Boards›Gameplay

A single number to measure premade winrate is misleading

Hyrum Graff··4 votes·1,178 views

It lumps together coordinated premades with uncoordinated premades.

The premade handicap should be balance the winrate of solos who communicate vs coordinated premades.

Premades should expect to lose if they don't use voice chat. Just like solos should expect to lose if they refuse to ping or chat.

Premades who do use voice chat should expect to win more than 50%. Just like solos who ping and/or chat.

This is just how math works. Every game, 50% of people win. If you have some people winning less than 50%, some also win more than 50%. This might mean the average premade winrate is higher or lower than 50%.. and that's OK.

Let's make up some numbers for an example.

Suppose solos who communicate have a 55% winrate. Riot modifies the premade handicap so that premades with voice chat also have a 55% winrate. This leaves premades without voice chat at a 40% winrate.

Scenario 1: 50% of premades use voice chat.

The average winrate for premades overall is 47.5%. That's OK, because premades that don't use voice chat shouldn't expect to win in a competitive queue.

Scenario 2: 90% of premades use voice chat.

In this case, the average winrate for premades overall is 53.5%. That's also OK, because solos who communicate win just as much; those that don't should stop bitching and start pinging.

If only 10% of premades actually use voice chat, but 50% of solos use pings/chat, solos are going to have a higher average winrate than premades*. And vice versa. Which -- again -- is OK as long as the solos who ARE pinging have the same winrate as premades in voice chat.

*Actually, there's more to it, like the winrate of solos who refuse to communicate might be different from the winrate of premades without voice. The point is still valid, it just isn't the ONLY thing that you need to consider.

https://am-a.akamaihd.net/image?f=https://news-a.akamaihd.net/public/images/articles/2016/july/DQA/graph-winratesforgroups.jpg?v=1

Premade winrate is like champion winrate: you need a lot of other numbers for context before it means anything.

A solo who communicates has a chance of getting matched with other players who refuse to communicate. The premade handicap is compensation for being able to avoid that chance. It's not an excuse for anybody, premade or solo, to refuse to communicate.

Advantage premades have that I didn't mention in this post:

  • Teammates guaranteed to have a compatible playstyle
  • Teammates who you've played with before
  • Probably something else I'm not thinking of

By the same logic as this post, we should balance the premade handicap around those that DO take advantage of this capacity, which might cause premade winrates overall to drop (or rise). And that's OK as long as a solo doing all the right things has the same chance to win as a premade doing all the right things.

Rammus OK?

17 Comments

Zielmann2 votes

Care to let us know how they can distinguish groups of people who queued together from other groups who queued together?

Also, any hard numbers you have with proof of a winrate disparity between the two would help. Otherwise this entire thread is speculation.

Hyrum Graff2 votes

{quoted}

Care to let us know how they can distinguish groups of people who queued together from other groups who queued together?

I'm not quite certain what you are asking.

Also, any hard numbers you have with proof of a winrate disparity between the two would help. Otherwise this entire thread is speculation.

I think this misses the point of the thread. I don't think you'd go into a thread titled Champions should be balanced around their performance by their mains in high elo and say "without hard numbers, this is just speculation!"

This thread is about how we should do balance. It's a philosophical argument; the numbers are just there to help people understand what I'm getting at. That's why they're made-up scenarios with lots of 'if's, 'then's, and 'might's.

Zielmann3 votes

I'm not quite certain what you are asking.

My point exactly. A group queues together and is using voice comms. Another group queues together and isn't. As far as the client is concerned, these are just two groups who queued together. On top of that, how would you handle a situation where some of the group is in voice comms and some aren't?

A single number to measure premade winrate is misleading

If a thread is intended to be of a philosophical nature, you shouldn't present it as fact. I don't disagree that it's right to put thought into data presented to you. But if you're going to lead off with a hard claim as opposed to a speculative statement, you need to back up that claim.

Hyrum Graff2 votes

{quoted}

My point exactly.

I was not sure what you were asking, so I asked you to rephrase/clarify. I am not sure how this relates to your point at all. Regardless, you did clarify so I can respond, now that I understand your point.

A group queues together and is using voice comms. Another group queues together and isn't. As far as the client is concerned, these are just two groups who queued together. On top of that, how would you handle a situation where some of the group is in voice comms and some aren't?

I don't know of any way to identify whether groups are in voice chat or not. Additionally, there are other advantages to being in a premade, as I noted at the end of the original post, that it's hard to separate from the advantage of voice chat. It's particularly difficult to separate the effect of voice chat specifically vs the effect of guaranteeing that your teammates will use pings/chat (at least without doing experiments where you know if a party is using voice).

Fortunately, I don't think you need to separate out each individual factor. Instead, you can look at things like:

  • Players who consistently queue with two different groups of people, but have a much higher winrate with one group
  • Players who consistently play with the same team vs those who just invite their whole friend list
  • Distribution of premade winrates. Is it a normal bell curve or something else?

If a thread is intended to be of a philosophical nature, you shouldn't present it as fact. I don't disagree that it's right to put thought into data presented to you. But if you're going to lead off with a hard claim as opposed to a speculative statement, you need to back up that claim.

Just so we're on the same page: I think my hardest claim is that the premade vs solo winrate needs context to be valuable. Do you agree?

If so, I think the logic is sound:

  • Assumption: coordinated premades (particularly, those using voice chat) have a higher winrate.
  • Fact: The winrate data we have aggregates data from premades using voice and those without.
  • Opinion: Balance should be done assuming premades are taking advantage of voice (and other premade - exclusive benefits)
  • Conclusion: The given statistic lacks important context.

Continuing to summarize stuff I've said in the comment thread here:

  • Fact: we can't easily obtain the relevant information about use of voice
  • Conclusion: The next best thing is to learn more about the variance in how premade status can affect winrate, so that the premade handicap can be tuned around premades that use their advantages well.

I'll note that I do think a premade using all their advantages to their fullest probably should have a higher winrate than those same players queueing solo. But I think that shooting for 50% is not the way to do it.

Maybe later I'll make a pretty graph to explain this; I think that might be a lot more accurate with a lot fewer words.

BloodyTompon2 votes

I'm going to lump a few different desperate things you've said because I think it's slightly easier to follow if they're converged, so bear with me.

Premade winrate is like champion winrate: you need a lot of other numbers for context before it means anything.

Yes and no. It really depends on a few things. For example, if we look at the recent stats Riot released, I think it's easy for those who have some stats to bear the curse of knowledge - i.e., they struggle to put themselves in the position of when they did not know stats. Imo these are quite good simplified stats for supporting the argument they made. They're short, brief and even if know sweet F.A. about stats, you can still see what they're trying to show.

Now, is win rate the be all end all? Absolutely not. Is there an a great amount of complexity hidden under that number? Yes there is.

Fortunately, I don't think you need to separate out each individual factor. Instead, you can look at things like:

Players who consistently queue with two different groups of people, but have a much higher winrate with one group Players who consistently play with the same team vs those who just invite their whole friend list Distribution of premade winrates. Is it a normal bell curve or something else?

This is what I would be most interested in too. Intuition tells us that the more people play together the better they get, so how exactly that learning curve works would be something I'd like to know. They wouldn't be able to track each permutation individually, so they'd have to use some heuristic and that's what I would be the most intrigued about.

Premades should expect to lose if they don't use voice chat. Just like solos should expect to lose if they refuse to ping or chat.

Premades who do use voice chat should expect to win more than 50%. Just like solos who ping and/or chat.

.

I'm using it to express a philosophical viewpoint, that premades should be balanced around players who are taking full advantage of them (like how Riot balances champs around their power in plat+, not bronze). [from a different thread -http://boards.na.leagueoflegends.com/en/c/gameplay-balance/ygGPafH3-dynamic-queue-will-never-have-balanced-matchmaking ]

I disagree with the second sentiment and possibly the first. I think something that kind of resembles a Utilitarian approach is best. Whilst the Utilitarian maximises happiness, imo the matchmaking should attempt to maximise the number of people as close to 50%. I can see the rationale for balancing the game around the high end, but I can't see how that holds when applied to matchmaking. To that end, I agree with the first sentiment if it's a consequence of maximising 50/50 chance, but disagree if it isn't because I don't believe it ought to be the goal.

Hyrum Graff1 votes

{quoted}

I'm going to lump a few different desperate things you've said because I think it's slightly easier to follow if they're converged, so bear with me.

Thanks, it's easier for me as well, especially on mobile (don't ask me why I'm writing this from a cell phone instead of opening my laptop that's literally 5 ft away).

Yes and no. It really depends on a few things. For example, if we look at the recent stats Riot released, I think it's easy for those who have some stats to bear the curse of knowledge - i.e., they struggle to put themselves in the position of when they did not know stats. Imo these are quite good simplified stats for supporting the argument they made. They're short, brief and even if know sweet F.A. about stats, you can still see what they're trying to show.

Now, is win rate the be all end all? Absolutely not. Is there an a great amount of complexity hidden under that number? Yes there is.

I agree that if you're going to pick a single stat to show they picked a good one. I'd still like to see more.

This is what I would be most interested in too. Intuition tells us that the more people play together the better they get, so how exactly that learning curve works would be something I'd like to know. They wouldn't be able to track each permutation individually, so they'd have to use some heuristic and that's what I would be the most intrigued about.

+1

I disagree with the second sentiment and possibly the first. I think something that kind of resembles a Utilitarian approach is best. Whilst the Utilitarian maximises happiness, imo the matchmaking should attempt to maximise the number of people as close to 50%. I can see the rationale for balancing the game around the high end, but I can't see how that holds when applied to matchmaking. To that end, I agree with the first sentiment if it's a consequence of maximising 50/50 chance, but disagree if it isn't because I don't believe it ought to be the goal.

I think that all league games should strive to be as balanced as possible. I talk more about what that means in this post. Tldr, equally skilled teams and also minimum skill variance between players.

Additionally, I think ranked matchmaking should provide a way for people in all premade sizes to demonstrate their skill through their rank.

That's where the parallel to champion balance comes in to play. It sucks to pick up a champ, fall in love and main them, but then at some point they just aren't viable and you need to learn another champ to keep climbing (looking at you, old garen).

Similarly, it sucks to get to a certain elo exclusively solo, but now that you reach a certain rank you need to queue in a premade in order to keep climbing, because the premade handicap is tuned around uncoordinated premades. (I'm aware this is little bit of a straw man since most players still play solo. But it is a possibility that we must consider, and think what message is sent by allowing this). It's being told your way of playing League isn't valid anymore.

For the record, it also sucks to get to diamond playing in a group of 4 or 5, then suddenly you can't any more, for the same reasons. I'm sympathetic to solo players but I also play in premades (as well as solo) so I think I'm fairly unbiased in this area. I'm more interested in solutions in this complex problem space.

I'd guess the best solution analyzes when certain players play as a team vs when they're just randoms who want to rank together and adjust the handicap accordingly. I still favor the approach of balancing around coordinated premades, especially to maintain the incentive to take advantage of the benefits available to premades, but this is a way of softening the blow of a harsher handicap for less coordinated premades. A compromise.

BloodyTompon1 votes

I agree that if you're going to pick a single stat to show they picked a good one. I'd still like to see more.

Did you see the ones posted in the comments section? They were win rate too, but a lot of people seemed to prefer those ones.

I think that all league games should strive to be as balanced as possible. I talk more about what that means in this post. Tldr, equally skilled teams and also minimum skill variance between players.

We're on the same page with this and have just phrased it differently.

Additionally, I think ranked matchmaking should provide a way for people in all premade sizes to demonstrate their skill through their rank.

I'm not picking up what you're putting down. I'm not understanding what you mean by this and consequently the parallels either.

Hyrum Graff1 votes

Aside, I love the conversation in this thread, it's a shame it didn't get more visibility.

If only I had a name like MM or LF that got me automatic upvotes, lol.

Zielmann1 votes

Sorry. I approached this all wrong and that's my bad.

In the interest of a real discussion, I would argue that whether or not the premade has voice comms is irrelevant. The data presented represents an overall average. Regardless of the distribution of comms vs. no comms, it works out to appear to favor the premades by maybe a couple percentage points. This sample would essentially contain an average distribution of groups with comms vs. without.

So, over time, a single player's experience as a non-5-man versus a 5-man group will give an average number of groups both with and without comms, likely resulting in approximately an average winrate against premades as a whole (slightly below 50% overall against premades).

Hyrum Graff1 votes

I'm on mobile atm, hence long delays. Just posted a reply to the other comment before I saw this. Some of that is relevant to this one.

Agreed that whether they have voice comms specifically is irrelevant. It's more about whether they take advantage of the opportunities afforded by premade status. I'll continue saying "voice" instead of that just because it's shorter, though.

I think that solo winrate vs 5's with voice should be just under 50%, and winrate vs 5's without should be a little over 50%.

I'd guess that most 5's don't really take full advantage. IF that's true, I'd want to see solos average winrate vs 5's be slightly above 50%. But we really need more info about the distribution of how well premades use their advantages to make that determination.

Which is kinda what I was saying in the first place :P

Maximum Morde1 votes

I disagree. I shouldn't need to ping or anything to pull a 50/50 winrate. The game should adjust my mMr to the point where I'm winning 50/50 without it. If that means it's lower, fine. I don't like people. I don't like chatting with people. I don't like holding their hands because they are literally too stupid to pay attention to the minimap.

Hyrum Graff3 votes

[{quoted}](name=Maximum Morde,realm=NA,application-id=3ErqAdtq,discussion-id=oPKpez4E,comment-id=0001,timestamp=2016-07-26T15:35:14.157+0000)

I disagree. I shouldn't need to ping or anything to pull a 50/50 winrate. The game should adjust my mMr to the point where I'm winning 50/50 without it. If that means it's lower, fine. I don't like people. I don't like chatting with people. I don't like holding their hands because they are literally too stupid to pay attention to the minimap.

Assuming you've plateaued, I agree. But you don't get to that mmr by winning 50/50.

Battlecast Sona2 votes

why are you playing a 5v5 game in a competitive environment if you don't even want to utilize smart pings...

Maximum Morde1 votes

Mostly because I like the Pentakill theme. So It's irrelevant to me how everyone else feels. I just make my pick and jam to Smite&Ignite.

Battlecast Sona1 votes

this is a post about ranked though. Premades have been in normals since the birth of the game - only crazies want them taken out of normal matches.

So I repeat my question - why are you coming into a competitive 5v5 environment if you don't want to even ping your team to communicate? That's all. I don't care if you wanna play normals and mute your whole team and the enemy team as soon as you enter the game and pretend your alt key doesn't exist, but if you're playing ranked and you're not communicating you're severely hurting your team's chances of winning out of the gate.

Maximum Morde1 votes

Duh, to get my end of season rewards, free champ skin and shiny border.

Poske1 votes

Riot cannot have accurate way of measuring. Its Dynamic Q after all and common sense. Also Riot always makes games harder for premades because of voice calls if you arent ussing 1 your problem

The main issue is existance of smurfs Smurfs just give riot bunch of innacurate DATA

I ll list how every fair player is fked up by Dynamic Q

If you are playing league of legends solo you are at disavantage since you arent cheating the system with Dynamic smurfing chances that the enemy team is are higher? Why well simple man 5 enemies 4 allies (Given that you dont abuse the system)

If you are playing league of legends with non smurf premades you are also at disadvantage but how so Sub heroCro You arent making much sense To put stuff in perspective

Premades win 50% of games what if i told you thats not true

Smurfing premades win 55% of games while actualy fair non smurfing premades win 45% of games.

Riot Raised matchmaking standard for premades to much because of innacurate data from Smurf Premades

Reasons i just listed are why everyone wants soloq

So Soloq players are fked up Non smurfing premades are fked up People who profit from Legal boosting are happy

More from Gameplay

3 Refunds per season, upvote if you want 3 refunds per season.
Gameplay764 comments
7,167
votes
3 Refunds per season, upvote if you want 3 refunds per season.
Gameplay156 comments
3,510
votes
UPVOTE FOR URF
Gameplay606 comments
2,440
votes
Yes plz.
Gameplay456 comments
1,721
votes
3 Refunds per season, up vote if you want 3 refunds per season.
Gameplay114 comments
1,716
votes
When is ANYONE at riot going to address the awful new player experience?
Gameplay634 comments
1,383
votes

More by Hyrum Graff

DoTs can feel impactful if you show how much damage they WILL do as soon as they hit
Gameplay356 comments
825
votes
@meddler @kateykhaos The dragon icons need tweaking for clarity
Gameplay155 comments
299
votes
Can we all agree that towers are in a pretty good spot right now?
Gameplay138 comments
194
votes
DQ has problems. So did the old system. Introducing Club Ranked
Gameplay182 comments
160
votes
ZED CHANGES ARE TARGETED AT FRUSTRATION, NOT POWER
Gameplay213 comments
146
votes
Small Ekko nerf: better counterplay through better ult clarity (QOL for enemies)
Gameplay90 comments
127
votes