Draft Analysis: The Defense
One of the biggest online competitions will conclude this week. With over 150 games played, joinDOTA's The Defense 3 is one of the biggest Western online competitions, featuring a prize pool of $20,000. This Saturday, Virtus Pro and Fnatic will battle for the first place in a tough best-of-five Grand Final. It is time to look back onto the entirety of the tournament and see which heroes are most popular and how trends evolve over the span of four months.
Rubick tops the list of most picked heroes, having been played in over half of all games. Both him and Shadow Demon were both chosen mostly in the second pick phase and are banned rarely compared to the other top picks such as Dark Seer. It is also interesting to see that carry heroes tend to be picked up in the second picking phase, with only Lone Druid being chosen in the first phase. There are many heroes who are both picked and banned frequently (see chart 'Popular heroes') and therefore appear in both lists - Batrider being the exception. He is banned in three quarters of all games, almost entirely in the first ban phase along with Magnus, Undying and Nyx. The hero banned the most in the second phase is Queen of Pain.
The 'Most Popular' chart looks at both pick and winrates combined. It shows that Batrider was picked or banned in 100 percent of all games, but banned more often than picked. Generally, these top contested heroes are banned more often than they are picked - excluding Darkseer who makes it through the bans about 50 percent of the time. Additionally Rubick and Shadow Demon, both of whom are in the top 3 picks, are banned less frequently. The chart to the right entitled 'Top 10 most popular heroes trending' looks at these top ten popular heroes over the span of the tournament. For this, the duration of the tournament was split down into periods of two weeks and the popularity rates were calculated for each of these periods. As you can see from the chart, the most contested hero, Batrider, was highly popular over the entire span of the tournament, just like Dark Seer and Magnus. After not being too favoured in the first half of December, Nyx quickly became a top pick; Undying's polularity however faded over time. Gyrocopter was not in the Top 10 for Most Contested Heroes - 16th place to be exact - but since he is the 'current flavour of the month', he was added to the chart. It is clearly visible that throughout all of December, January and the beginning of February, he was rarely picked but suddenly became the most popular hero for the last two periods.
To find out which heroes have the highest and lowest win rates, only the 50 percent of the most picked heroes were chosen - those with more than 11 games played. This is important since it makes sure that heroes with a 100 percent winrate who have only been picked once will not make into the list. Surprisingly, the most picked or most popular heroes are far from being the most successful ones. Invoker, who won 10 out of 14 games, tops the list. Heroes which appear in both this and the top 10 picked heroes are Lifestealer, Keeper of the Light and Wisp. It seems that to be a successful pick, a hero has to be strong in early game to score early kills and have a decent transition into late games as well. Carries are naturally very strong in late game, but Phantom Lancer's nuke and Lifestealer's slow are also very good in early game. On the other hand, Enchantress and Tiny are two prime examples on good early and late game, especially when picking up an Aghanims Scepter. It is interesting that we see top picks appear in the list of the least successful heroes, such as Luna, Magnus and Shadow Demon. The carries in this list have a far worse winning rate than Lifestealer or Phantom Lancer and it could be that they were outcarried both in the late and early game by these two.
Versatility of heroes picked in the first and second picking phase
The results: The average first phase pick can be put on 1.56 different lanes, the average second phase pick on 1.48. However, this difference is not significant (Statistical sidenote: chi² (2) = 0.952, p > .66). This means that heroes which are picked earlier in the draft are not more versatile than the heroes which are picked later.
More information on The Defense Season 3: GosuGamers Coverage Hub
GosuBet for the Grand Final: VP vs. Fnatic.EU