Originally published as part of Team Liquid’s TI3 Retrospective Part 1 (Writers: phantasmal, TanGeng // Gfx: riptide, Heyoka // Editors: Firebolt145, ScintilliaSD, TheEmulator // Photography and art via Valve and R1CH)
With all the questions during The International asking why Alliance was always on Dire and Na`Vi was always on Radiant, I thought I’d take some time to clear up how team selection worked. From there we can look at the patterns in team selection to see what they reveal about each team’s preferences and also how those preferences might have influenced the Radiant/Dire and 1st/2nd pick win rates over the entire event.
Let’s get the easy stuff out of the way first. According to 2p.com, each prelim series begins with a coin flip. The winning team then gets to choose either a side or a draft order, and the losing team gets to choose the remaining option. In game two this order is flipped, with the loser of the initial coin flip getting the first selection.
For the playoffs the process is similar, but instead of a coin flip, the initial selection goes to the team with priority in the matchup. In the first round of the playoffs, priority is determined on seeding. For all other rounds, priority goes to Upper Bracket teams over Lower Bracket teams. If both teams are from the same bracket, priority goes to the team with the higher win rate.
With that out of the way, let’s look at how the prelims went down. We can’t determine who won the coin flips, but we know that every team in a group has an identical schedule to every other team in the group (minus themselves of course), and that every team got first selection in precisely half of their games. This makes it an environment that is ideal for a quick aggregate analysis.
To start things off, it is my belief that teams almost always preferred 1st pick to 2nd pick. The simplest evidence for this is Group B. There are only two teams in this group that had second pick more than 8 times: Rattlesnake and Alliance. These are also the two teams that showed the greatest overall preference for Dire in the entire tournament. Group B is also peculiar in that only 4 out of the 56 games featured a team that had both Radiant and 2nd Pick. The simplest explanation for these trends is that Group B was largely full of teams that preferred both 1st pick and Dire. In this environment, if you select Dire, I would select 1st pick. If you select 1st pick, I would select Dire. As a result over 90% of the games in this group featured 1st pick Radiant and 2nd pick Dire.
Group A is a more difficult case because the teams in this division seem to be split over their side preference. In this group 33 out of 56 games featured a team with both Radiant and 2nd pick. If we continue to assume that no one actively tries to get 2nd pick, we can conclude that Group A had a slight preference towards Radiant overall.
Another thing we can do with Group A is try to determine which teams preferred which side. If we assume that no teams actively sought 2nd pick, whichever side the teams play more often when they have 2nd pick reveals their side preference. Using this test, Na`Vi was unsurprisingly the most pro-Radiant, followed by mouz, Zenith, and Fnatic. On the other end of the spectrum, DK and LGD.cn were the most pro-Dire, with MUFC showing a slight Dire preference. Dignitas showed no consistent preference in either direction.
Another interesting fact when we look at the patterns in 2nd pick across both groups is that only 3 out of 16 teams played more than one game on both Radiant and Dire when they had 2nd pick: Fnatic, Dignitas, and TongFu. It is my suspicion that these three teams had the most malleable side preferences depending on who they were facing in their current match and what they wanted to run.
So with side preferences established, which teams prioritized 1st pick over side? This is actually difficult to answer from the aggregate data. Ideally we would look at specific game patterns, but trying to cover all of that here would be impossible. Based on what I’ve looked at, my belief is that Orange is the team that most favored 1st pick in the entire tournament. Close behind were iG, Liquid, DK, MUFC, and possibly LGD.cn. The teams that most favored side over 1st pick were Rattlesnake, Alliance, Na`Vi, Fnatic, and mouz.
So to summarize everything, in Group A Na`Vi, mouz, and Fnatic all exhibited a strong preference for Radiant over everything, but mouz and Fnatic showed more inconsistencies than Na`Vi. DK, LGD.cn, and MUFC all showed a 1st pick then Dire preference. Zenith was probably 1st pick then Radiant. Dignitas was all over the place and may have been responding to specific match conditions.
In Group B, Alliance and Rattlesnake showed only side preferences, with both preferring Dire to everything else. TongFu behaved similarly to Dignitas and exhibited no clear overall preference. Every other team appeared to prefer 1st pick then Dire, with VP being the most inconsistent in this preference.
With all that out of the way, I want to turn my attention to the win rates for sides and pick orders. To be frank, my earnest advice is that when it comes to TI3 do not trust the pick/side win rates. It’s not merely that it’s a small sample; pro games as a whole are always going to be a small sample so you just try to make do. The bigger problem is that the sample is in no way random. To illustrate this, let’s look at the overall win rates:
What’s going on is most clearly displayed by 2nd pick Radiant. That pick/side combo has 54% win rate over 57 games, which looks pretty substantial. Unfortunately, 22 out of the 57 2nd pick Radiant games in the entire tournament were played by Na`Vi, who had a 72% win rate when on that side/pick combination. Given the degree team preferences warped the distribution these win rates cannot be considered reliable.
More importantly, consider that the top 3 teams all had different philosophies when it came to pick/side combo, and they all had pretty clear motives behind these preferences. For Orange, they were heavily dependent on grabbing Visage, who was the most drafted hero of the tournament. Securing first pick gave them the best shot at grabbing this highly coveted hero before their opponents. For Na`Vi, being on the Radiant side opened up their options in the hard lane by allowing them to farm ancients with either Mirana or Windrunner. For Alliance, being on the Dire side fit into their overall strategy to force early Roshan fights that they could win by utilizing buybacks.
The lesson to take from this is that you’re likely better off adopting a side/pick preference that fits what your team wants to accomplish. Sure, the statisticians might be able to tell you which side is marginally better on average, but surely you’re not striving to be just an average team. If you want to be a real contender, you need to go the extra step, look at the benefits of each choice in full detail, and then find a way to utilize those benefits in your overall team strategy.