Information Cascades in Esports
Every year Valve Corporation Hosts the “World Cup” for their most popular video game called Dota 2. This tournament is known as “The International ” and is generally hosted in July or August every year since 2011 when the Inaugural International took place. This past July the tournament was held in Seattle’s Key Arena with over 10,000 live spectators and a price pool of nearly 11 Million Dollars, 5 of which went to the first place team. The stakes are high and that create top level play and draws top level players from all over the world to compete for a chance to be the best in the world, and a millionaire professional gamer all at the same time.
Dota 2 is a 5 versus 5 multiplayer game in which teams select heroes from a pool of ~110 possible choices in a draft style selection at the start of each game. Each team’s captain can ban heroes they don’t wish the other team to have and select heroes that they want to their team to play. The heroes that are currently popular in the pro Dota 2 Scene are considered part of the metagame. The Metagame is, simply put, the trends of what strategies and heroes players are tending to use at the current time. The commonly used measure of a hero’s popularity is their pick/ban rate. This is the number of professional games in which the hero was either picked or banned divided by the total number of professional games played. There are two main ways that a hero’s pick/ban rate changes significantly. The first is that everyone 4 months or so the game’s designer (known only by his username “Icefrog”) changes the abilities of heroes in order to make the game more fair. When this happens heroes that are buffed (abilities and powers made stronger) tend to have increased pick/ban rates. The other reason that heroes get picked/banned more is simply that other teams are picking/banning that hero despite no changes to the balance of the game occurring near that rise in popularity. This behaves similarly to the Information Cascades as discussed in class. To illustrate this phenomenon data from three heroes will be analyzed before and during their rise to popularity during the International 2014.
The Three most picked heroes at the International 2014 were named Skywrath Mage, Razor, and Shadow Shaman. The last balance change in Dota 2 before the International 2014 (which started on July 8th and concluded on July 21st) was on March 8th 2014. From March to June Skywrath Mage, Razor, and Shadow Shaman has 3%, 14% and 30% pick/ban rates. Less than 10% means the hero is almost never picked, 30% mean it’s sometimes picked but not often. The truly popular heroes have 70%+ pick/ban rates. So when The International 2014 begins all of the sudden in July/August Razor has a 94% pick/ban rate, Skywrath Mage has a 75% pick/ban rate, and Shadow Shaman has a 70% pick/ban rate. They are the three most picked heroes in the entire tournament, but from a balance standpoint nothing changed about these heroes as the tournament started. What did change is that teams started picking them. At first teams that used them well and won games picked the heroes and all of the sudden everyone was picking an banning them because they were now considered good within the current metagame. This is almost exactly like an information cascade. People are selecting these heroes because other people selected these heroes and ignoring the data they have which says that they weren’t picked after being changed last, and haven’t been changed since then. Based on that information it seems probable that behavior similar to an information cascade affects the drafting minds of captains in Dota 2.
The last thing to analyze is whether or not this cascade was a good one or a bad one. To do this all one has to do is examine the win rates of the three heroes during the International 2014. Despite being the most picked hero Razor only won 42.4% of the games he was in, and Skywrath Mage only won 40.4% of the games he was in. Shadow Shaman did win 56% of games he was in on the other hand. This is a nice example of how information cascades can lead to false conclusions. With a 10 million dollar prize pool hanging in the balance captains were picking heroes based on a flawed assumption that they were good. The two most picked heroes lost more games than they won which hardly suggests that picking them was a good idea. Interestingly enough the winning team Newbee has only picked razor twice in their history of play with a record of 1-1. Perhaps this distinction is what helped them win the whole tournament though avoiding of the pitfalls of information cascades.
Sources:
Graph of Hero Popularity Over Time http://www.datdota.com/hero_popularity.php?hero_popularity=0&hero1=101&hero2=27&hero3=15&hero4=&hero5=
Hero Stats for the International 4 : http://www.datdota.com/tournament.php?q=285&tournament=The%20International%202014&p=heroes
Information about the International 2014 2014: http://www.dota2.com/international/overview/