Win Percentages in Ranked

trueApollo·Aug 4, 2014·2 votes·903 views

Hey guys,

I'm wondering if your win % affects who you are matched with in ranked games. I have searched the boards and the ranked FAQ and can't seem to find an answer. I ask because i have over a 50% win ratio, but am matched with people below a 50% win ratio every ranked game I play (I'm speaking of people on my team and on the enemy team). That seems counterintuitive, they should be matching you with people who win roughly the same number (percentage) of games as you, isn't that the purpose of the ranking system? Especially if I have a ratio that says i win more than i lose, i think that should be taken in to account.

Anyways sorry for the rant, just wondering if anyone knows. Thanks!

11 Comments

ThoroniulAug 4, 20141 votes

I have not had a problem with win ratio affecting my games. I have been at a steady 53% win ratio for a while. The things that effect whether you win or not really are how you play the game. Going for objectives is more worthwhile than fishing for kills. Go for dragon with your team as soon as its up. And be aware of where the other team is.

I know that sounds obvious and is probably not even helpful. What determines if you win or not is not some number representing some past games, but how you do in the current game you are facing. Don't be intimidated if you fight people a few tiers above you. Chances are they are about the same skill level. If you happen to fight platinum or diamonds outside of ranked, it is not much different. Pick your fights, get objectives, and ward.

The more matches you play, the more accurate your mmr will represent your skill level.

trueApolloAug 4, 20142 votes

Thanks for the reply! I agree 100% that it is more about how you play the game than it is about win percentages. The only reason I asked is because if you look at the logic, you and I win more than half our games (>50% win ratio). If we are matched with people who win less than half of their games (<50% win ratio), the math would say that we have a lower chance of winning those games.

I'm mostly curious because like i said, that seems counterintuitive in a ranking system. The ranking system should say, "these 10 people all win more than 50% of their games, let's group them together," and "these 10 people all lose more than 50% of their games, let's match them together." Hope that makes sense.

I'm sure this is due to something that I am unaware of in the ELO system. Just trying to find out what that is :)

ThoroniulAug 5, 20141 votes

It could be just coincidence that you are matched up with people of lower average win rate. Also correlation is not causation, just because you see a pattern does not guarantee it is actually there. Also take a look at the opposite team, they might also be in the same situation.

If you take a significant amount of data points like 60% of your games of 1000 games you were matched up with people below the .5 ratio. Then you could argue a stronger argument.

trueApolloAug 5, 20141 votes

I love looking at data, let's give this a looksee. I'll use my last 10 games as an example, as there is no way to pull the data from more games than that. I will pull this data from the League website or lolking. For reference, my win percentage is 54.9%, so i will not list it below.

Here my team's win percentages for my last 10 games. I am omitting the other team as that does not coincide with my argument that players who win more than 50% of their games should be matched together, but are not. Feel free to look up the enemy team if you'd like.

I will list only the player's win percentage is so as not to name names.

Game 1: Loss My Team 51% 51% 49% 48%

Game 2: Loss 49% 49% 49% 51%

Game 3: Loss 49% 49% 48% 47%

Game 4: Win 48% 47% 50% 46%

Game 5: Loss 48% 47% 51% 49%

Game 6: Win 48% 43% 52% 49%

Game 7: Win 48% 47% 49% 49%

Game 8: Win 49% 48% 51% 58%

Game 9: Win 45% 50% 48% 51%

Game 10: Win 52% 49% 49% 51%

So from the data above, you can see that from a 10 game data pool, i am ranked with at least two out of four teammates with below 50% win percentage 10/10 times, in most cases (6/10) three of my four teammates are below 50%.

I know it may seem like i am "splitting hairs" so bear with me, but even a 49% win rate still means you lose more than half of your games. Using these games as a reference, 28/40 players i was matched with have <50% win ratio.

I understand that the more you play, the more accurate your ranking becomes. This thread was not created as a compliant. On the contrary, what i am looking for is to understand how the ranking system decides who to place on your team.

ThoroniulAug 5, 20141 votes

I kind of like talking about this stuff, and im going to throw random thoughts.

First of all to answer your question of how it works. It just makes the team average mmr very similar. so it will be team1_mmr ~= team2_mmr.

a 49% win rate still means you lose more than half of your games.

This statement is a fallacy. a 49% win rate means **on average ** you lost more games than you have won. I was taught to be a stickler to how things are stated. This number is blind to what will happen, and a win rate will change with every additional data point added.

These are numbers, and they do little to asses(assessment? can't spell :( ) a persons skill level, and tell absolutely nothing about how a person will improve.

Its not the most perfect thing. I personally think that riot can perhaps improve the queue by looking at more things than mmr, like competency level at the game split into tiers. I would not judge a silver person on the same level as a platinum. The current system is just a large wall one has to climb, and once they are there they will be matched in more competent games where everyone knows how to team fight and stuff.

If you have a win rate higher than .5 and hold it constantly, then you will get too a higher tier. Though the question is how long.

I have been thinking about it for a long time before this thread, and what matters is not the win rate, but the significance of the win rate. I am just writing some incomplete "rules" describing this problem

I wrote this while thinking about it.

n//number of wins r//win ratio

round(abs( n(1-2r)/r ))

//I can't write more complex notation in this chat so i can't write it completely in one summarized line.

This essentially gives wins - losses

so 100 wins with a 50% win rate will be 0 300 wins with a 52% win rate will be 23 1000 wins with a 70% win rate is just impossible at 571

There clearly is some metric where we can say its impossible(nearly not to see that its impossible, but just so improbably that its absurd to think its possible)

I would also consider winning 10 ranked games in a row improbable.

I would consider having a win-loss difference of 20 really good. This is the point where my personal bias would consider this person skilled. When doing a game I have evaluate my opponents to guess how " skilled " my opponent is with win loss difference along with their initial game play and lvl 6+, and 8+ game play.

I kind of have an odd way off looking at my opponent because if I assume they play anything like I do, then initial evaluations are useless. I purposely play like an utter noob, and make mistakes. It does not matter as long as my two initial objectives are complete. Take minimal damage, take no turret damage. It does not matter if i miss early farm because I will catch up in it anyway. Though it usually is better to farm if you can. Its only in extreme cases where I would opt to giving up all farm.

Wins and losses fluctuate rather evenly.

ThoroniulAug 5, 20141 votes

if r >= .5 round(abs( n(1-2r)/r )) else if r < .5 -1*round(abs( n(1-2r)/r )) end

trueApolloAug 5, 20141 votes

This statement is a fallacy. a 49% win rate means on average you lost more games than you have won

Agreed, I could have stated this more coherently, but you are correct.

This number is blind to what will happen, and a win rate will change with every additional data point added.

These are numbers, and they do little to asses(assessment? can't spell :( ) a persons skill level, and tell absolutely nothing about how a person will improve.

I think this is where we might disagree somewhat. While your win percentage does not dictate how you play in your current game, it is the only measurable metric (barring KDA, which is not used to calculate anything in the current system) that portrays how you have historically played in games. I believe your wins/losses are what Riot uses to calculate your MMR, so while these things are not mutually exclusive, your win percentage is indicative of how you have played historically.

My original statement still stands true through all this. If you have won >50% of your games on average, it is counterintuitive in a ranking system to match you with people who have won <50% of their games on average. Your chances of winning mathematically decrease.

ThoroniulAug 6, 20141 votes

I think this is where we might disagree somewhat. While your win percentage does not dictate how you play in your current game, it is the only measurable metric (barring KDA, which is not used to calculate anything in the current system) that portrays how you have historically played in games.

It may be an indicator, but i mean it quite literally. It contains nothing about what will happen. Sure it can be used to predict things to a degree, that is what statistics was made for.

trueApolloAug 6, 20141 votes

The problem with saying win percentage as an indicator has no meaning is that If you remove the weight of this indicator, the whole League ranking system falls apart. It may just be an indicator, but it is the metric that is used to place you in the ranking system.

I think what you said earlier is the core of this thought process. Assuming you maintain a >50% win ratio, the current League ranking system is NOT and indicator of how skilled you are, but is, in your words "just a large wall one has to climb, and once they are there they will be matched in more competent games where everyone knows how to team fight and stuff." I believe this is true.

Again, my thesis holds strong. Match >50% people with other >50% people, and vice versa.

trueApolloAug 16, 20141 votes

I'm actually kind of tired of trying to defend the ranking system. It doesn't make sense more than half the time, and i doubt someone would be able to give me a logical explanation for what you see below. This was one of my last games:

My Team win %: 54% (Me) 50% 44% 45% 47%

Enemy Team win %: 50% 56% 50% 50% 52%

What is interesting to me about this is that the other team had 2,131 games total combined, and my team had 725 games total combined. They had almost exactly 3X the amount of games we did!

They've played more total games, meaning that the ranking system should be better at matching them with appropriate skill levels.

I'm sorry, but there is noting i can think of that would explain why those two teams were matched together. I read somewhere that the longer you wait in queue, the less optimal the games are. I was in queue for <30 seconds for the game above.

Micavity CatAug 4, 20141 votes

its not only how well you play the game but your teammates to. been finding myself in matches where i come out 8-2 and we lost because my teammates all hat 6-10 deaths lol