One thing I talk about a lot is that you can model teams in regards to which players they rely upon the most in comparison to all other teams in a competition.
Before I break down what I mean by that, here is what I found: Teams that rely on their DPS players too much have the worst odds to win tournaments. In fact, when looking at the 9 past tournaments for which there is sufficient data, only 3 out of the 18 finalists are more reliant on their DPS players than the average team at that tournament. Additionally, out of the 9 teams that relied on their DPS the most in those tournaments, only 1 even made it into the top 50%.
In the PTK-Model a team is assigned a style depending on what players/heroes they rely upon (in comparison to all other teams). A team can be a “Tank-Team”, “DPS-Team”, “Support-Team” or a combination of those (for example a “DPS/Tank-Team”).
To ‘rely upon’ a role X means the following: Let’s say in the considered tournament heroes of role X get 20% of their teams’ kills (PTK) on average. A team then ‘relies upon’ role X if that teams’ heroes of role X get a PTK higher than 20%.
For example, in APEX S2 every team for which the DPS heroes got more than 25.4% of their teams’ kills would, corresponding to the PTK-Model, be considered a “DPS-Team”. For this tournament Lunatic-Hai was a Support-Team, RunAway a Tank-Team and LW Blue a DPS/Tank-Team.
The team that relies most on a specific role in a tournament is the team with the highest PTK for that role. In APEX S2 the team that relied most on their Supports was Lunatic-Hai as can be seen above. The team that relied most on their DPS players was Cloud 9 and the most reliant on Tanks was Conbox Spirit.
In case you wonder why the percentages do not add up to 100%: The displayed values show the average PTK of a single hero. If a team played two Support heroes, we take their weighted average. That way these numbers are not as dependent on the meta or style of a team – Triple DPS teams would have much higher total DPS-PTK values than Triple Tank teams.
Saying a team ‘relies upon’ a role is not really an accurate description of what is going on, but it was the best term I could find. If someone has a better way of describing what the PTK-Model expresses please feel free to let me know.
DPS-Teams have the worst results
Now that we understand how the model works, let’s look at the results. There are nine tournaments for which there is sufficient data: DreamHack Winter, MLG Vegas, IEM Gyeonggi, APEX Season 2, Winter Premiere, OW PIT EU, OW PIT NA, Overwatch Rumble and TakeOver 2.
As mentioned earlier, only 3 DPS-Teams made the Top 2 in those 9 tournaments. When compared to the 9 Tank-Teams and 11 Support-Teams that made it, it gets hard to deny the actual importance of good Tank and Support players. In case you wonder why the numbers don’t add up to 18, remember that teams can have a combination of styles, i.e. look at LW Blue above (DPS/Tank-Team).
Things get more interesting once you look at the extremes. While it is true that a team that is heavily reliant on their DPS players does rarely get into the final, it is also true that you can’t have “bad” DPS players.
As the following graphic shows only one of the 9 teams that had the highest PTK values for DPS even got into the top 50%. The most outstanding results are those of Support-Teams: 5 of the highest Support-PTK Teams made it into a final and 7 of the teams with the lowest Support-PTK at their tournament finished in the bottom half.
All of this indicates that it is far more important to have above average Tank and especially Support players than to have above average DPS players. In other words it does not necessarily mean that the best players on a team have to be the Supports, but a good Support player can have a bigger impact on a team than a good DPS player can have as an individual.
Anyone who is interested can check out which teams where assigned which styles by the PTK-Model on our PTK-Model page. For reference I will leave you with the Top 2 teams from those 9 tournaments that had sufficient data (and here is also a link for that specific query):