Piercing the Veil: Selection Priority
Within the game of Dota 2 there is an obvious fog of war. It represents uncertainty — forcing players to make instinctual, logical, yet best-effort decisions. After the game is done, that fog’s veil is pierced — the players can review the game in it’s entirety and see where their intuition deviated from reality. Their information goes from an imperfect to a perfect state — a critical part of the learning process (a really cool article on this is available here). Well, almost perfect…
Some small things are not always accessible publicly. One of these is the decision making behind how it is decided which team plays on which faction and which team has first pick. Like traditional sports, this can have a significant impact on the game.
There are four ways that I’ve seen this handled in events, and the first two of these methods are very similar, as you’ll see below.
Type A (Valve events + ~65% of significant non-Valve events)
- Game One has a coin toss. Winner has “Selection Priority” and can select between Radiant, Dire, First Pick or Second Pick. Loser picks between remaining options.
- Game Two Selection Priority flips to the other team, and teams repick accordingly.
- Game Three (and Five, Seven, Nine) we have *new* coin toss for that game, and the winner of that coin toss has Selection Priority (which flips over for the following even-numbered Game only).
- For the first round of Winner’s Bracket and Loser’s Bracket, the team with the higher seed has Selection Priority in Games #1 and #3. Similarly, for the Grand Finals, the team coming from the Winner’s Bracket has Selection Priority in Games #1, #3 & #5.
Type B (~35% of significant non-Valve events)
- as A#1 above
- as A#2 above
- The team with Selection Priority in Game One has it in all odd-numbered games, and the team with Selection Priority in Game Two gets it in all even-numbered games.
Type C (historically a format used by joinDota)
- as A#1 above
- Game Two we see side/pick swap over. If you had First Pick and Dire, you now have Second Pick and Radiant (for example).
- We continue to see flips in each subsequent game.
Type D (I can’t even recall where I saw this used)
- as A#1 above
- Loser of Game One has Selection Priority in Game Two
- Loser of Game Two has Selection Priority in Game Three
- Loser of Game <X> has Selection Priority in Game <X+1>
What’s wrong with these methods?
Type D (the ‘alternating format’) was a system which forced teams to limit sides/picks to the fairest possible compromise because if you had an insane combination (like First Pick + Dire in the TI5 meta) in game #1, you’d have to play against it in game #2. Also a huge advantage outright for the team who won the initial coin toss. This method leads to longer series on average.
Type C, the old format by joinDota forced teams to play with Factions or Pick Selection they’re not used to — often teams would both prefer the opposite of what they got. This method also lead to longer series on average.
Type A and B are very similar, but with a key difference. The Valve rationale is that selection priority is just an arbitrary yet as-fair-as-possible method of determining sides/pick order, and that just knowing that you definitely have Selection Priority in Game #3 (a deciding game in a bo3) as you start game #1 is in itself an advantage. You’re able to prepare for multiple games knowing you can dictate how the draft for future games begins (and your opponent has to prepare for all eventualities, with no knowledge of what you’re considering). This leads to substantially unfair preparation time for the team with more information.
Is Choosing Always An Advantage?
Not all teams are the same, so naturally one question which arose given the popular Type A/B formats is “can our team defer Selection Priority?”, as in can the team with Selection Priority force the other team to choose first between Radiant/Dire/First Pick/Second Pick and then they pick between the remaining options?
The first time I thought about this was at The Frankfurt Major, when OG (up 2–0 against Secret) took Second Pick in the 3rd game of the Grand Final — essentially a statement that they wanted to react, not lead the draft. They went on to lose the game (but win the match), and remained the only team of the Frankfurt Major Main Event to pick Second Pick. One of the players that I’ve seen caring about deferring is Swindlezz — he’s deferred in a few events I’ve worked on. This is super important for teams who view not having Selection Priority as a bonus, and hence are being punished for doing well in the group stage (see Type A, comment # 4 above).
Fast forward, and some other teams now intentionally break minor rules (like they join the lobby 10 seconds late on purpose), for which the prescribed penalty is ‘loss of Selection Priority for that game’. This suggests that for lots of team Selection Priority isn’t always viewed as a benefit and statistically there’s not a strong correlation between Selection Priority and victory once you offset the average advantage gained from max(‘radiant and first pick’ - ‘dire and second pick’, ‘dire and first pick’ — ‘radiant and second pick’). Some teams might perceive it as an advantage, and for some teams there are pretty significant deviations and so there’s no reason not to just allow deferring as part of the rules (or at the very least remove it as a “perk” of doing well in the group stage).
Free Common Access
So, now you hopefully understand how Selection Priority works in tournaments we can get to the real question here — why is it important? Well, it’s a core part of the game setup. It’s the first important decision you make in a match. It’s also our (as external viewers) first insight into a teams core strategy and understanding of the game and the metagame — their decision to prioritize getting their heroes of choice or their faction of choice; or to entirely react to their opponents. It’s more complex and more significant than playing White or Black in chess.
Yet unlike which side each team is playing, or what heroes they picked — Selection Priority isn’t recorded anywhere public. Sometimes casters make reference to which team chose what, but this is based on what they see in the pre-game lobby (or is told to them by some illustrious statsman via the production feed), and isn’t consistently communicated. This means that this information is very incomplete. If it becomes a highlight in a specific match, let’s say between EG and VP that EG keeps picking Dire or First Pick when they have Selection Priority, then for other teams who will eventually draft against EG they have more information about EG than EG has about them. It also incentivizes collectives of teams (whether collectives by nationality, by friendship or by ownership) to share information between themselves on what opponents like to do. Sometimes at events players who are knocked out become consultants for other teams for the rest of the event, so this information can also become part of that working relationship.
I believe that this information is significant yet simple enough to be stored in the replay and WebAPI (and hence parsable by your favourite stats websites [coming soon, please stop pestering me]) and visible in the in-game client. There should be no benefit by having access to this data where others don’t, and no loss for having casters of your game discussing it as a talking point when they are silent about it for other series.
I actually manually recorded the Selection Priority for almost every game I’ve done stats for — but to release this information during the event without any warning that it will be done also seems like an unfair thing to do. Releasing it after the event & patch seems like a fair thing to do though, so here we go:
