This thread is going to be replaced soon

General Discussion
1

THIS THREAD IS GOING TO BE SUPERSEDED BY A NEW ANALYSIS. I AM CURRENTLY IN THE PROCESS OF REBUILDING THE DATABASE FOUND AT HOTSAPI (WHICH HEROES PROFILE USES) AND I WILL RUN A NEW ANALYSIS WITH FRESH NUMBERS.

This is a response to another thread that you can find here.

That thread tries to draw a conclusion that solo players are at a disadvantage, but the conclusion is unsupported.

Below is the data they used from a Reddit thread that used an API to pull data for 16,820 games.

To the uninformed, this may seem like it’s significant enough to draw a conclusion from, but there is simply not enough data to suggest that since solo players ended up with a 49.6% winrate that they’re at a disadvantage.

You can find the Statistics Math below.

Team Composition Win Lose Games Count Win % Lose % 95% Confidence % of Games
5 Stack 857 756 1613 53% 47% ±2% 10%
1 Solo Players + 4 Stack 700 606 1306 54% 46% ±3% 8%
3 Stack + 2 Stack 201 199 400 50% 50% ±5% 2%
2 Solo Players + 3 Stack 1437 1426 2863 50% 50% ±2% 17%
1 Solo Players + 2 Stack + 2 Stack 677 666 1343 50% 50% ±3% 8%
3 Solo Players + 2 Stack 2544 2590 5134 50% 50% ±1% 31%
5 Solo Players 1994 2167 4161 48% 52% ±2% 25%
Overall 8410 8410 16820 50% 50% ±1% 100%

Alright, here goes. Let’s take all of the games that have a possible solo player in them. This means 1 Solo + 4 Stack, 2 Solo + 3 Stack, 1 Solo + 2 Stack + 2 Stack, 3 Solo + 2 stack, and 5 Solo.

Total Wins = 7352
Total Losses = 7455
Total Matches = 14807
Winrate = 49.6% winrate

Disadvantage as a Solo Player
7352 wins, 7455 losses, 14807 total, 49.6% winrate.

H_0: Winrate = 50%
H_a: Winrate < 50%

p = 0.19865

Conclusion: Fail to reject the null, you cannot draw a conclusion that a solo player is at a disadvantage.

5 Stack
857 wins, 756 losses, 1613 total, 53% winrate.

H_0: Winrate = 50%
H_a: Winrate > 50%

p = 0.00595

Conclusion: Reject the null, there is strong evidence to suggest an advantage with a 5 stack.

1 solo + 4 Stack
700 wins, 606 losses, 1306 total, 54% winrate.

H_0: Winrate = 50%
H_a: Winrate > 50%

p = 0.00465

Conclusion: Reject the null, there is strong evidence to suggest an advantage with a 4 stack.

I’m not going to type out the rest because you fail to reject the null hypothesis on all of them so 3 stacks and 2 stacks probably don’t have an advantage. However, it does appear to be supported that Blizzard favors 4 and 5 man stacks.

HOWEVER, the hypothesis that a solo player is at a disadvantage is unsupported.

Blizzard should reassess the MMR bias for 4 and 5 man teams, but there’s not evidence to suggest that queueing as a solo player in SL will cause you to lose matches.

In other words, Storm League appears to be working as intended. Whether or not 4 and 5 man stacks are truly intended to be favored is another story, but there is less than a 0.5% chance that 4 and 5 man stacks are not favored.

Source: Engineering Statistics class at Arizona State University

TL:DR

  • 5 mans are favored to win.
  • 4 mans are favored to win.
  • 5 solo players are favored to lose.
  • There is not enough evidence to support the theory that queuing as a solo player in any composition where you have solo queued puts you at a disadvantage. More data must be collected.

EDIT:

The reddit thread in question.

https://www.reddit.com/r/heroesofthestorm/comments/hrw1u9/do_5_stacks_have_an_advantage_the_stats_revealed/

EDIT 2:

Edited the title to make the intent clearer. It means you need more data to conclude that solo players are at a disadvantage, it does not prove that they aren’t.

15 Likes
2

Its nice to see some actual data on all this. Its interesting to see that even the most “unbalanced” team is only 4% from 50%.

Also for a bit of clarification, what does the “95% Confidence” mean and why 16,820 game isnt enough games to reach a proper conclusion?

3

It also depends how the 5 man is. Is the 5 man casual team full of 5 solo players that formed a team in LFG then you still have a chance to beat them.

The only problem with 5 stacks is if they are a 5 man try hard team full of smurfs that has 20 wins and 0 loses the past 20 games in a raw.

5 Likes
4

Does the data show any stacks (2, 3, 4 or 5) just versing a team of 5 solo players? What is the WR with that data?

Data shows the win/loss percentages based on team compositions not what stack vs what stack.

This data doesn’t tell me anything about how solo players are at a “disadvantage” but its assumed they’re not

2 Likes
5

as far as i know, its extremely unlikely that a team of X stacks will face 5 solos. thats why it takes longer to find games when you party up.

6

Sure, I can answer that.

95% confidence means that in order to be 95% confident (meaning only 5% error) the true winrate is ± the 95% confidence value. It is otherwise known as a confidence interval.

For instance, let’s take the 5 stack.
53% win
47% lose
±2% Confidence Interval at 95% Confidence.

This means that you can be 95% confident that the true winrate is anywhere between 51% and 55%.

The reason why 16,820 games is not enough to reach a proper conclusion for Solo Queue is because there is a 19.87% chance that the winrate being below 50% is wrong.

The p-value in the calculation tells you the confidence that the alternate hypothesis is wrong. A p-value of 0.19865 means that there is a 19.87% chance.

When you fail to reject the null hypothesis, you cannot immediately conclude that the null hypothesis is correct. You just fail to reject it.

Alternatively, however, if you reject the null hypothesis, you have a greater than 95% chance of the alternate hypothesis being correct. So thus, if you reject the null, it’s safe to say your alternative hypothesis is correct.

9 Likes
7

Since solo players has lower winrate then 4-5 stacks then 4-5 man stacks are favered big time.

No need to discuess this more reallly Solo players are in a disadvantage just from those data alone.

1 Like
8

I believe if you look at the bottom of the reddit thread, they broke down every type of matchup vs every other type of matchup.

There’s too many hypotheses to test, so you’d have to pick one for me to calculate it because I’m not calculating them all.

It’s not supposed to, it’s supposed to tell you that there’s not enough evidence to support the hypothesis that solo players are at a disadvantage. It does not prove that they aren’t, it’s saying that based on those numbers, there’s not enough evidence.

9

No, that’s where the conclusion is wrong.

You can’t just look at the number and conclude something, you have to be confident in your conclusion.

For example. Let’s say I flip a coin 100 times, I land on heads 55 times, and tails 45 times.

Someone might say, “Look, your coin is heavily favored to land on heads because it did 55% of the time” but that simply is not the case. There is no evidence based on that small amount of data to support the hypothesis that the chance to get heads is greater than 50%.

When I do a 1 prop z test, my p-value is .1587. I fail to reject the null hypothesis that the chance to land on heads is 50%, and this means that there is a 15.87% chance that I’m wrong by saying that heads is favored.

So no, you can’t just look at data and make a conclusion, you have to be CONFIDENT that your data is representative.

5 Likes
10

You know there are multiple more factors in this game then that like:

How many trolls does the enemy team has compared to the 5 man.
How bad is the enemy teams team comp compared to the 5 man
How many players on the enemy team are lightly going to afk because they get tilted/angry compared to the 5 man

Thats what solo que players complains the most about. The 5 man never has a troll or a feeder like 90% of the time while the solo team has all of them to handicap thier team even more.

2 Likes
11

This table shows the wins and loses for a 5 Solo Players against each of the listed team compositions

5 Solo Players vs Team Composition Win Lose Games Count Win % Lose % 95% Confidence % of Games
5 Stack 2 3 5 40% 60% ±33% 0%
1 Solo Players + 4 Stack 138 189 327 42% 58% ±5% 8%
3 Stack + 2 Stack 37 37 74 50% 50% ±11% 2%
2 Solo Players + 3 Stack 344 380 724 48% 52% ±4% 17%
1 Solo Players + 2 Stack + 2 Stack 130 148 278 47% 53% ±6% 7%
3 Solo Players + 2 Stack 599 666 1265 47% 53% ±3% 30%
5 Solo Players 744 744 1488 50% 50% ±0% 36%
Overall 1994 2167 4161 48% 52% ±2% 100%
6 Likes
12

I hope you do realize that what you just said makes a stronger case that solo players aren’t at a disadvantage.

If the chance for a troll/feeder in a 5 stack is 0%, then all you’re saying is that Blizzard might more correctly assess the true skill of a 5 man team and the reason why the data shows they have an advantage is because they can’t control for trolls/feeders. That would mean that Blizzard’s assumption to not handicap 5 man teams as much is possibly correct.

You’re going down a whole other rabbit hole when you make statements like that, but it only weakens your case, not strengthens it.

1 Like
13

Also the data dont show how many of those games were vs a 5 man try hard team. That were the problem starts really. casual 5 stacks are not the problem in this game.

2 Likes
14

So interestingly, if you do 5 solo players you end up with the following.

H_0: Winrate = 50%
H_a: Winrate < 50%

1994 wins, 2167 losses, 4161 games, 48% winrate.

p-value = 0.00366

Conclusion: Reject the null hypothesis, there is strong evidence to suggest that 5 solo players are at a disadvantage.

So yeah, if you get 5 solo players on your team, you’re at a disadvantage. This means that in roughly 25% of your matches you will be at a disadvantage.

1 Like
15

Again, your attempts to dissect the data only make your case weaker.

See the post above this which supports the conclusion that 5 solo players are at a disadvantage.

EDIT: No idea why auto correct changed dissect to disseminate.

16

Thanks

Is this in ranked or QM or both?

Be interesting to see, but interesting to see that 4 stack are dominating.

17

According to the reddit thread, the data is collected for Storm League.

18

This for ranked only. The other thing to note is that larger parties are fairly rare. 4 and 5 stacks make up 18% of the games total, and the total number of matches that featured a 5 stack vs 5 solo players was 5. This isn’t even 1% of the matches played by solo queuers.

5 Likes
19

One thing that baffles me tho is how perfectly and accurately ALL are to 50% winrate. The MM seems to be able to set the scene so well to not keep matches fair and even, but fair and even to 50%.

20

TL:DR for anyone interested. I’ll throw it up in the OP as well.

  • 5 mans are favored to win.
  • 4 mans are favored to win.
  • 5 solo players are favored to lose.
  • There is not enough evidence to support the theory that queuing as a solo player in any composition where you have solo queued puts you at a disadvantage. More data must be collected.
2 Likes