Experiments in Guild Wars GvG: The Autumn Ladder Season Playoffs
During the weekend of October 21-22, 2006, we ran the playoffs that ended the Autumn Ladder Season. But this time we did things a little differently than ever before. Instead of the normal single elimination tournament, we ran four rounds of Swiss play with pre-set maps, and then cut to the top four guilds for standard, single-elimination semi-finals.
Why did we do this? What exactly do we mean by Swiss play? What does "opponent's match-win percentage" mean?
These are all good questions we'd like to answer.
First, the easy answer. We did this to test out new ways of running high level Guild Wars competition, because we are always looking for ways to improve GvG and other forms of PvP.
Our Current Tournament System
The current ladder system has been in use since Guild Wars launched, and while it provides a fun way for guilds to keep track of their GvG results, it doesn't provide the truest measure of a guild's strength. Without going into a long discussion on this topic alone, there are too many ways to gain ELO points under the current system that don't necessarily translate to a true strength rating for each guild. This is why we have ended most seasons with a tournament to allow the most skilled guilds their opportunity to shine.
Single elimination tournaments, which we have always previously used, have the advantage of clearly showing who has won and moved on because the losers are eliminated from the tournament. Each round has been run as a best two wins out of three battles to provide each guild a chance to excel on their chosen map and to let them adjust builds between rounds. This can make each match exciting for the audience to watch on observer mode because it shows the build and map strategy taking place between the two competitors, as well as the tactics of each battle.
Unfortunately, this also puts a lot of pressure on the results of each of these battles (lose twice and you are done) and puts a large amount of importance on pairings (a bad matchup for a guild in the first round can mean an early exit).
Let us explain. In the metagame of Guild Wars, there are several types of builds primarily used (spike, balanced, Condition overload, etc.) as well as various map and tactical choices that guilds employ. Because there are so many ways to play, most guilds tend to specialize, becoming stronger in some areas while accepting their limitations in others (though any good guild must have a deep understanding of the current metagame and build with that in mind).
This can mean that Guild A is inherently more likely to defeat Guild B if Guild B customarily employs strategies and tactics that are weak against Guild A's strength. And Guild A would have an increased chance of defeat at the hands of Guild C if that guild usually plays with strengths that Guild A is weak against. And so on. This sort of rock-paper-scissors (dynamite-horses-chocolate pie, etc.) circle is one of the true strengths of Guilds Wars PvP, because it means there are many ways to play and win.
But, even though every high level guild should practice as many strategies and tactics as possible, some guilds will always have a starting advantage over others because nobody and no guild can be good at everything. So, some guilds will gain this advantage in the first round of a tournament while others lose it, based entirely upon the first round pairing.
In a single elimination event, if you lose, you are done! Thus, this potential first round advantage takes on more importance than was originally intended.
In addition, single elimination tournaments only allow a losing guild one chance to test out their skills against another high level guild. Not only do extra rounds provide more chances to win, they provide more time to play against the best competition in the field. For a new guild just starting to delve into advanced play, it can be disappointing to have to wait until the end of the next season to play in a high-level, public event again. It can also put an inordinate amount of pressure on new players to do well immediately, and has even led to some guilds breaking up. This result is antithetical to the long term health of competitive Guild Wars, which we want to encourage instead of discourage.
So, instead of jumping right from the uncertain ladder to the pressure of single elimination, we added rounds of Swiss play. This allows guilds to compete in meaningful matches without the results of a single round's pairing having such a huge effect on the final outcome, or the pressure of knowing that losing one match can end their playoff experience.
Swiss Tournament System
So, now that you know why we tried this grand experiment, let us tell you what the Swiss tournament system is all about. Swiss tournaments allow guilds to compete in a set number of rounds, regardless of how well they do. Thus, they get to pit themselves against several different opponents in a high-level, public competition and form a better idea of how their strategic choices work against a wider variety of challengers.
The number of rounds run is the base two logarithm of the number of competing teams rounded up. This means two rounds of Swiss play can handle up to four guilds, three rounds up to eight, and for 16 guilds, we run four rounds of Swiss.
After the first round, each subsequent round pairs guilds against other guilds with the same (or almost the same) match win record. For the second round, the eight winning guilds from the first round are randomly paired against each other, and the eight losing guilds are randomly paired against each other.
This system helps make each match more compelling for the guilds involved as they are playing against opponents who have done as well as they have. Instead of potentially steamrolling over an inferior opponent (or getting steamrolled by a superior one) every guild gets the challenge of more balanced competition.
At the end of the Swiss rounds, we have a clear picture of the ordering of the field from top to bottom. For a field of sixteen, the Swiss rounds should produce four competing guilds that are clearly the strongest out of the initial sixteen on that day. We then cut to these top four guilds for the more standard single-elimination rounds.
So, how do we determine which four guilds make the cut at the end of the Swiss rounds? After all, at the end of four rounds, we would have one guild with a 4-0 record and the possibility of four with a 3-1 record, making a total of five guilds. Obviously the undefeated guild moves on. But the other four have identical win/loss records. Here, we apply a test using opponent's match-win percentage as a tie-breaker. Hold on, this requires a little math.
Opponent's Match-Win Percentage
Under the Swiss scoring system we tested, teams earn 3 match points for each win, 1 point for any draw, and 0 points for a loss.
If a battle had not been completed after 45 minutes passed on the match clock, that battle would be called a loss for both guilds, each receiving 0 match points.
Each guild's match-win percentage is that guild's accumulated match points, divided by 3 times the number of rounds in which they competed, or 0.33, whichever is greater. (Establishing a minimum match-win percentage of 0.33 limits the effect low performances have when calculating and comparing opponent's match-win percentages).
- For example, if a guild's record is 2-1-1 (2 wins, 1 loss, and 1 draw) after 4 rounds of Swiss play. It has earned 7 match points, and its match win percentage is 7/(4*3) = .583
- A guild's opponent's match-win percentage is the average match-win percentage of each opponent that guild faced (ignoring any rounds for which the guild received a bye).
- For example, using the above example, let's say the guild that finished 2-1-1 played against opponents whose match records were 2-2-0, 1-2-1, 3-1-0, and 0-4-0. Therefore, the guild's opponent's match-win percentage is:
(6/12 + 4/12 + 9/12 + 0/12)
Replacing the 0/12 with 0.33 (see above minimum) and translating to the decimal system we get:
0.50 + 0.33 + 0.67 + 0.33)
This equation becomes 1.83/4, or 0.46.
While this may seem a bit complicated, it basically comes down to using the strength of your schedule as the tie-breaker. The guild that defeats guilds with better records is rated higher in the tie-breakers than a guild that defeated weaker opponents.
Swiss tournaments also have the advantage of taking less time (as every round starts at the same time) and can allow many more guilds to compete in larger events (which we will be working toward in the future).
One other change we incorporated for this experiment was to use specified maps for each round of Swiss play. Since each match consisted of only one battle (to keep the overall length of the event reasonable), we wanted to give each guild as much information as possible in advance to balance the competitive field. While allowing a higher ranked guild to make the map selection works in best two-out-of-three battle matches (because each guild gains this tactical advantage at least once), in a single battle match, that advantage would have been too much.
Instead, we selected four diverse GvG maps and announced well in advance which maps would be used in each round. This allowed competing guilds time to practice on those maps, producing a more level playing field.
Now you know more about the rules variations we tested during the recent Autumn Season. The initial feedback has been positive, but we still have plenty more to test and to learn. Look for the rules of the Winterfest season event to incorporate some of these potential changes, as well as more innovations, such as larger numbers of participants, a wider spread in the prize pool, and others.
We are always trying to move Guild Wars to the fore of competitive e-sports and are open to your ideas and suggestions. Please write to firstname.lastname@example.org with your comments.
ArenaNet Tournament Coordinator