Composition Pool Diversity beats Hero Pool Diversity

June 06th
written by Karahol


The true strength of a team is really hard to evaluate in the constantly changing environment of Overwatch. I believe that one indicator of the strength of a team is how many compositions its players can play without losing efficiency. In other words, how well the players’ hero pools can be combined rather than how deep of a hero pool each player has.

Having gone over all of Rogue’s games before leaving for Korea I checked the players’ hero pools and I tried to envision the possibilities of counter strategies they could run with their players’ hero pools. Simultaneously, I also tried to see their weakness composition-wise and the striking one was having the same player on Genji and D.Va, two key-heroes in some compositions; and I have mentioned this multiple times 

This got me thinking that somehow teams must have players covering a role and have certain heroes in their hero pool that do not limit the team combinations but only expand them.


Composition Styles are too generic

Back in the day, when the scene of Overwatch was just starting to shape up a lot of people, Harsha being a notable among them, valued player flexibility and deep hero pools very highly. I wasn’t entirely convinced on that and therefore I sat down and did my research on the hero pool versatility split in two parts: a) Hero Mentality (part 1, part 2, part 3) and b) Ultimate Inconsistencies (part 1, part 2).

The reason I wasn’t entirely convinced was because some heroes are more important for their Ultimates, while others are more important for their skill set and some for their entire arsenal. So, what I considered more important was to have players who can pick up heroes when needed but not necessarily to be able to play 6-7 heroes at the highest possible level. I deemed it more necessary to have the ability to play certain compositions per map phase rather than certain heroes or composition styles throughout the map.

By composition styles I mean what everyone has in mind when saying “2-2-2″, “Triple Tank”, “Triple DPS”, “Triple Support” etc. These are composition styles that have many variations.


For example, you can play 2-2-2 with the following combinations of heroes:

Tracer – Genji – D.Va – Winston – Ana – Lucio

Tracer – Genji – D.Va – Winston – Zenyatta – Lucio

Pharah – Genji – D.Va – Winston – Mercy – Lucio

Sodier76 – Tracer – Reinhardt – Roadhog – Ana – Lucio

Soldier76 – Genji- Reinhardt – Zarya – Ana – Lucio

There are too many variations of the 2-2-2 setup that one can explore to be mentioned in this article, but I want to point out that that depending on a team’s players’ hero pools, some of these combinations may not be available. Taking Rogue as an example, they can play both “Triple DPS” and “2-2-2″ styles, but the variations of those styles they can play are too limited.

They can’t run a “2-2-2″ variation where D.Va & Genji are both in the composition efficiently because they have one player covering both roles. And they can’t run a Triple DPS composition with no Genji because NiCOGDH hasn’t shown us that he can play any other hero apart from Genji and D.Va mostly. Therefore, they can’t even pull a “Pharah – Tracer – Soldier76″ Triple DPS composition if they need to.


Composition Pool Diversity versus Hero Pool Diversity

Again drawing information from Rogue’s run in APEX S3, we can see some really interesting stuff when comparing their composition diversity to Lunatic-Hai’s.

If one checks the team’s entire hero pool in the tournament, one will notice that Lunatic-Hai doesn’t really have a much bigger hero pool in absolute numbers. The two teams’ differences probably boil down to Lunatic-Hai having some Roadhog and Rogue having Pharah in their hero pools, respectively.


However, if you break down the hero pool on a per player basis and try to draw the combinations of the hero pools in terms of compositions, then you see the significant advantage Lunatic-Hai has over Rogue. Both Zunba and Esca can play Soldier76. Both Ryujehong and Esca can play Widowmaker. WhoRU can play Tracer as well. Lunatic-Hai can play more combinations of the “Triple DPS” & “2-2-2″ styles than Rogue.

In other words, Lunatic-Hai has more aces up their sleeves so as to unlock points and maps with strategies built around these compositions.


One of the crucial aspects of combat is to find a way to surprise your enemy, to hold the element of surprise. In a game like Overwatch and especially on maps like Hybrid or Assault, startling your opponent almost results in an easy win. And at the same time, you force your opponents to practice more for future matches in order to be prepared to counter all your tactics while you keep devising approaches and making your game even more unpredictable. One more aspect greatly affected by the composition pool diversity is the Ultimate Management. In some cases, teams can’t play certain heroes and are forced to use heroes that have really slow ultimate charge rate, like Zarya, and therefore they set themselves up for failure in this area as well. Or they can’t even get their Ultimate game up and working until very late into the game.


How to measure a composition’s strength  

Right now Winston’s Lab offers all the compositions a team has played per match/tournament/time frame with winrates based on the winrate of the team on a map level. No matter how well or badly they did with a certain composition on a map, as long as they won the map, their composition’s winrate is considered positive. This is skewing the results, however.

What we need in the future is a way to see the comp’s strength in regard to kills/deaths/ultimates/healing etc numbers and fights won/lost. With such a formula, we can have a metric that shows us exactly the strength of a team’s diversity in comps; the Composition Pool Diversity metric (or CPD metric). Then we can compare teams more accurately on a composition level and this will help us to further evaluate the role and performance of players on a personal level too. Once we know how well the whole team does with a specific composition, then we can weight the performance of the pieces far easier and more accurately.

Even if a player is a specialist on Tracer, like SoOn, it’s very different to play in a 2-2-2 composition style, where there are only 2 DPS players and therefore he needs to carry more than in a Triple DPS style, with 3 DPS, where he’s not required to carry as much and his performance may be somewhat weaker.


The Proper Pieces 

If this is truly the future of Overwatch, where you can get value out of every hero under certain circumstances, then one should start prioritizing getting the right pieces for his team rather than having only a team that can yield good results in the short term. Especially once Overwatch League starts, it will be far more important for teams to have players that can flex between a certain range of heroes while specializing on 1-2 of them, rather than specialists who play the meta heroes to the best possible level. Obviously, larger rosters with 8-9 players can cover these weaknesses but it really depends on how the whole structure of the league will play out.

Eventually, in my opinion, players with practice will be able to reach the same kind of expertise that we see right now some players have achieved and the deciding factor won’t be the mechanics but the strategy execution. If a team can’t adapt its strategy to execute a counter-strategy, then in games like Overwatch, this will become a glaring weakness.

Team EnVyUs had this weakness back in APEX S2, but they have transformed and now they look far better against the Korean powerhouses. Cloud 9 also had this weakness in the same tournament, with players that prioritized kills over playing the objective and the game in general, but now they look in far better shape and they will be a big threat once they get more practice together. It’s not always necessary to adapt by switching pieces. In some cases, being able to re-allocate the players on different roles is more important, because their skill set may drop out of favor for a short while but it may be needed in the future.

This is the case with EscA in Lunatic-Hai, who was forced to practice Tracer after the departure of LeeTaeJun but now he’s starting to get more comfortable on her and at the same still retains his previous hitscan hero pool (Widowmaker, McCree, Soldier76).


Closing Thoughts

Evaluating players’ performances in a team environment is really hard, but with the data WinstonsLab has provided us thus far and will continue in the future, I am very positive that we will reach the level of creating metrics that represent a team’s and player’s performance more clearly very soon. It’s hard but it’s doable.

Stay tuned for more next level analysis.

– Karahol