Statistics on 2v2/3v3/4v4 team composition

Post a Reply

1 2 3 Next All

StartingCrafty

United States32 Posts

August 22 2010 23:55 GMT

Poll: What is your 2v2 arranged team composition?

Protoss/Zerg (113)

21%

Terran/Protoss (111)

21%

Terran/Zerg (106)

20%

Random/Random (63)

12%

Protoss/Random (30)

6%

Zerg/Random (28)

5%

Zerg/Zerg (27)

5%

Terran/Terran (21)

4%

Protoss/Protoss (20)

4%

Terran/Random (19)

4%

538 total votes

Your vote: What is your 2v2 arranged team composition?

(Vote): Terran/Zerg
(Vote): Terran/Protoss
(Vote): Protoss/Zerg
(Vote): Terran/Random
(Vote): Protoss/Random
(Vote): Zerg/Random
(Vote): Random/Random
(Vote): Zerg/Zerg
(Vote): Terran/Terran
(Vote): Protoss/Protoss

It seems like most of the posts about Starcraft stats are focused on 1v1 win rates/race distributions/etc, but I thought it would be interesting to take a look at how the 2v2/3v3/4v4 leagues break down.

My data is based upon http://www.sc2ranks.com, I wrote some scripts to derive some stats on team composition that they didn't have available.

2v2 team composition:
+ Show Spoiler +

So, let's start out with the Top 100 2v2 arranged teams in the North America diamond league:

1 Terran 1 Protoss : 30
1 Terran 1 Zerg : 28
1 Protoss 1 Zerg : 16
2 Random : 7
1 Zerg 1 Random : 4
2 Protoss : 4
1 Protoss 1 Random : 3
1 Terran 1 Random : 3
2 Zerg : 2
2 Terran : 2

Not a huge sample size, but it's clear much more people like to take a pair of races than play random, and Protoss/Zerg is the least popular pair by a significant amount.

So now let's take the Top 1000 2v2 arranged teams in north america diamond:

1 Terran 1 Protoss : 298
1 Terran 1 Zerg : 185
1 Protoss 1 Zerg : 142
2 Random : 98
1 Protoss 1 Random : 72
1 Terran 1 Random : 61
1 Zerg 1 Random : 47
2 Protoss : 31
2 Terran : 31
2 Zerg : 25

The distribution here is pretty nice, all paired teams are on top, followed by all teams with a random, followed by all double teams. Terran/Protoss pulls away significantly from the others.

And just in case you're curious, these are the Top 1000 arranged teams in diamond across regions:

1 Terran 1 Protoss : 268
1 Terran 1 Zerg : 243
1 Protoss 1 Zerg : 175
2 Random : 74
1 Terran 1 Random : 55
1 Zerg 1 Random : 44
1 Protoss 1 Random : 44
2 Protoss : 32
2 Zerg : 28
2 Terran : 27

Very similar distribution, the main difference is that Terran/Zerg moved up a bit.

The platinum league is very similar, the main difference is that Terran/Zerg is tied with Protoss/Zerg, it drops behind Protoss/Zerg in the gold league.. The silver league also has Terran/Zerg behind Protoss/Zerg, and double Protoss/double Terran teams are much more common. 2v2 Bronze is nearly identical to silver.

It's worth noting that Terran/Protoss is at the top of every single league I looked at, twice as popular as the next most common team in the bronze league, and with a substantial edge in a few others. I'd guess it's the relative popularity of the two compared to Zerg, and the fact that Terran/Zerg is higher than Protoss Zerg in diamond may be because of the current strength of Terran.

Summary: The top three combinations at all regions and levels are Terran/Protoss, Terran/Zerg, Protoss/Zerg. Terran/Protoss is the most popular everywhere, Terran/Zerg is more popular than Protoss/Zerg in diamond and below it elsewhere.

3v3 team composition

+ Show Spoiler +

So moving on, the top 100 3v3 teams in us diamond:

1 Terran 1 Protoss 1 Zerg : 43
2 Terran 1 Protoss : 10
1 Terran 1 Zerg 1 Random : 6
1 Terran 2 Zerg : 6
1 Terran 2 Protoss : 6
1 Terran 1 Protoss 1 Random : 5
1 Protoss 1 Zerg 1 Random : 4
2 Terran 1 Zerg : 4
3 Random : 3
1 Zerg 2 Random : 3
1 Protoss 2 Random : 2
2 Terran 1 Random : 2
1 Protoss 2 Zerg : 2
2 Zerg 1 Random : 1
2 Protoss 1 Zerg : 1
3 Protoss : 1

As you might expect, the most obvious trend is that most teams prefer to mix it up and have one of each race in the mix, with some other significant outliers.

There are only 600 or so teams in us diamond at the moment, but thanks to the server being wonky I can only grab the Top 400 3v3 in us diamond:

1 Terran 1 Protoss 1 Zerg : 133
1 Terran 1 Protoss 1 Random : 33
1 Terran 1 Zerg 1 Random : 25
3 Random : 23
1 Terran 2 Protoss : 21
2 Terran 1 Protoss : 21
1 Protoss 1 Zerg 1 Random : 16
1 Protoss 2 Zerg : 15
2 Terran 1 Zerg : 15
1 Protoss 2 Random : 14
1 Terran 2 Random : 14
1 Terran 2 Zerg : 14
2 Protoss 1 Zerg : 14
1 Zerg 2 Random : 12
2 Protoss 1 Random : 8
2 Zerg 1 Random : 7
2 Terran 1 Random : 7
3 Terran : 2
3 Zerg : 1
3 Protoss : 1

As you can see, 5 out of the top 6 combos have at least one terran on their team, combos with Terran/Protoss and Terran/Zerg outnumber Protoss/Zerg as usual. The picture for all regions at diamond is pretty much identical, in the platinum/gold leagues TPZ is still the most popular triplet but you get a lot more TTP and TPP

Summary: Terran/Zerg/Protoss, aka one of each race is the most popular triplet at all levels, otherwise one of the other two races plus a random, with triple random having a significant presence as well.

4v4 team composition

+ Show Spoiler +

There's only 74 teams in 4v4 Diamond so let's jump straight to the top 1000 in us platinum:

1 Terran 1 Protoss 1 Zerg 1 Random : 93
2 Terran 1 Protoss 1 Zerg : 82
1 Terran 1 Protoss 2 Zerg : 79
1 Terran 2 Protoss 1 Zerg : 74
1 Terran 1 Protoss 2 Random : 62
1 Terran 2 Protoss 1 Random : 53
1 Protoss 3 Random : 50
4 Random : 41
2 Terran 2 Protoss : 40
2 Terran 1 Protoss 1 Random : 39
1 Protoss 1 Zerg 2 Random : 38
1 Terran 1 Zerg 2 Random : 32
2 Protoss 1 Zerg 1 Random : 32
1 Terran 3 Random : 30
1 Zerg 3 Random : 25
2 Terran 2 Random : 25
2 Protoss 2 Random : 23
2 Terran 1 Zerg 1 Random : 22
2 Protoss 2 Zerg : 22
3 Protoss 1 Zerg : 16
1 Terran 3 Protoss : 16
1 Protoss 2 Zerg 1 Random : 15
1 Terran 2 Zerg 1 Random : 14
2 Terran 2 Zerg : 13
3 Terran 1 Protoss : 12
3 Terran 1 Zerg : 10
1 Protoss 3 Zerg : 6
3 Terran 1 Random : 5
1 Terran 3 Zerg : 5
2 Zerg 2 Random : 4
4 Terran : 4
3 Protoss 1 Random : 3
3 Zerg 1 Random : 2
4 Zerg : 2
4 Protoss : 1

So let's get a more coherent breakdown of that. These are all the teams with a guaranteed one of each race:

1 Terran 1 Protoss 1 Zerg 1 Random : 93
2 Terran 1 Protoss 1 Zerg : 82
1 Terran 1 Protoss 2 Zerg : 79
1 Terran 2 Protoss 1 Zerg : 74

That's 328 total, showing that most people still prefer diverse teams. After that, Terran/Protoss based Teams are pretty popular:

1 Terran 1 Protoss 2 Random : 62
1 Terran 2 Protoss 1 Random : 53
1 Protoss 3 Random : 50
2 Terran 2 Protoss : 40
2 Terran 1 Protoss 1 Random : 39

253 falling into that. It's worth noting that there are 41 teams with four randoms, but there aren't any other huge groupings that jump out at me.

Summary: Out of the top 1000 platinum teams (only 79 players in diamond at the moment) 328 teams have one of each race plus a random or a duplicate of some other race, 253 teams are based around Terran/Protoss with the other slots being random, 41 teams are quadruple randoms.

And finally some stats on league distribution by gametype, for just the US region:

For 1v1s:

Diamond contains the top 7% of all players
Platinum contains the 7% to 21% range
Gold contains the 21% to 36% range
Silver contains the 36% to 58% range
Bronze contains the bottom 42% of all players

For 2v2s:

Diamond contains the top 4% of all players
Platinum contains the 4% to 18% range
Gold contains the 18% to 41% range
Silver contains the 41% to 64% range
Bronze contains the bottom 36% of all players

2v2s are 75% players in picked teams, the other 25% are randoms. The diamond league is 63% players in picked teams.

For 3v3s:

Diamond contains the top 2% of all players
Platinum contains the 2% to 14% range
Gold contains the 14% to 34% range
Silver contains the 34% to 56% range
Bronze contains the bottom 44% of all players

3v3s are 78% players in picked teams, 22% in randoms. The diamond league is 32% players in picked teams.

For 4v4s:

Diamond contains the top 3% of all players
Platinum contains the 3% to 14% range
Gold contains the 14% to 38% range
Silver contains the 38% to 62% range
Bronze contains the bottom 38% of all players

4v4s are 72% players in picked teams, 28% in randoms. The diamond league is 13% players in picked teams.

Ultimately the fact that the diamond leagues for 3v3s and 4v4s are dominated by randoms makes me think team composition isn't going to be a huge deal in the 2v2/3v3/4v4 scene, and I think players play for very different reasons than 1v1 games, which probably encourages a lot more variety in composing teams. The trends are interesting to see so far, I'd be interested if any of these numbers change significantly as the game progresses.

Let me know if there's anything here that doesn't make sense or anything you're personally curious about.

Morayfire73

United States298 Posts

August 23 2010 01:08 GMT

Nice data, never knew that diamond in 4v4, and 3v3 was so elusive, makes me feel special. However your data in for random groups is probably off, since if you play say a 4v4 with a party of 3 and pug one person, you truly aren't a random group, but still count as one. And getting one random person, isn't truly a full random team.

StartingCrafty

United States32 Posts

August 23 2010 02:06 GMT

That's probably true, sc2ranks only lists teams in two categories, "full" teams of 4, and everyone else is classified as a "random". The fact that the bulk of players are in picked teams could be overstating things too, since a group of 6 people could form 20 possible 3v3 teams, and be counted as 60 players in picked teams if they all went through placement. I'd be curious how many of the randoms play in 3v3s/4v4s by themselves, or with a partner or two.

Congrats on making it to diamond though, me and my friends are slumming down in Gold for 2v2s and 3v3s, and I rarely play enough with the same set of 4 people to be ranked in 4v4s, which makes me wonder if the team leagues will get more bronze-heavy over the long run with a lot more idle teams.

StartingCrafty

United States32 Posts

August 23 2010 16:42 GMT

I main Zerg btw, and I'm typically with Protoss/Zerg in 2v2s, and Terran/Zerg/Protoss or Protoss/Protoss/Zerg in 3v3s.

StartingCrafty

United States32 Posts

August 26 2010 17:37 GMT

A lot more voters than posters. Anyway, here's as much of the 3v3 team composition as you can fit in a poll.

Poll: What is your 3v3 Arranged Team Composition?

Terran/Protoss/Zerg (8)

22%

Double Random. (5)

14%

Two different races, one random (4)

11%

Triple Random. (4)

11%

Terran/Terran/Protoss (3)

8%

Terran/Terran/Zerg (3)

8%

Terran/Protoss/Protoss (2)

6%

Protoss/Protoss/Zerg (2)

6%

Protoss/Zerg/Zerg (2)

6%

Two of the same race, one random (2)

6%

Terran/Zerg/Zerg (1)

3%

Three of the same race. (0)

0%

36 total votes

Your vote: What is your 3v3 Arranged Team Composition?

(Vote): Terran/Protoss/Zerg
(Vote): Terran/Protoss/Protoss
(Vote): Terran/Zerg/Zerg
(Vote): Protoss/Protoss/Zerg
(Vote): Protoss/Zerg/Zerg
(Vote): Terran/Terran/Protoss
(Vote): Terran/Terran/Zerg
(Vote): Two different races, one random
(Vote): Two of the same race, one random
(Vote): Double Random.
(Vote): Triple Random.
(Vote): Three of the same race.

Heyoka

Katowice25012 Posts

August 26 2010 17:41 GMT

These are some cool numbers, I wonder if you can extract any useful meaning from any of it. I feel almost like the game is still so new and the team game so undeveloped that its not very meaningful yet.

StartingCrafty

United States32 Posts

August 26 2010 17:45 GMT

On August 27 2010 02:41 heyoka wrote:
These are some cool numbers, I wonder if you can extract any useful meaning from any of it. I feel almost like the game is still so new and the team game so undeveloped that its not very meaningful yet.

I think it shows you much more about what's a popular team combination then what's a successful one. I don't think team games will ever be as competitive as 1v1s, so it's likely teams are decided much more according to what people find fun to play together than what is a dominant strategy. Even if three zerg players always six-pooling could make it to diamond, not many people would enjoy that style of play with their friends because of the lack of variety.

But it'll be interesting to see if the diamond leagues ever diverge into a few optimal combinations over the long run, or if it's always a mix.

StartingCrafty

United States32 Posts

August 26 2010 18:00 GMT

One early comment now that we have exactly 100 voters in the 2v2 poll: Team Liquid is extremely similar to the outside world, with our top 4 combinations being identical to the top 4 in diamond. The main difference is that Protoss/Zerg and Terran/Protoss seem to have switched places, which is interesting because Terran/Protoss was the most popular in every league I looked at. I wonder if that's because of the forum's dislike for Terran, or if battle.net simply has a higher number of Terran players because of the campaign, and it's intended to be the starter race.

me_viet

Australia1350 Posts

August 26 2010 18:18 GMT

I think the reason P/T or Z/T is so popular is because of how strong the t1 combo is. Zlots to tank and marines to do some nice dps. That actually destroys most other armies. I don't think any other t1 armies can beat beat a Zlot/Marine army.

t2 Siege tanks/ Sentries with FF would be pretty imba too.

ta2

125 Posts

August 26 2010 18:33 GMT

#10

In popularity terms then.

In 2v2: TP > TZ > PZ, therefore T > P > Z in 2v2.

In 3v3: TPZ > TPR > TZR therefore T > P > Z in 3v3.

In 4v4: TTPZ > TPZZ ~= TPPZ therefore T > Z ~= P in 4v4.

Though the difference in popularity between P and Z in 3v3 and 4v4 is not statistically significant. Terran is clearly favoured in all matchups though.

ltortoise

633 Posts

August 26 2010 19:55 GMT

#11

In team games, I think zergs are the best to have on your team. You have to protect them in the early game, but they reward you with mutalisks which are just awesome.

StartingCrafty

United States32 Posts

August 26 2010 21:05 GMT

#12

On August 27 2010 04:55 ltortoise wrote:
In team games, I think zergs are the best to have on your team. You have to protect them in the early game, but they reward you with mutalisks which are just awesome.

Yep, I love a mobile Muta army, and I'm always a fan of the team working together to let someone tech up and pull off something cool.

Gnial

Canada907 Posts

August 26 2010 21:28 GMT

#13

People like Husky claim that PP is a disadvantaged combo, and the data seems to reflect that sentiment. Pretty much all my closer friends are protoss players, so I've been playing a TON of PP and PPP. It took us a lot of games to figure out just how complimentary a PP combo can be. Once you learn how to identify and defend against certain cheeses like zealot/reaper, you get the opportunity to create extremely complimentary multi-tech-path armies in the early-mid game, like void ray/phoenix/colossus, or templar/immortal, to name a few. Perhaps that explains why pure protoss combos are the least popular of all the combos - noone has had the patience or drive to figure out how to deal with reaper/zealot attacks, and other tactics that are usually strong against protoss. Hell, we almost all gave up too, haha, but once you figure it out it is so powerful.

I predict the PP combo will gain greater representation in top 2v2 diamond in the coming months.

Alsn

Sweden995 Posts

August 26 2010 21:49 GMT

#14

Lets make some assumptions, they aren't nearly true, but bear with me.

Lets assume 2v2 is perfectly balanced across all combinations. Lets then assume that the distribution(popularity) of all those combinations is not evenly distributed. What would then be the expected the result when taking a sample(such as top 1000)? Distribution in that set would exactly mirror the popularity of the combinations.

Thus, either the game is perfectly balanced and your numbers show popularity or the game is not perfectly balanced(extremely likely) and without numbers showing popularity of the combinations over the entire population you can't really draw any conclusions about balance.

Also note that neither can you assume that a sample "top x" does not necessarily reflect popularity either. It's entirely likely that toss/terran is stupidly overpowered but people just don't enjoy playing the combination; thus misrepresenting the combination downwards making it look like it is more balanced than actual.

All in all, it's very hard to draw any conclusions from this, but something it does tell you is that if you are at the top in 2v2, you are most likely to be matched up in distributions similar to your data. Simply because the matchmaking system pits you against teams with similar rating to yours.

StartingCrafty

United States32 Posts

August 26 2010 21:49 GMT

#15

People like Husky claim that PP is a disadvantaged combo, and the data seems to reflect that sentiment.

I wonder though if any combination that only allows for one race isn't going to be a disadvantaged combo simply because your opponents narrow in the range of possible options they're playing against, there's less stuff they could have to counter.

And given that the majority of diamond players in 3v3 and 4v4 don't play with a full arranged team, my instincts would be to say that diamond level teams are going to form partially around what they find successful, but partially around the players who may prefer forming a diverse team where they fill different roles.

ta2

125 Posts

August 26 2010 21:52 GMT

#16

On August 27 2010 06:05 StartingCrafty wrote:

Show nested quote +

Yep, I love a mobile Muta army, and I'm always a fan of the team working together to let someone tech up and pull off something cool.

And brood lords!!

StartingCrafty

United States32 Posts

August 26 2010 21:56 GMT

#17

On August 27 2010 06:49 Alsn wrote:
Lets make some assumptions, they aren't nearly true, but bear with me.

Lets assume 2v2 is perfectly balanced across all combinations. Lets then assume that the distribution(popularity) of all those combinations is not evenly distributed. What would then be the expected the result when taking a sample(such as top 1000)? Distribution in that set would exactly mirror the popularity of the combinations.

Thus, either the game is perfectly balanced and your numbers show popularity or the game is not perfectly balanced(extremely likely) and without numbers showing popularity of the combinations you can't really draw any conclusions about balance.

Also note that neither can you assume that a sample "top x" does not necessarily reflect popularity either. It's entirely likely that toss/terran is stupidly overpowered but people just don't enjoy playing the combination; thus misrepresenting the combination downwards making it look like it is more balanced than actual.

All in all, it's very hard to draw any conclusions from this, but something it does tell you is that if you are at the top in 2v2, you are most likely to be matched up in distributions similar to your data. Simply because the matchmaking system pits you against teams with similar rating to yours.

It is going to be extremely hard to separate whether combinations are popular because of relative balance issues or whether it's simply what people enjoy playing as more. The most useful analysis might be to compare popularities across leagues, where for example Protoss/Zerg is much lower in diamond compared to the lower leagues. The fact that leagues show some differences in popularity would suggest that people may begin to gravitate towards certain combinations or switch away from them when they get to the top.

But the most telling thing may be to track this historically, and see if the diamond leagues for all these things gradually shift, and particularly whether they shift after a patch that alters the balance, that would give you some idea if people choose teams based upon how much they like the races as opposed to balance concerns.

StartingCrafty

United States32 Posts

August 26 2010 21:58 GMT

#18

On August 27 2010 06:52 ta2 wrote:

Show nested quote +

And brood lords!!

I love them as a closer for long games, particularly when you can focus on them and your allies can provide anti-air support. If we were better maybe we wouldn't let things get to that point, but it always seems great when you can get them.

ta2

125 Posts

August 26 2010 22:11 GMT

#19

On August 27 2010 06:58 StartingCrafty wrote:

Show nested quote +

Here's how I like to go in 3v3 or 4v4 when I'm in an able team:

1 base lings into roaches just for defense whilst teammates wall-in
2nd base
spire
pump mutas for harass, until they get defenses up + flyer attacks 1
infestation pit + 2nd spire
hive + 3rd base + flyer carapace 1
greater spire + pump corruptors
broodlords
rest of upgrades

Get 2 utility hatcheries - you will have SOOOO much spare minerals that you will just have to make lings. Alternatively you could send minerals to your allies or pump spine crawlers, though that's not really an option usually as your creep screws up your teammates. Also don't forget to get melee upgrades for the broodlings.

StartingCrafty

United States32 Posts

August 27 2010 00:06 GMT

#20

Just for the sake of completeness, here's a poll for the 4v4 options, as neatly as I think it'll fit into 12 options and with a focus on spelling out the most popular ones.

Poll: What is your 4v4 Arranged Team Composition?

Quadruple Random (4)

25%

Terran/Terran/Protoss/Zerg (3)

19%

Terran/Protoss/Protoss/Zerg (2)

13%

Triple Random (2)

13%

Terran/Protoss/Zerg/Random (1)

6%

Two of one race, two of another (e.g., TTPP) (1)

6%

Two of one race, one of another, plus a random (e.g., TTPR) (1)

6%

Two of one race, plus two randoms. (e.g., TTRR) (1)

6%

Two different races plus two randoms. (e.g., TPRR) (1)

6%

Terran/Protoss/Zerg/Zerg (0)

0%

Three of one race, one of another including random(e.g. TTTP or TTTR) (0)

0%

Four of one race. (e.g., TTTT) (0)

0%

16 total votes

Your vote: What is your 4v4 Arranged Team Composition?

(Vote): Terran/Terran/Protoss/Zerg
(Vote): Terran/Protoss/Protoss/Zerg
(Vote): Terran/Protoss/Zerg/Zerg
(Vote): Terran/Protoss/Zerg/Random
(Vote): Two of one race, two of another (e.g., TTPP)
(Vote): Three of one race, one of another including random(e.g. TTTP or TTTR)
(Vote): Four of one race. (e.g., TTTT)
(Vote): Two of one race, one of another, plus a random (e.g., TTPR)
(Vote): Two of one race, plus two randoms. (e.g., TTRR)
(Vote): Two different races plus two randoms. (e.g., TPRR)
(Vote): Triple Random
(Vote): Quadruple Random

1 2 3 Next All

Please or register to reply.

Statistics on 2v2/3v3/4v4 team composition

Completed

Ongoing

Upcoming