set of standings and GSH adds up to GanZi [5-1], viOLet [3-3], and MC [1-5].
GanZi vs. vioLet 2-1
GanZi vs. MC 3-0
viOLet vs. MC 2-1
Tie for 5/6/7
Heart vs. PartinG 3-0
Heart vs. DongRaeGu 1-2
DongRaeGu vs. PartinG 1-2
Our next tie was not as cut and dry, even for the referees. There was a delay in announcing the finalists while tournament admins reconsidered the standings and the rules they themselves had written — perhaps realizing the error of their ways when faced with a real world example. PartinG had beaten DongRaeGu, Heart had beaten PartinG, and DongRaeGu had beaten Heart. Unlike the previous example, MRH settled nothing. Moving on to GSH, PartinG’s loss to Heart was [0-3] while the other two matches went [1-2]. This ruling gave Heart and DongRaeGu passage to the next round and sent PartinG home. Had MLG used GSO before GSH, Heart would have broken out of the tie and taken fifth place. PartinG and DongRaeGu would remain tied with a GSO of 9-12, but with PartinG having won the match between them during the round robin, he would have taken sixth place instead.
DongRaeGu advanced on the back of Heart’s performance, not his own. Both he and PartinG had been shutout by one other player in the group. DongRaeGu’s [0-3] loss to MC was not factored into the tiebreaker, because MC was not part of the tie. Heart was, so PartinG’s [0-3] loss in their match cost him the tournament. Some looked at this situation and cried foul — PartinG had beaten DongRaeGu, but was eliminated, while DongRaeGu received a second chance he ultimately used to win the entire tournament. I don’t find it to be unfair to PartinG so much as it is peculiar. If PartinG traded his [0-3] loss to Heart with a [1-2] loss to MC, viOLet, or MarineKing, he would have advanced while the eventual winner of the tournament finished in 7th place. DongRaeGu’s exceptional performance in the final bracket is compelling evidence that he deserved to be champion, but the slightest variation in the rules would have denied him the chance to even try. On the other hand, we have no idea what PartinG was capable of. To implement rules that generates this sort of situation is unnecessary and harms the integrity of the entire tournament.
Looking at this web of results, we could cite all sorts of supposed injustice. MC holds the only win in the group over MarineKing, making him arguably the most fit to face defending champion and top seed MarineKing in a championship match, but the tiebreaker rules cost him both the bye into the semi-finals and match choice. Of the other tied players, PartinG had the only win over GanZi, second place finisher and automatic semi-finalist. While the head to head results tell us how two players lie in relation to each other, a tournament is an attempt to give us a relative ranking of many players all together. In this light, small errors further down the rankings such as transposing two players who will always finish between 5th and 7th is trivial compared to the catastrophe of eliminating a potential champion much too early. If MarineKing and GanZi finish first and second, it could be argued that a win against one of those two players should be worth more when choosing which mediocre performing player(s) should continue and which should not, since those wins suggest a higher likelihood for winning the entire tournament. A tournament could even choose to move past immediate results all together and use initial seeding,[12] results from previous tournaments, ladder ranking, or simulate matches with a rating system.
Valid or not, these arguments are all just nitpicking. I bring them up not to lobby for more complicated tiebreaker standards, but to show that all methods are sorely lacking in integrity. The goal of a tournament for the competitors is to put their skills to the test, under pressure, for all the marbles. The goal for the spectators is to see exciting matches and cheer on their favorite players in battle. By relying on a multitude of criteria, any criteria, rather than playoff matches, tournaments rob the players of an opportunity to compete knowing what exactly is on the table, and fans are robbed of the nail-biting suspense of watching their favorite player try to rise from near elimination. How can anyone be happy when it becomes clear the deciding match in a one player’s tournament happened hours earlier, without anyone realizing it? The solution to the problem isn’t to argue over methods slightly more fair than flipping a coin, the solution is to default to the notion that Starcraft tournaments should settle these issues with Starcraft, not arithmetic.
The rationale for having these rules in the first place likely comes from the administrative side of the tournament. It’s plenty difficult behind the scenes to keep everything from crumbling down without the added stress of falling behind schedule as you try to jam extra games into the itinerary. But considering what is at stake, is this valid? These tournaments consist of a very large number of matches — the group stage alone for MLG Spring Arena 1 guaranteed 84 games, while the tiered format used by DreamHack in Stockholm gave us over 300 before the RO16. Is adding a few sudden death matches such a logistical problem that it must be avoided whenever possible, even when this means eliminating a player who has earned consideration? Conventional sports all implement some sort of overtime, in spite of being aired on television and operating on much stricter schedules, and any sports fan will agree that overtime games are among the most memorable. A tournament is in the business of broadcasting games people want to watch. Why are they all so quick to avoid a match with so much on the line in favor of eliminating players based on technicalities?
Notes
1. As of four months after HomeStory Cup IV, the only other international offline event Delphi has had the opportunity to enter is the DreamHack Stockholm Open. For players who aren’t invited with travel paid to every event on the circuit, big chances don’t come up very often.^
2. Chronologically, the last match of the group was viOLet vs. MarineKing. Going into the match, MarineKing was in a must-win situation. If he lost, Delphi would advance to the second group stage instead. viOLet would move on regardless of the outcome, but held Delphi's fate in his hands. viOLet swiftly lost two straight. When they met again in the quarterfinals of the same tournament that same weekend, viOLet won the new series [3-2].^
3. For 2011 Code S seasons in January, March, and May, GSL used a psuedo round-robin system that cancelled matches between players out of contention in their group stages. Starting in July and since, they have used an elimination format instead. They also used round-robin in 2011’s Arena of Legends 1, Arena of Legends: Team Ace, and Blizzard Cup (all in 2011). Since Blizzard Cup they have hosted only one invitational, Arena of Legends: King of Kongs, and did not use round robin. While nothing is set in stone, it gives the impression that GSL prefers other formats.^
4. While there was significant controversy over DreamHack not awarding first place in the example group to TLO in consideration of his awesome beard, wildcard spots based on facial hair is beyond the scope of this article. ^
5. In this particular tournament, second place matched Protosser against Polt, the winner of Group C, while MorroW went on to face Group C runner up Socke. ^
6. More common based on history, not probability. In a simulation, both three way ties are equally probable. It just doesn’t seem to happen that way. ^
7. I don’t recall hearing any viewers complain about the extra games. ^
8. MC went on to win the entire tournament, worth $35,000 USD. Had he not received first place in his group, he would have not only needed to win four matches rather than three, but he would have played either a slightly or entirely different sequence of players in the playoffs. ^
9. As an example from the same tournament, Socke won four matches in a row after the group stage before he was HuK's first opponent in the elimination bracket. Finishing four rungs higher gave HuK four byes. First place in the group is slotted into a different (and more advantageous) bracket rather than being given one extra bye. ^
10. Getting to choose the pairings is only an advantage if a player is capable of identifying their best option. viOLet chose Heart. He lost, but we can't know whether he would have fared any better against DongRaeGu, his only other choice. In the next round, MarineKing also chose to face Heart over DongRaeGu. While his victory over Heart and subsequent loss to DongRaeGu suggests that MarineKing’s decision meant finishing in second place rather than losing one round prior, I can’t help but point out that the semifinals were BO5 and the finals were BO7. Before losing the series in the seventh game, MarineKing was leading the championship [3-2] — choosing to play DongRaeGu earlier in a shorter series may have won him the tournament. The only thing we know for sure is that Heart must not spend a lot of Friday nights alone. ^
11.The South Korean qualifiers for MLG Winter Arena was a veritable gauntlet of some of the strongest competition in the world. 52 players — 24 of whom started that GSL season in Code S and 11 in Code A — in a double elimination bracket with just eight advancing. Even a very talented player with a marked advantage over the opposition could easily stumble over so many matches against such skilled opponents. ^
12. Though it didn't come into practice, the rules for Spring Arena do in fact use seeding should they be unable to avoid extra games (see Round Robin Format, #3). According to these rules, had the three way tie gone to overtime, DongRaeGu (seeded #2) would have received fifth place automatically, with the winner of a BO1 match between Heart (seeded #3) and PartinG (seeded #5/6) advancing in sixth and the loser eliminated. ^
All tournament results obtained from Liquipedia.
Player links goto the TLPD database appropriate for the tournament.
Thanks to primadog for encouraging me to present this essay here and spending quite a bit of time showing me how to format the text for TL, especially since he added many pictures which I stubbornly and ungratefully removed.
