Dota 2: Getting Picky About Pick Order

Everyone that plays Dota 2 has a different opinion about how important the draft is. Feelings range from “We lost to the draft” to “The draft doesn’t matter, just play your hero”. I’m of the opinion that drafting is like picking a path down a hill. The choices you make determine your ease in getting to the bottom unscathed, but you can still catch yourself on a root or patch of difficult terrain that causes you to slip or outright fall over. Today, we’re going to take a look at just how important drafting can be. 

Before we can launch into this riveting examination, there are a few things that have to be cleared up and/or defined. First, not all of the games have been played on the ticket. All of the matches have been played and reported, but because match reporting does not include pick order, games that were not ticketed are a blank spot in the data. To make matters worse, some teams ticketed game 1 but did not ticket the rest of the series. For some of these series, this is because their opponent forfeited the second game, meaning there was nothing to put on the ticket. It’s unclear how many games fall under this category. It’s very common to make adjustments to the draft in games 2 and 3 based on teams’ prior performance and observations. Because the data for these matches isn’t available or the match wasn’t played at all, this will affect the data.

Second, I’m trying to avoid talking about drafting ambiguity, or the hiding of information about which hero will lane where and the farm priority for each. Even though it’s a subject that I could talk about for hours, all of the data I’ll reference and show is based on what position each hero was played as, not what they were picked as. Because of this, the data will have a specific kind of blind spot: certain teams can pick heroes that can fit multiple positions and they decide where the hero will go later on in the draft. That is an entirely different subject that could be taken apart and analyzed for hours, which I will not do in this particular article. “But Torrasque, how can you be certain that they played that hero as that position?” I’m afraid you’re just going to have to take my word for it or look through the data on Dotabuff yourself.

Third, because the league is still ongoing, I will not divulge too much information on each team. Some teams will still believe I have said too much, but if we followed the wishes of these teams, we wouldn’t be able to talk to them about anything even slightly interesting or important until they finished the season. Some players are so closed-mouth that they don’t even divulge anything until they are out of CSL entirely!

The last, and arguably most important, point is that pick order is more a framework on how to draft well than a rigid guideline on how to win every Dota game ever. I maintain that when it comes to drafting, pick order is probably the most important thing to figure out early on, as everything else hinges upon it. There are still countless other things to master, including picking good heroes, not giving away too much information, and picking heroes your opponent wants to ban before they can ban them (not to mention everything else involving banning). There’s a reason these teams have gotten as far as they are, regardless of any apparent skill gaps between them and the rest of the competition. 


The games that we’ll be using for this analysis are the regular season and playoff games played by the University of Toronto, McMaster University, the University of Waterloo, New York University, Stony Brook University, UC Santa Cruz, and the University of Wisconsin Madison. 

See the full-size image here.

When you look at the data from 196 ticketed games played by those teams, a couple of very interesting patterns emerge. In 93% of these teams’ drafts, a core hero was picked last. Mid heroes were picked last 65% of the time, carries were picked last 18% of the time, and offlaners were picked 10% of the time. It’s expected to save your last pick for your mid hero, since they typically need the highest amount of information to decide who is the best pick for the match and for their lane matchup. It was surprising to see that 7% of the time, a support was picked last. Some teams did it much more than others, but whether that is because of their drafting style or the nature of those specific drafts requires a much more focused analysis on those specific games and teams.

The next pattern I noticed was the preference for the position chosen in the first pick phase. It shouldn’t come as a surprise that with core heroes being picked last in 93% of games in our data set, positions 4 and 5 were picked in the first phase 62% and 57% respectively. Offlane was picked in the opening phase 43% of the time, with mid and carry at a meager 12% and 26%, respectively. At first, I was surprised to see the amount of teams picking their mid hero as early as that, even if it is comparatively lower, but I think it makes sense if you break it down. Imagine you’re drafting and your team really likes Leshrac or Puck. Several of your players can play them in almost every role. Later on in the draft, you can decide where to lane that hero and who it will be played by. This gives you an advantage if your opponent misreads your draft and assumes it to be a support. The flexibility of many heroes in Dota as well as the diverse hero pools of players can lead to very surprising drafts!

See the full-size image here.

Continuing on the subject of openers, it was interesting to see how certain early picks would unfold and how often teams picked in a certain way. For example, out of 196 games, roughly 25% of them started with the offlane and position 4 support chosen in the first phase. Of those games, about 73% of them followed by picking the other support and carry in the next pick phase, followed by the mid hero. That’s about 18% of all games! While I wouldn’t advise jumping to the conclusion that the team you’re facing is going to pick a specific way just because you saw some other teams do so, simply carrying that knowledge in the back of your head may give you the edge you need when you see the pattern begin to form.

Many professional and collegiate players share the opinion that picking two supports in the first phase is bad. As Kyohei “Poxy” Ikeda said in this previous draft article, “Picking two supports in the first phase isn’t good, especially when there are many good or broken cores available.” 43 drafts of the data-set had teams doing just that. If we take Poxy’s words to heart, 22% of games started with a less-than-good opener. As I discussed previously, picking two supports in the first phase potentially opens a team to strategies that a savvy opponent can exploit, although picking flexible heroes that may or may not be supports can mask that weakness. Regardless of your approach, I would recommend picking two supports in the first phase only if you know exactly what you're doing.


Drafting is very important, but it’s only as important as you make it. This sounds strange, but hear me out. The order you pick your heroes in is the basis of all drafting skills, but it’s still essential that you build other vital skills and try different strategies to see what works and what doesn’t. No one person or team’s word is law. Just because you’ve seen someone make it down the hill successfully a specific way doesn’t mean that way is the only way. Better paths are found all the time, and the search for new paths is one of the most enjoyable features of both professional and casual Dota. Remember to keep your eyes open and don’t go too fast--unless you like flying over the handlebars, I won’t judge. Best of luck in your next draft!


You must be signed in to leave a comment.

Sign In