What's In A Draft: TI3 Group Stage Draft Analysis

by UnoPolak

Dota has always had a close link with statistics, dating back to the spreadsheet of numbers at the end of a game in the original WC3 mod. Valve has continued this tradition with live stats for spectators, a WebAPI for match history, and a basic replay parser that Statsman Bruno has expanded upon. Now there is a shiny new tool called Skadi, allowing even more information to be parsed from replays. For example, we can efficiently determine not only which heroes were picked and banned by a specific team in a bunch of replays, but how long it took them to do it. The International 3 pitted the sixteen best teams against each other, so where better to look to see how top captains drafted? We will be looking at just the group stages since there are an equal number of games for each team, and we should be able to establish general trends as well as tease out the differences between teams. Spoiler alert: Na`Vi and Alliance fans will not be disappointed.

Overall Draft Times

The International 3 took place during patch 6.78c. This patch had five 30 second bans, five 40 second picks, and 110 seconds of reserve time, giving a maximum possible time of 460 seconds per team. You can see the full breakdown of teams below, but two deserve special mention.

Fastest Draft in the West: Virtus.Pro had the shortest individual draft of 63 seconds in a victory over LGD.int (match page), which breaks down to an average of 6 seconds for each pick or ban. The next lowest individual draft was 118 seconds by Rattlesnake, which is nearly a full minute longer. Virtus.Pro also had the lowest average draft time of 253 seconds, so apparently CIS Dota can be as aggressive in the draft as it is on the battlefield.

Slow and Steady Wins the Race: Alliance had both the longest individual draft of 449 seconds in a victory over Orange (match page) and the highest average draft time of 401 seconds. They banned only 23 different heroes, tied for lowest during the group stages, and still spent nearly all of the allotted time for each individual ban.

In fact, the top three teams in the entire tournament were all in the four longest average draft times during the group stages. Of course just taking longer in the draft does not guarantee success, but it does suggest that these teams were very methodical, taking as much time as possible to think through the draft in order to set up their team for success. We’ll look at exactly how they did this a little later.

Pick/side preference

Another thing that we can look at is the preference for specific sides or pick orders. Before each match, a coin flip was performed backstage, with the winner choosing either a side or first/second pick, and the loser selecting whatever was left. Unfortunately these decisions were not recorded, so we can only make educated guesses. The general consensus seems to have been that Dire is the better side with the much talked about “Dire advantage” around Roshan, and FP being the better pick order since you get both the overall first and last pick. The entire group stage saw 75 games pitting Radiant with FP against Dire with SP, and only 38 games of Radiant with SP against Dire with FP. Interestingly, teams with both Dire and FP only went 16-22 in the group stages though, so either the consensus was wrong, or execution simply trumps all.

 

Dire and/or Bust: LGD, Rattesnake, DK and Alliance got Dire at least 11 times each. It is difficult to believe that they won the coin flip that many times, but this does indicate that if given the choice Dire was their preference. There was mixed success, ranging from Alliance’s 11-0 run on Dire to Rattlesnake’s 2-10.

The Middle Road: Dignitas appeared to have no bias to either side or pick order with a nearly equal distribution of Radiant/Dire and FP/SP. We cannot speculate if they had specific preferences against individual opponents either since they only played two games against each other team.

Bucking the Trend: Na`Vi, mouz and Fnatic all showed a preference for Radiant over either Dire or even FP, opting for Radiant and SP 9, 8, and 7 times, respectively. Na`Vi’s captain Puppey has stated in the past that he prefers the Radiant jungle, and Funn1k’s ancient farming on Radiant during the group stages was much commented on during the International, so they obviously had a plan which led them to a 10-2 record as Radiant. The other two did not fare quite as well with records of 4-6 and 5-4.

Individual Pick/Ban Timings

The effect of Radiant vs. Dire on the individual pick and ban times is negligible compared to who has first pick (FP) and second pick (SP), so we will be focusing on pick order only. The averages from all of the group stage games follow trends that are not very surprising if you are familiar with drafting. However, there are a few teams that do wildly different things from the others, and that is what we will be highlighting as we go along.

Draft Preparation: Teams had been preparing for months before TI3, and a lot of this effort would have been spent preparing their drafts. The first few bans and picks would have been enumerated and weighed long before each game, and as a result was that the very first ban from each team generally took less than 20 seconds. A few teams took a little longer, but curiously enough, none more so than Zenith. Iceiceice frequently sat there with the first overall ban watching as up to 49 seconds cut into Zenith’s reserve time before making a selection. Was this somehow another mind game coming from the enigmatic Iceiceice?



Back-to-Back Decisions: There are a few points during the pick/ban phase during which a team has to makes two decisions in a row. The first one is when the team with SP gets back-to-back picks in the first picking stage. Both of these picks are given a full 40 second clock. However,  we found that the second pick only took 15 seconds on average, which is just under half the 33 seconds for the first pick. The most extreme example of this back-to-back picking was Team Liquid, who frequently spent 30 seconds of their reserve time for that first pick before spending less than 3 seconds of the allotted 40 for the second pick. One school of thought is that by quickly locking in that second pick you give your opponent less overall time to plan. Contrast this to Na`Vi, who used less time on the first pick (47 seconds) and a lot more time for the second pick (42 seconds). Two explanations for this difference come to mind. The first and obvious reason is that Na`Vi do not choose their second hero before they pick the first and thus need to spend the time thinking about their choice. The more interesting reason is that Na`Vi usually decide on their heroes just as quickly as Liquid, but choose to take their time and discuss the rest of their draft before locking their selection in. This does give the opposing team an extra 40 seconds to think about their own draft, but if they don’t know what your second pick is, the time is more effective for you than them. This “information advantage” is not unique to Na`Vi either.

Time as a Resource

We are used to thinking of health, mana, and gold as resources, but what about draft time? Each team can lock in their pick or ban before the allotted time has run out, after which the remaining time is lost. We can add up that lost allotted time during the draft for each team, and a very clear trend appears.

Waste Not, Want Not: Alliance never lost more than 65 of the total allotted 350 seconds, whereas LGD.int, TongFu, Rattlesnake, VP, LGD and Dignitas lost at least 70 seconds in every single draft. I believe this is another example of a team using an information advantage to increase the effectiveness of their draft relative to their opponents, similar to how Na`Vi extended their SP second pick but in this case expanded to every single pick and ban. Their hero pool was relatively small with the fewest number of different heroes banned at 23 and an average number of different heroes picked at 34. It is possible that Alliance spent all of this additional time considering exactly which hero to pick or ban, but their hero pool is does not suggest they were considering a much wider number of heroes than their opponents so it is more likely that they were already looking ahead in the draft.



Play to Your Strength: Orange had the second lowest lost allotted time, but used their reserve time very differently to Alliance. Alliance preferred to use their reserve time during the middle portion of the draft, specifically the second ban phase. On the other hand Orange preferred FP which also gave them the last overall pick, something that the opponent could not counterpick at all. They placed a lot of emphasis on that last overall pick by stockpiling their reserve time and entering that last pick with an average of 49 seconds of reserve time, compared to Alliance’s 17 seconds. A few other teams averaged as much reserve time on their last pick as Orange, but none came close to spending the average of 37 seconds of reserve time on that last pick like Orange did. During TI3 there were quite a few comments about the number of different heros that Mushi was playing, and nearly half the time Mushi’s hero was picked at the very end. From this we can say that Orange’s strategy revolved around building a flexible team without expending too much reserve time and then finding that perfect hero to complete the draft at the very end.

Conclusion

The top three teams of TI3 (Alliance, Na`Vi, and Orange) each displayed a methodical approach to drafting. They used the majority of the allotted time for each pick and ban as well as their reserve time. In regards to the allotted time usage it is highly probable that they were leveraging an information advantage over the opponent, keeping their decision hidden while continuing to think about the next step in the draft while the opponent was still unsure about the current pick or ban. Of course, other teams also used the majority of the allotted time to much less success, so one must remember that this is not a game-changing advantage, but an interesting one to analyse nonetheless. The way these top teams utilized their reserve time varied quite a bit, with pinpointed bans in the second phase from Alliance and exotic last picks from Orange. The competition at TI3 is stiff, but these three teams in particular found that bit of efficiency in the draft to help set themselves up for success.

---

About the writer:
UnoPolak started playing Dota 2 in early 2012 and is now an administrator of the Dota Noobs community. More recently he started contributing to the open-source Skadi project and is one of the founding members of SkadiStats, a group whose aim is the in-depth analysis of professional replays.