The Math of Climbing in League of Legends

JRobin31··50 votes·37,846 views

WIN RATE: a win or loss in LoL counts as approximately 20 LP. If your MMR is higher a win might count more and a loss might count less. If your MMR is lower, a win might count less and a loss might count more. So it takes approximately 5 more wins than losses to gain 100 LP. Most promotions are best of three which takes roughly two games to win and promoting a League is a best of five which takes 3 games to win. Therefore, it takes about 36 more wins than losses to climb one full League.

Win Rate / Expected Number of Games to Climb One League 51% / 1800 55% / 360 60%/ 180 65% / 120

This means that if you are only slightly better than the players you are matched with it can take a thousand games or more to climb, but if you are significantly better than the players you are matched with, then the number of games you have to play drops dramatically. This is the mathematics behind players that say you "belong" at your current rank. The reason you "belong" there is because even if you are "better" than the players you get matched with, you aren't significantly better (you need to be getting more than 55% win rate). But that isn't actually the important take away you should be getting from this mathematics. The important take away here isn't where other players tell you you belong. The important take away is: how much time are you willing to spend to climb. Maybe you do belong in the next League up, but are you willing to play 400 games to get there instead of the 120 it takes some other guy?

INFLUENCE ON THE GAME: You are one of ten players in the game (not five) and that means that all other things being equal each player has about a 10% influence on the outcome of the game. So if your influence on the game is just as profound as the other players, then you are the determining influence for one out of every ten games. Perhaps your Match History looks like this:

  1. opponent AFK -> easy win
  2. teammate AFK -> easy loss
  3. enemy smurf -> easy loss
  4. teammate hard carried -> easy win
  5. team imploded with negativity -> easy loss
  6. enemy team imploded with negativity -> easy win
  7. teammate feeds -> easy loss
  8. enemy feeds -> easy win
  9. close game but teammate picked off death timers OP -> easy loss 10. The outcome of this game was directly dependent on you

This is the mathematics behind players that complain that they can't climb. They often blame teammates for feeding or AFK or troll and they have the feeling that their own skill at the game just doesn't matter. And they get told that they shouldn't blame their teammates. But, of course, it is obvious from looking at the match history that many of their games were losses due to poorly performing teammates. They aren't actually technically "wrong" when they place blame on their teammates. So why shouldn't they blame them? The answer is that blaming random players won't change the mathematics of climbing. These wins and losses appear to be due to random factors beyond control. It is the equivalent of blaming the rain for spoiling your picnic. Yes, the rain spoiled your picnic; you aren't wrong. But you also can't control the rain... so why are you blaming it? You should've anticipated that it could rain and been prepared. So how do you do that? The answer is that you have to increase your influence on the game beyond the normal 10% that everyone is assumed to contribute.

SPENDING YOUR TIME: It should be clear by now that players have to spend time climbing. A player with a 55% win rate has to be committed to playing 300-400 games to see a result. That's actually a lot of time spent playing the game and if most of that time, you aren't the primary factor in who wins, then you need a strategy to deal with the rain.

Mechanics: For this we have to go back to the advice commonly given to players: improve your mechanics. It has been said by top players that getting to the highest ranks in LoL is all about improving your mechanics. Macro game, item builds, Runes, Masteries, blah, blah, blah - all of that is secondary. It's all a clever distraction. It's only important if you actually have the mechanics to back it up. It doesn't matter if teamfighting was the "right" call, if you don't have the mechanics to turn that fight to victory. This advice of improving mechanics comes from coaches that trained in Korea solo queue. It comes from Challenger players that confessed that they didn't think pro play was about mechanics until they entered pro play and discovered that other pro players were doing things they didn't even know were possible. You might think that you've climbed past the point of needing to improve your mechanics, but I'm saying that never actually happens; you can always have better mechanics.

Time and Attention in the Game: So let's say you are playing the game and you miss a cs or... you die. Maybe the chat blows up with every player putting in their two cents about why you suck. And maybe you get irked and type something back or maybe you don't type anything, but you can't get your teammate's negativity off your mind. This is all a distraction - a waste of time. While you are thinking about that missed CS, your lane opponent could land a skill shot and the next thing you know you are respawning on the fountain and you are still thinking about how bummed you are that you died while you chat with your jungler about how he isn't ganking for you and while that happens you miss three more CS because you are trying to play safe and not feed. There is a lot of mental attention being wasted and you don't deserve that win - you got carried. Here's the real deal: the reason for most deaths is obvious. Watch your replay. It's obvious. Most of the time, there isn't a whole lot to think about. If you are chatting about death (your own or a teammate's), then you are probably wasting time. Talking about it doesn't magically resurrect anyone or fix any problems. This is because most deaths are due to mechanical error. If you had simply been paying attention and on top of your game, you would not have died.

What this all boils down to is that you can improve your mechanics all the way to challenger and into the pro scene and mechanics is the primary factor in how fast you climb ranked queues. So how are you spending your time? If nine out of every ten games is out of your control and you are playing at least 300 games to get to the next league, then you have a huge amount of time. If you finish a game and didn't improve, then you didn't change the rate at which you climb.

This is the mathematics of climbing: to climb faster, you need to increase your win rate. to increase your win rate, you need to have more influence on the game. to have more influence on the game, you need to die less. to die less, you need to improve your mechanics. to improve your mechanics, you have to stop wasting time.

TL;DR: Once you do this, you will realize that whether you win or lose a game isn't actually that important. In fact, the mathematics tells us that it isn't that important.

124 Comments

Fondling Gems15 votes

#The Math of Climbing in League of Legends (Memes & Games version)

http://images.mentalfloss.com/sites/default/files/styles/article_640x430/public/156210661.jpg

jaymc11303 votes

[{quoted}](name=Fondling Gems,realm=NA,application-id=3ErqAdtq,discussion-id=zwEa6kq9,comment-id=0009,timestamp=2017-03-07T23:10:44.464+0000)

#The Math of Climbing in League of Legends (Memes & Games version)

http://images.mentalfloss.com/sites/default/files/styles/article_640x430/public/156210661.jpg

This meme is not Laughing Fish approved. I'm sorry sir, but you'll have to pay a fine for subprime meme posting.

Fondling Gems3 votes

However did Laughing Fish ever approve your ability to say that something isn't Laughing Fish approved?

jaymc11302 votes

[{quoted}](name=Fondling Gems,realm=NA,application-id=3ErqAdtq,discussion-id=zwEa6kq9,comment-id=000900000000,timestamp=2017-03-07T23:21:41.051+0000)

However did Laughing Fish ever approve your ability to say that something isn't Laughing Fish approved?

Of course not...

I pay him in favors of an intimate sort to overlook my lack of proper documented authorization.

Fondling Gems4 votes

Could I pay you in favours of an intimate sort to overlook my transgression of LF approval?

jaymc11305 votes

[{quoted}](name=Fondling Gems,realm=NA,application-id=3ErqAdtq,discussion-id=zwEa6kq9,comment-id=00090000000000000000,timestamp=2017-03-07T23:41:31.062+0000)

Could I pay you in favours of an intimate sort to overlook my transgression of LF approval?

I like where this is going.

Kai Hong1 votes

Me too, can I get in on some of this? ;) :P

Eedat12 votes

This math isn't correct. Last season I climbed from S5 to G5 with a 67% winrate in 75 games.

You can't "just assume" that you get 20 LP for a win and lose 20LP for a loss. This assumption makes all of your math completely meaningless

EDIT: actually 65 games considering 10 of those were placements

JRobin312 votes

I give average number of games. 20 LP is very rough approximation. 75 games is not surprising. Math is still good :D

Eedat3 votes

It was actually 65 games because 10 of them were placements. The thing is though your "rough approximation" multiplied across hundreds of games adds up to A LOT. I climbed in about half the number of games that you approximated

AdeBug2 votes

seeing how higher winrate means lower number of total games required (exponentially, not proportionally), and the highest considered winrate in this article is 65%, anything above that would mean less than 120 games required, per avg. (prolly around 60 at a 70%winrate but I could be wrong) I don't see anything wrong with that

jaymc11307 votes

{quoted}

This math isn't correct. Last season I climbed from S5 to G5 with a 67% winrate in 75 games.

You can't "just assume" that you get 20 LP for a win and lose 20LP for a loss. This assumption makes all of your math completely meaningless

EDIT: actually 65 games considering 10 of those were placements

The 20 MMR gain is a standardized measurement that consists of the average rate for gains, it is the same in all elo systems for any game when opponents are of equal skill level.

The one thing op forgot to account for was MMR gain relative to lp gain inflation since the league system and MMR system are two separate things but only MMR is used for the ladder while LP and divisions are merely a fictitious cover sheet. It's very tough to calculate specific MMR gains relative to LP gains with respect to win rate because the two systems are separate and thus the average of 20 MMR is an approximate substitution that rises or falls based on the delta (change) of your win rate.

For further information about the separate systems and how they interact with one another, you can find it here: http://boards.na.leagueoflegends.com/en/c/gameplay-balance/E9nkYATc-psa-regarding-ranked-placement-and-seedings-for-all-who-are-confused

In general, what that information means in relation to the math in this thread is that as your win rate increase accelerates, so too do your MMR and LP gains which will decrease the total number of games required to hit your climb goal thresh hold.

Eedat6 votes

You're missing the point. You can't assume a standard static LP gain in a dynamic system. The end result is false and any assumptions based on that are false as well.

For example, you can easily climb with a 50% winrate as long as you maintain an MMR that is above your current division.

The assumption that it takes 1800 games to climb a tier is a blatant falsehood in this situation

JRobin311 votes

Except that there are players who have taken 1000+ games climb to Silver V from Bronze V! 1800 games isn't that unrealistic if your win rate is really that low. Just because there may be strategies for maintaining your MMR and climbing a little bit quicker or just because you never took 1000+ games to climb from Bronze V to Silver V doesn't make my values terribly unrealistic. You may be getting caught up on the details. I'm explaining the math of climbing by using approximations and you are trying to dispute that by giving an example where it took fewer games.

So... how useful is your information for helping players who have played a couple hundred games and can't climb to Gold V? The answer is: not useful at all unless they want to hire you to boost them.

Does your wonderful 65 game climb to Gold V benefit anyone looking to duplicate your feat? Does it explain why it took longer for some players to get to Gold V? Is it even a record for how many games it takes to get to Gold V? I mean aren't there players who have done it in even fewer games?

So how does this climb to Gold V in 65 games really invalidate what I said? Is this the common experience? Do most players who are Gold V take 65 games to get there? Is 65 games a reasonable expectation?

Eedat4 votes

Thats nice and all but again you're missing the point. You're just assuming one way or the other. Are there people who took 1000 games to climb out of bronze? Yes. Are there people who took 500 to do the same? Yes. How about 100 games? Yes. Does all of their winrates directly coordinate to how fast they climbed? NO.

You're just picking specific examples that agree with your point while completely ignoring anything that doesn't

jaymc11302 votes

{quoted}

Thats nice and all but again you're missing the point. You're just assuming one way or the other. Are there people who took 1000 games to climb out of bronze? Yes. Are there people who took 500 to do the same? Yes. How about 100 games? Yes. Does all of their winrates directly coordinate to how fast they climbed? NO.

You're just picking specific examples that agree with your point while completely ignoring anything that doesn't

This statement is why I said to revise the ops original "climb a tier" to "climb to true skill level".

If you use that definition you can account for all players in all ranges and all outlier scenarios. The math says so, the volume of statistical data on the subject is immense and irrefutable.

Not that the OP doesn't have some minor flaws in his premise, he does, but the basic gist of it is pretty spot on.

Let's do an example with you, if you wouldn't mind. Would you feel bothered if I link your op.gg and use your last season as an example?

Eedat4 votes

You can if you want. Just for some extra info I only played 2 games once I hit Gold (1W-1L) and I went 7/10 in my placements I think

FROScountered1 votes

{quoted}

This math isn't correct. Last season I climbed from S5 to G5 with a 67% winrate in 75 games.

You can't "just assume" that you get 20 LP for a win and lose 20LP for a loss. This assumption makes all of your math completely meaningless

EDIT: actually 65 games considering 10 of those were placements

These wins and losses appear to be due to random factors beyond control.

So control your pants befor it explode [slayer-jinx-wink]

jaymc11305 votes

Hey there, jaymc here, resident ranked systems expert and tester, 4 time diamond and former master tier player.

Just popped into the thread to comment on how similar your numbers are to my 5 seasons worth of collected and saved data regarding the function of systems. What you describe here I would say is fairly accurate and some truly quality content to be sharing, many people should find this stuff useful as tools to plan and enact climbing strategies.

I will say that these figures are only accurate for accounts that have held a previous rank in a previous season since all fresh accounts receive bonus MMR gains for the first 50 lifetime ranked matches and this can skew the numbers some as less games will be required to climb the same distance.

Overall this is a pretty informative post though and I applaud your diligence in information gathering. Well done mate, these figures are a little on the high side for total games required to climb but easily within the bottom range limits supported by my own years worth of spreadsheet data. Good job on the thread.

BluePolarizer4 votes

Macro is very important actually. The #1 reason why players lose is not because of their mechanics. It is because they got caught at the wrong place at the wrong time, backed at the wrong place at the wrong time, etc. Mechanics should be a given. You should learn to cs with minimal brain usage. And if they aren't a given, you need to be playing Malphite and Annie only every single game until they are a given.

This is not true in silver but if you want to get to plat+ you need to have good macro. Good mechanics is necessary but not sufficient.

Amak3 votes

The best guides I've read suggest different things that ultimately poke holes in this theory.

The first is when to play. Playing at times of the day when children are more likely to play is a bad idea because they are more likely to make bad decisions from champion choice to in-game tactics. Playing late at night is a bad idea because often you're playing with people who are from different time zones and potentially different language backgrounds, so ping and communication are both issues.

The second is when to dodge. Almost every high-ranking guide I've read describes the art of dodging a likely loss because the LP loss is minimal while preserving your MMR.

In short, it's not simply about improving your mechanics. My own stats show this. In S6 I climbed from Silver to Gold almost solely on the back of Zac Jungle. My KDA is 5.32 and my winrate is 60% with him. I played 174 games from, literally, Silver V to Gold V. I made into Gold just days before the season ended.

S7 is a little strange because now there are two queues, but I'm Silver V in both currently (thanks Riot for "lowering the ladder" after I finally get into Gold). Even so, I'm at 4.50 KDA, 50% winrate, and 104 games with Zac Jungle. The difference? I play vastly different hours of the day this season because I got a different job.

Even so, my personal performance is largely unchanged. I still have a solid KDA with Zac, lower largely due to losing more games rather than my own personal mechanics, yet my winrate has plummeted. Strangely, despite the general disapproval for Flex, I've found that Flex feels more like S6, and I am climbing Flex much faster than Solo. I don't even want to play Solo anymore, because my winrate there is closer to 40%.

Personally, I don't think there is any one factor more important than others. I still do what I can to improve personally, but the tone of this article implies that the fault is solely with individual mechanics, which is patently absurd.

GamerThink2 votes

I highly doubt "higher mmr=more points on win", or this would mean the system is totally flawed! XD Probably you meant the other way around! ;)

Snowman Arc1 votes

It does work like that. On my bronze smurf, I get 26 LP for wins and lose 14, while on my gold main, I get 23 (not a big difference, since I maintain a good win rate in both which means my MMR is probably high in both). Still you can see the small difference in win rates can affect the LP gains. On the smurf, I only need 4 wins to get to promos, where I need 5 on my main.

GamerThink1 votes

Bronze=lower MMR Gold=higher MMR

Is there something I should know?

Snowman Arc1 votes

Yes. A silver guy can have the same MMR as a bronze guy, depending on their recent performances in games. A 30% winrate silver 3 guy in his last 100 games might have lower MMR than a 70% winrate bronze 2 guy.

GamerThink1 votes

Ok, sure, but more or less the MMR depends on your position, thus if your team has lower MMR than your opponents you should get less LP, am I right?

Incarceration1 votes

Meanwhile I turned a 48% winrate to a 59% winrate after 126 games, but I still feel like I'm stuck in Silver.

jaymc11302 votes

{quoted}

Meanwhile I turned a 48% winrate to a 59% winrate after 126 games, but I still feel like I'm stuck in Silver.

You're only stuck if you choose to be. There are tons of subtle tricks to help artificially manage your win rate and combine this with the obvious other improvements you've been making and you might maintain or even improve that win rate past 60-65%.

It sounds cliche, but optimizing your play sessions can make a big difference.

Ultimate Deadpoo1 votes

Your math of the game thing is wrong.

Lets assume your playing mid lane. Each of the other two lanes are out of your control and therefore have a 50/50 chance of doing well or feeding. Simple math means that 25% of the time both of the other two lanes will feed and lose, meaning unless you are really doing well along with your jungler your proboably gonna lose 25% of your games. 25% of the time the other two lanes will do well and you don't need to do anything really amazing, so 25% of the time your gonna win. The other 50% is determined by you, so the math actually favors you if you are in a solo lane.

This is backed up by the fact the best players in league tend to have just above a 75% winrate with their champions, as they are able to win the games they influence directly, win the games they get carried on and sometimes carry games they should of lost.

The worst players have around a 25% win rate for the opposite reason.

I'd do the math for the jungle, adc, and support but that math is kinda complicated and I just know that for those roles you have a bit more RNG in what happens.

JRobin311 votes

I agree that different roles have a different amount of relative influence on the game, but... why repeat myself five times?

That said, the win lane - win game method of calculating influence isn't really that accurate. A common complaint players have is that they win lane but can't win game which directly suggests that they aren't having this outrageous 50% influence. Players that can make 75% win rate are players that have powerfully developed their influence over the game (or happen to be smurfing) and although I talk about players having a 10% influence on their games, I don't mean that this is in fact how much influence a given player actually has. For example, some players attribute their wins to the development of powerful shot calling and communication skills. The professional NA LCS player known as "Hai" is an example of a player who appears to able to boost the ability of his team to win through shot calling and communication. It doesn't appear to matter what role he plays (jungle, mid, support...). He influences the outcome in a powerful way that many other players don't. The closest other comparison I would make is probably "Xpeke" in the EU LCS who led his team to several victories when he played as a sub in roles he never played professionally before and didn't really have time to practice. The wins were attributed to his experience and shot calling and the overall influence he had on his team.

This sort of influence isn't necessarily about the role a player plays and it certainly isn't the result of the chances that lanes win or lose. Two different players playing in the same role can have a very different amount of influence on the game and that influence goes beyond them "winning" or "losing" their match up and it isn't just 50-50 chance of doing well or feeding in the other lanes anymore!

We don't really know how much influence a given player has on his games, but strong players don't settle for having just as much influence on the outcome of the game as any other player has (if all players had the same amount of influence then everyone would have 10%). Strong players are looking to increase the influence they have on the game and they can leverage that influence to achieve higher win rates.

Chikapi1 votes

I actually agree with most of what you said. In terms of MMR/tier, location isn't important. direction (as well as speed and acceleration) is.

One thing I would point out as a diamond support main is that, macro game and shot calling is more important than a clever distraction, especially for less mechanics-demanding champs (which exists in every lane). For instance:

  • Learning the route to roam without stepping on wards as a midlaner
  • When and where to tp as toplaner
  • Knowing the possible lurking positions of enemy assassins as adc, and direct your team to peel accordingly
  • Needless to say, how to ward/counterward as support

All these things are pretty much mechanics independent, but if done correctly, basically gives you "easy mode" on the execution part.

While diamond players can carry @ silver/gold with mechanics alone (and quickly), when playing against players of your level (or similar), getting an edge before the skirmish/teamfight can mean a lot.

JRobin311 votes

I guess I just need a better way to talk about learning from the games you play. I talk about it as mechanics, but maybe that's not the best way to describe it. I see pathing as a part of mechanics but it's become clear that many higher level players just don't see it that way. They see it as the macro game and appear to regard it as essential - even more essential than what are considered to be mechanics to the point where they see Gold level "mechanics" (or even Silver level mechanics) as sufficient for Platinum (or Platinum level mechanics as sufficient for Challenger) if your "macro" game skills are where they need to be and that even very skilled mechanical players can't make it to Platinum if their macro game doesn't get to par.

The point is that if you want to use your time efficiently, then you need to be using it to improve regardless of whether or not you win or lose.

If the win or loss was out of your control and I'm saying that most of the games you play are out of your control (insert flipping coin meme here), then what were you really spending your time doing? Patting yourself on the back and/or flaming teammates? Unconsciously affirming your unconscious mistakes? Or were you really improving your game? And I suspect that really good players are always improving their game even though the outcome of most of the games they play is out of their control.

SzGamer2271 votes

It's not actually as simple as this. MMR is a big part of the League ranked system; You will always be matched with people with the same MMR as you, so your win rate will never be above or below 50% for long. As a result, climbing (or falling) comes down to whether your MMR is higher or lower than your current League. If the difference between your MMR is such that you earn 25 LP for winning and lose 15 LP for a loss, you can expect to gain an average of 5 LP per game, resulting in an average of 20 games needed to climb 100 LP.

JRobin311 votes

That's certainly an interesting way to look at climbing (and I agree that it is not that simple), but if you want to look at things that way, you still have to toss MMR out the window and look at it in terms of LP losses and gains. Players can't see their MMR; they can only infer it. LP gains and losses are visible to players. And the question remains as to how your way of looking at climbing actually helps us understand climbing. Can we draw any insights from that way of looking at it?

Can it give us an estimate for the number of games played to advance a league? Well, first you have to figure out some additional relationships between MMR and LP gains and losses. This could get very messy, but it could also give us better estimates than the ones I projected. That's totally legit mathematics: you have a method and you apply it. I don't talk about error and that's a messy problem too. To be really legit I ought to talk about error. So my apologies.

Just blindly applying your method, we might say 5 LP gain per game is a good estimate. Then we would say that means 20 games per Division. Then we would say, the system is trying to give you 50% win rate so you gain about half a win per promo game so we could estimate 4 promo games for a Division and six promo games for a League. So to advance a League, a player would play about 205 + 44 + 6 = 122 games. Actually we would want to account for various players so we would wind up with a table like this:

avg LP Gain / # games to advance one League 1 LP / 522 games 2 LP / 273 games 3 LP / 186.7 games 4 LP / 147 games 5 LP / 122 games

The assumption is that player aren't skipping divisions, but it's also probably safe to say that you are going to be seeing much larger average LP gain if you are skipping divisions.

So we can say that if your average LP gain is 5 or more, then you are going to see progress relatively quickly, but if your average LP gain is 1 or less, then you are in for a looong climb. From there you could draw the same conclusions I wrote in the OP, but it just wouldn't be as elegant. LP gains are skewed during the first games players play so they can't use those early average LP gains to estimate their long term climbing speed, but maybe it's good estimate for short term climbs?

The elegant thing about my method is that win rate is completely transparent to the player and MMR is not. I'm not saying my method of approximation is the most accurate approximation there is, but it is a useful approximation that helps us understand the process of climbing in ranked.

I think that if I wanted to talk about "climbing to your true ELO", then I might want to take the MMR approach because the meaning of "true ELO" is sort of tangled up with this idea of MMR. We don't actually know what a player's "true ELO" is and that's why I talk about climbing a League instead of climbing to "true ELO". The League system is what RIOT actually shows players and what actually determines end of season rewards. "true ELO" can neither be seen nor does it determine end of season rewards. Also players aren't trying to climb to their "true ELO". Players are trying to climb to a certain rank usually a certain League (i.e. the next league up or Gold V). They have to believe that they belong there even if their MMR is terrible and they want to see evidence that they do belong there. Maybe a third party site can provide MMR estimates, but win rate is more accessible.

The MMR issue is relevant. You aren't wrong! Don't think I'm telling you that you are wrong. Players often get told on the boards that the way they see things is wrong (okay often times it is wrong), but it isn't always wrong. Sometimes it's just a different way of looking at things.

SzGamer2271 votes

The problem I see with your method is that the system is actively trying to get all players to a 50% win rate, and whether you are climbing steadily or falling very quickly, the system is going to match you into games where you have a 50% chance to win. If you try to estimate how long it will take you to climb, you will see that after playing a large number of ranked games your win rate is going to consistently be 50%, giving the illusion that your climb is stagnant or at least impossibly slow, when it is still just as dynamic as it always is and you could very well be rising or falling at that moment.

LP gain and loss is just as transparent as win rate, and since the system's way of normalizing LP gain/loss is by directly adjusting your ranked placement (rather than win rate which is normalized via MMR, a number completely unrelated to your rank), you can get an accurate picture of how fast you're climbing whether you've played 5 ranked games or 500.

Here are the simple rules. If the difference between your LP gain and loss is 0, then the system is telling you that your MMR aligns perfectly with your rank, and at that moment the only factors in climbing or falling rank is going to be randomness, which means you can expect not to climb or fall. If you're earning more for winning than you are for losing, that is encouragement to continue playing ranked games, as the system is nudging you to a higher rank. If you are losing more LP than you are gaining, you've lost your edge or were somehow placed at a rank that is above your MMR, which means that you should not play ranked and try honing your skills in normals before coming back if you don't want to drop a division.

These rules hold up at all stages of your ranked climb. If you have a spike in win rate that drives you to a higher MMR, you will see a short-term increase in your win rate as the system works to adjust your MMR to normalize your win rate, but then your win rate will plateau once more and revert back to 50%. However, the MMR increase caused by that spike in win rate persists into the long-term, so even after your win rate returns to 50%, you will notice that you're still gaining more LP for your wins than you are losing for losses, which means you're still climbing. This climb will not cease until the player achieves a rank that matches their MMR, and LP gain/loss is the same once again. Meanwhile, your win rate gives no such indication of your climb, and all you can infer is that your "lucky streak" has ended and you're no longer climbing.

JRobin311 votes

The biggest problem with what you've said is that the system doesn't actually match players according to win rate. It doesn't take a player with a 65% win rate and match them with a player with a 35% win rate in an effort to make their win rate be 50%.

The system uses the MMR ("match making ratio") to determine match ups. So while win rate tends to 50% over a large number of games, the system isn't actually trying to sabotage player win rates.

Personally, I've found LP gains and losses method to work well for short term predictions (as in how many games I need to get to the next Division), but that it has been a poor predictor for long term climbing (as in the number of games needed to get to the next League). This may be precisely because win rates generally aren't 50% and because MMRs (and the associated LP gains and losses) fluctuate a lot.

MMR estimates are eventually useful as predictors of a player's true ELO, but they aren't as useful as you think they are for predicting the number of games needed to climb a League precisely because you make the assumption that win rates are 50% and... they aren't.

SzGamer2271 votes

{quoted}

The biggest problem with what you've said is that the system doesn't actually match players according to win rate.

I think you are operating under a critical misunderstanding.

This is exactly the flaw I see with what you've said.

As you just pointed out, MMR fluctuates a lot, but the thing is that MMR fluctuates with win rate, as a change in your win rate causes a change in your MMR. A change in your MMR in turn leads to a change in your rank, but MMR directly affects how your rank is adjusted, while win rate does so indirectly by adjusting MMR (which, in turn, adjusts your rank). Since MMR's affect on climbing can be observed by taking the difference between your LP gained versus LP lost, you can avoid the complication of a middleman that MMR causes when you try to make predictions on ranked climb from win rate.

I'm going to make this clear now before proceeding: You cannot make long term predictions on your ranked climb in any case or by any means. Since the player's skill, experience, and mental state is constantly in flux, which causes numerous and unpredictable short-term changes in win rate, MMR, and so forth over the course of a ranked career, it is safe to assume that an accurate assessment of ranked climb over the course of hundreds of games can never be accurately assessed, and any indication that this might be false is likely caused by sheer fluke. Short-term predictions are therefore the most you can hope to accurately assess, and the most effective way to make short-term predictions is to look at what directly determines your ranked adjustments in short term, which is LP adjustment after matches.

With the context of short-term assessments now in place, allow me to elaborate on the flaws I see in your method. Estimating ranked climb by looking at your win rate operates under a number of unrealistic assumptions. The biggest of these offending assumptions is that your MMR will stay unchanged for a large number of games if your win rate is anything but 50%. On the contrary, the system is designed to assess your skill level and match you in an environment where other players match your skill level, making the game fair. If you are on a win streak, that win streak will be short lived, as the system is going to increase your MMR to match your elevated skill level, thus returning you to a 50% win rate.

Thus, assuming that MMR is static when win rate is in flux causes inaccuracy in other areas as a side effect. I've identified two of these side effects in the following paragraphs.

Firstly, your win rate will never be constant unless your skill level is rising or falling faster than your MMR is being adjusted; you will never have 330 games in a row with a 55% win rate unless you are rapidly getting much better at the game and consistently improving so that your MMR never catches up to your actual skill level, and the notion of maintaining such consistent and incredible improvement for such a long period of time would be absurd.

Secondly, your LP gain/loss will not be constant or predictable if you have a win rate that strays from being 50%; if your win rate is significantly higher or lower than 50%, your MMR is in flux, which means that the amount of LP you gain and lose cannot be generalized to be a certain number, and the approximated 20 LP gain/loss is especially inaccurate as LP gain/loss are only equal when your MMR is the same as your rank and you settle at a 50% win rate.

To convert win rates into average LP adjustment over time as you do in the main post, you operate under the assumption that MMR does not change when you consistently win more than 50% of your games, and thus that you can maintain an above-average win rate for an extended period of time, and that your LP adjustments will remain consistent at 20 LP gain/loss for a win/loss. These assumptions are, at best, hard to believe.

On the contrary, no such assumptions are made when you consider only your LP adjustments. If you consider only how much more LP you gain for winning than you do for losing, you don't need to make any assumptions at all, and since LP gain is the purest and most direct factor in ranked adjustment, estimating or extrapolating based on MMR will be consistently more accurate. Thus why I see the difference in LP gained/lost as the best indicator of ranked progress.

JRobin311 votes

My method actually doesn't care about MMR. MMR could be static or in flux or whatever. It actually doesn't matter. That's why the method is better at predicting.

  1. MMR is harder to track. If you want to use MMR to predict, then you first have to have a method of estimating MMR. There are third party websites that will give players estimates of their MMR.

  2. MMR is about where a player currently stands. MMR changes with every game and always represents where a player currently stands (not where a player will stand or climb to in the future). It doesn't predict a player's increase or decrease in MMR. Players can estimate their true ELO based on MMR at the end of last season before MMR resets.

  3. Win rate has to be taken over a large enough sample size and as a statistic actually represents a trend in the player's climbing. That makes it actually potentially useful as a predictor in the first place. If your win rate is at 50%, then your MMR has probably stabilized and thus you aren't climbing regardless of the method you use to predict.

Maybe we can make a predictor using last season's rank and MMR to predict how many games it will take a player to reach that same rank and MMR this season. It may even be a better predictor. That doesn't make my prediction based on win rate a bad prediction. We still need data: lots of data, much more data than a simple win rate predictor.

I said before that I don't regard the win rate prediction as the best possible predictor, but I do regard it as one of the more useful predictors out there because it's simple and anyone can actually use the method without visiting third party websites.

LP Gain/Loss that isn't based on MMR is also useful as a predictor because you shed the dependence on MMR and go directly to LP loss and LP gain. I won't say LP loss and gain is a better or worse predictor because there are benefits and drawbacks to both methods. For example, LP losses and gains can't predict the outcome of promos.

SzGamer2271 votes

{quoted}

My method actually doesn't care about MMR. MMR could be static or in flux or whatever. It actually doesn't matter. That's why the method is better at predicting.

That's just the problem. Your preferred method ignores MMR, when MMR is the main determining factor in adjusting and determining rank in a system that is based on matchmaking.

This is why the win/loss method is inaccurate. Win rate does not determine rank. Win rate determines MMR which determines LP gained/loss which determines rank.

The transformation of win rate into ranked looks like this:

Win rate -> MMR -> LP -> Rank

If you look only at win rate, you disregard MMR and LP gain and loss, which are both vital parts of what determines rank. The uncertainty of MMR has a lot more impact on ranked progression than win rate does, because win rate will always average out to be 50% in anything but the short-run due to matchmaking.

The point I've been arguing since my very first post has been tat predictions based on win rate are not accurate, and watching how much LP you gain and lose after matches is not only more accurate, but it's just as easy to use. You already recognize that looking at LP gained/lost is a good method of prediction; I'm just trying to show you that win rate is not.

Regarding your listed off points one by one:

1-2: You don't have to track MMR. You look at the LP you gained or lost after each ranked game and work out how much more you gain than lose. All you need to know is that if you gain more LP for winning a match than you lose LP for a lost match, your MMR is higher than your rank, and you are climbing the ranked ladder. You don't need to see or know exactly what you're MMR is. If you're worried about long term-predictions, you're never going to get anything very accurate, because the system is so dynamic that patterns only ever hold in short-run in any case.

3: Win rate is just as dynamic as MMR, since they're both exerting constant change on one another. Looking at your win rate over your entire ranked career means nothing because there is no consistency in a system with matchmaking. If being placed above 67% of players literally meant that you had a higher win rate than 67% of players, looking at your win rate would be perfect. But since win rate, MMR, and LP gain/loss are all dynamically shifting from game to game to create fair matches, the closest you can get to an absolute predictor is LP gained/lost.

When you only have to worry about LP turning into your Rank, everything becomes simpler and more accurate.

Regarding the handling of Promo Series when predicting rate of climb:

Since your MMR is constantly adjusted to keep you at a 50% win rate, you have a 50% chance to win each Promo Series. Since the nature of winning or losing Promo Series is purely random (one of the major flaws of the current ranked climb system) you have to factor them in as they come. Divisional Promo Series add an average of 2.5 games to your ranked climb whenever they come up. League Advancement Series add an average of 4.125 games. You have a 50% chance to pass on your first try, but if you fail you have to add another Series in addition to the games you have to play to get back to 100 LP (Promo Series can really mess up your ranked climb, and it's all random, so predictions become harder to shape and less meaningful).

I'd like to destroy the notion that long-term predictions are at all possible if I can. The ranked system in League makes anything beyond short- to medium-term predictions hard to obtain and vague at best.

JRobin311 votes

Ah, I see. You don't think that making these kinds of estimates are useful or meaningful for players in the first place. So let me explain why a long term prediction is useful in the first place.

So, I hop onto league of legends and I play a my provisionals and let's say I get placed in Bronze III. Maybe that seems kind of low to me, but I was Silver V last season and I'm expected to be placed lower and have to climb. In fact, Bronze III is obviously lower than my actual rank. So now the question is: how much time do I need to spend to grind back to Silver V? This is the sort of question thousands of players are actually wondering the answer to. If it's going to take a thousand games, then maybe I won't push super hard to get to Silver V. But if I can get back in just a few hundred games, then maybe it's worthwhile. In short: can I get there before the end of the season? This is why estimating how many games who have to play to get to a particular rank is so incredibly useful for players to have some idea about. There really isn't any question that having some method of predicting the number of games you need to play is useful to players.

So if you really are just trying to argue that making predictions is useless, then you are just straight up wrong. Predictions are always useful. Players need to make decisions about how much time they need to spend or can afford to spend playing. Making that decision is a vacuum of information is worse than making that decision based upon some sort of information. Is there any guarantee a player can reach a certain rank if they play enough games? No, it's always an open question. But the better question is how much time and energy are you prepared to put in to get there?

With that out of the way, let's get back to the other issues you are raising.

  1. You say win rate is just as dynamic as MMR. Guess what? It isn't.
  2. You say win rate tends to 50%. I say: you have to get there first.
  3. You also say you can't maintain a win rate over a large number of games. I say: players maintain win rates over large numbers of games all the time. There is even a post in this thread giving as example of it. In fact, I say you have to play at least a certain number of games to get a decent idea of your win rate because the outcomes of matches are confused by games you can't control the outcome of.

In short, your understanding of win rate is lacking.

Okay, but is win rate really better at predicting the number of games needed to advance?

As I've pointed out: not necessarily. Of win rate, average LP gain, and MMR,

MMR is the worst predictor. This is because is MMR dynamic. If I look up MMR on a third party website, it's pretty much going to say that my MMR places me more or less at about my current rank plus or minus a Division. It's a horrible predictor because it is designed to tell you where you are currently and not designed to tell you where you maybe going. You can't really use your current MMR to predict how many games in the long term. At start of season, my MMR reset somewhere closer to 1200. It blatantly doesn't exemplify my actual skill** because I'm supposed to grind back to my actual level**.

Average LP gain is a great predictor. LP gained (or lost) moves you closer or farther from your next promotion (or demotion). And it takes a small number of games to estimate your average LP gain! I see LP gain as a middle of the road predictor because while you can get a quick estimate, the average LP gain also appears to be heavily influenced by the system initially. That means that LP gain might be harder to maintain in the long term and will fall off as a predictor. Moreover, LP gain doesn't predict at all how many game you will need to play to promote. For the most part average LP gain will get you to your promos, but not through your promos (and no it isn't 50-50, if it's really 50-50 then newsflash: you aren't actually climbing, you are actually stuck to within a Division or two of your rank).

Win rate may be the best long term predictor of the three mentioned because it can be used to predict not only the number of games needed to get to a promo, but also the number of games needed to get through a promo. Win rate can't be estimated accurately in too small a number of games though. You will have to commit to playing at least enough to get a decent estimate of your win rate. That is a drawback to using win rate as an estimator. But good news, if you have an obscenely high win rate you won't need to use it as a predictor!

But will win rate accurately predict how many games it takes for you to get from Bronze V to Challenger? That's a really long term prediction and I would agree that it is outside the scope of these simple prediction models. If you want to use win rate to predict, you should use it to predict over the course of a more reasonable climb - for example climbing to the next League - and not for predicting your climb from Bronze V to Challenger or for predicting your climb to your true ELO.

So given that win rate does have limitations as a predictor. What can we do?

The answer is that in order to take prediction to the next level, we would have to move beyond simple theoretical prediction models and enter the realm of hard data prediction models. What if we gathered the data on the number of games players actually played to reach certain ranks? The things we could do with that are far superior to the academic discussion we are having here.

If you are trying to say win rate can't be used to predict, then your effort in that regard is futile. Win rate is a predictor. If I ask you what the chances are that I win my next game and you say 50-50, then you are straight up wrong. It isn't even debatable. The only real question is how well win rate can be used to predict. How accurate is it really? I haven't answered that question. This is a simple model based upon my experience playing and reasonable estimations and I agree that it is subject to error. It remains a useful predictor for players thinking about how much time they want to invest playing League of Legends... (the most useful of any of the simple models discussed here I might add).

If they want a better more accurate model, then such a model can be created (if it hasn't already) out of the enormous amount of currently existing data on players. The outcomes of millions of games has been recorded and could be analyzed to give us much better predictions. I daresay we could have reasonable predictions of how many games it takes to go from Bronze V to Challenger if someone was willing to put in the work analyzing the existing data.

SzGamer2271 votes

Okay okay okay, calm down.

I did not say that making predictions is useless, I said that long term predictions in a highly dynamic ranked system tend to be vague, as variability to to randomness and constant change makes it inaccurate.

MMR and Win Rate are both equally dynamic, as they both influence each other. An increase in win rate above 50% causes MMR to be increased, which in turn causes win rate to decrease once more.

You don't seem to grasp the implications of a system that matchmakes you with players of your skill level. Win rate is not constant, and if it is constant, it's because it is at 50% and MMR and win rate are at equilibrium. If your skill level is not accurately represented by your MMR, you win rate will change, then your MMR will change, then your win rate will be nudged back to 50%. Nothing is constant in this system, not even your actual skill level, which increases as you get better at the game and decreases when you're on tilt or off your game.

I did not say that win rate couldn't be constant. In fact, I said:

{quoted}

Your win rate will never be constant unless your skill level is rising or falling faster than your MMR is being adjusted.

In most cases however, a Silver 5 player in Bronze 3 will start by being matched with Bronze 3 players, have a higher than average win rate, have their MMR increased to match that of equally skilled Silver 5 players, then rise through the ELOs as they earn more LP for matches than they do for losing. They temporarily experience an above average win rate, but to create more fairness in matches, MMR is adjusted to achieve a 50% win rate for all players, and climb continues by LP adjustment rather than win rate.

Everything about the ranked system tries to oppose the notion that you can maintain a win rate that strays from 50% in a majority of cases.

Now, the only reliable and accurate long-term predictions would have to involve knowing all the invisible variables behind the dynamic fluctuation of the ranked system. How and when is MMR adjusted with respect to win rate, and how much is it adjusted by? How exactly does difference between ELO and MMR translate into difference between LP gain and loss? What win rate does a player in X MMR have when matched with players in Y MMR? With all of these factors nailed down, you could mathematically figure out an average for exactly how long an player placed in X ELO would have their MMR shifted from X to Y, how much LP they would gain during the transition, and finally how many more games it would take for ELO to climb to come into equilibrium with Y MMR.

The tools we have do not allow such long-term predictions, but as you have said, someone with the power to gather all of the information required could definitely make accurate long term predictions.

I would urge you to shift your focus from trying to make long-term predictions to using a series of short-term predictions one after another to gauge your ranked climb. After all, it's a ranked ladder, which means you should climb it one rung at a time. ;D

JRobin311 votes

To really talk more about this we would have to get estimates on how many games it takes for the system to accurately estimate your true ELO using MMR. If you are climbing... the system doesn't actually know how far up the ladder to move you. It just moves you "one rung at a time" so to speak. If you have exceptional performances you can skip divisions.

I want to harp on this notion of 50% win rate again. Win rates aren't 50% while players are climbing. If win rates are 50%, then the system thinks your MMR is about where it should be and your rank is a reflection of your MMR. The system doesn't oppose the notion of maintaining greater than 50% win rates and it just fine with you going on losing or winning streaks that hurt or benefit your rank. If anything the system is okay with adjusting your MMR upwards every time you win a game!

Is that unfair? If win rates aren't 50%, then what about creating fair matches? Well, the notion that fairness in matches is the result of your win rate is flawed. The matchup is only actually fair if the system has accurately measured everyone's true ELO already...

That's why we would need to explore the number of games the system takes to accurately rank players (using MMR). You appear to believe that this process takes place very quickly. I believe that the process takes a variable amount of time depending on the player, that in general it doesn't happen too quickly, and that the system never actually represents your true ELO but rather only gets you very very close.

For example, let's say that we played a tournament with five million LoL players. It would take about twenty rounds of single elimination games to decide on a set of top five players. Twenty games! But what about the other players in the tournament?

Well, in order to rank them, we would have to play multiple eliminations and do a lot of bookkeeping, but, in general, it would take more and more rounds more than twenty as we went through the process of figuring out where each player should be. The bottom five could be figured out in about twenty games but the middle of the pack would remain an enigma for many many more games.

This example illustrates how a ranking system can take a variable amount of time to rank players. In a few cases, the ranking takes place quickly, but in the majority of cases, the ranking is much slower.

So how quickly does LoL rank the majority of players?

SzGamer2271 votes

Win rate is higher while you are climbing, but you can't make predictions based on it because it's changing.

You have a 60% win rate. You're climbing by winning more than losing. The system recognizes that you are playing better than people of your rank, so it adjusts your MMR to make matches more fair.

After a few games you now have a 55% win rate. You're climbing just as fast as before, because LP gains are now increased to compensate. You're still a little ahead of the players you're with, so it continues to adjust your MMR.

By the end of your climb, your win rate is 50%, and the rest of your climb is left to LP gains and losses getting you the extra step of the way.

Was your win rate above 50% for that climb? Yes. However, your win rate was consistently changing even if by a small margin each game. Making a prediction that assumed you stayed at a consistent win rate is less accurate as a result.

Additionally, when you estimate your win rate by looking at past games, you're going to get results that are outdated, because your current win rate is ever lower on the slope while your past games look at the average of the previous ground you've covered, causing an overestimation and skewing results of a prediction. There is a balancing act that you have to perform: Too few games analyzed results in a large amount of variation and therefore low accuracy, but to many takes into account too many games that have a win rate that is too different from what you are experiencing currently to remain relevant.

As to your metaphor about a tournament bracket system, this is incompatible because it's an entirely different system. That's like making an assumption about apples by looking exclusively at oranges. Sure they're both fruit, and brackets and Elo are both ranking systems, but you're not going to get anywhere trying to argue that apples can yield orange juice.

The very structure of the two systems is completely different. In brackets, you play out a number of matches with people who have won and lost the same number of games as you until you play out all your games and end up at a final rank. In an Elo system, rank isn't as solid, and you gain and lose rank not only based on winning and losing but also on how your current Elo matches against your opponent's, so you don't have to be equal in skill or rank to your opponent to be matched with them and your final rank is adjusted accordingly. League of Legends has an extensive player base however, and it can afford to place people of similar Elo against each other to make matches more fair, but that doesn't mean the systems are still fundamentally different. For example, a player can be matched with another player with 200 more games of experience and the system will operate just fine, because an Elo system does not necessarily require every player to have played the same number of games as their opponent.

And yes, it takes a long time for win rate and MMR and such to settle, which is why it's so hard to make predictions. If you could magically ascertain these things based on a single match, you could be a lot more accurate, but you have to constantly make predictions based on information that gets more dated with every new game, taking the average altitude of the slope behind you as you continue to levels you haven't yet reached.

JRobin311 votes

Time to clear up the half-truths again...

{quoted}

Win rate is higher while you are climbing, but you can't make predictions based on it because it's changing.

Truthhood: win rate is higher when you are climbing Falsehood: you can't make predictions based on it because it is changing

You have a 60% win rate. You're climbing by winning more than losing. The system recognizes that you are playing better than people of your rank, so it adjusts your MMR to make matches more fair.

Truthhood: you climb by winning more than losing Truthhood: the system adjusts your MMR to make matches more fair

After a few games you now have a 55% win rate. You're climbing just as fast as before, because LP gains are now increased to compensate. You're still a little ahead of the players you're with, so it continues to adjust your MMR.

By the end of your climb, your win rate is 50%, and the rest of your climb is left to LP gains and losses getting you the extra step of the way.

Falsehood: by the end of your climb your win rate is 50% Correction: your win rate is approximately 50% Truthhood: the rest of your climb is left to LP gains and losses getting you the extra step of the way

Was your win rate above 50% for that climb? Yes. However, your win rate was consistently changing even if by a small margin each game. Making a prediction that assumed you stayed at a consistent win rate is less accurate as a result.

Irrelevant? "is less accurate as a result" less accurate than what? than making no prediction?

Additionally, when you estimate your win rate by looking at past games, you're going to get results that are outdated, because your current win rate is ever lower on the slope while your past games look at the average of the previous ground you've covered, causing an overestimation and skewing results of a prediction. There is a balancing act that you have to perform: Too few games analyzed results in a large amount of variation and therefore low accuracy, but to many takes into account too many games that have a win rate that is too different from what you are experiencing currently to remain relevant.

Redundant Truthhood: when you estimate your win rate by looking at past games, you are going to get results that are outdated

As to your metaphor about a tournament bracket system, this is incompatible because it's an entirely different system. That's like making an assumption about apples by looking exclusively at oranges. Sure they're both fruit, and brackets and Elo are both ranking systems, but you're not going to get anywhere trying to argue that apples can yield orange juice.

Metaphor was an example to provide insight into how ranking systems function in general.

The very structure of the two systems is completely different. In brackets, you play out a number of matches with people who have won and lost the same number of games as you until you play out all your games and end up at a final rank. In an Elo system, rank isn't as solid, and you gain and lose rank not only based on winning and losing but also on how your current Elo matches against your opponent's, so you don't have to be equal in skill or rank to your opponent to be matched with them and your final rank is adjusted accordingly. League of Legends has an extensive player base however, and it can afford to place people of similar Elo against each other to make matches more fair, but that doesn't mean the systems are still fundamentally different. For example, a player can be matched with another player with 200 more games of experience and the system will operate just fine, because an Elo system does not necessarily require every player to have played the same number of games as their opponent.

None of those structural differences is relevant to the point being made.

And yes, it takes a long time for win rate and MMR and such to settle, which is why it's so hard to make predictions. If you could magically ascertain these things based on a single match, you could be a lot more accurate, but you have to constantly make predictions based on information that gets more dated with every new game, taking the average altitude of the slope behind you as you continue to levels you haven't yet reached.

Truthhood: it takes a long time for win rate and MMR to settle

Conclusion: We seem to be going over the same points again and again.

To summarize your position: win rate can't be used to predict the number of games it takes to climb with any accuracy because win rate is changing

To summarize my position: win rate is the easiest way to estimate your chances of winning the next game you play and it takes a long time for win rate to adjust to 50%. therefore, it can be used to make an easy estimate of the number of games you will need to play.

Our positions aren't mutually exclusive simply because we don't have any statements that express the accuracy of the prediction method. Our difference is academic and based on intuition.

I use the win rate as an estimate for climbing one full League. You argue that that's just too long term and that nothing beyond climbing a Division can be estimated with any accuracy.

My response is that there's really no point in estimating to get to the next division when what you are really interested in is getting to the next League. So although it may be nice to have "accurate" estimates of the number of games to get to the next Division, the question players really have is how many to get to the next League. I regard more than a League to be too long term and only getting to the next division to be too short term (unless you are climbing in Division I to get to the next League). The reason for that is that end of season rewards are for the most part tied to League - not Division. The reason I regard that estimate as not too bad, is the absence of other more accurate estimates.

We agree that it would be possible to obtain more accurate estimates using a different method. However, that other method would not be as simple and would require more knowledge about the system.

Are you going to make another post saying that win rate can't be used to estimate the number of games needed? We've been over it many times now. In the absence of a better estimator, it is the best estimator. My position on this won't change unless we come up with a better way to estimate the number of games. In which case, I will happily adopt that method over this one.

Ðelofasht1 votes

I think the math is very good, and reflects my climb from Silver this and last season to mid platinum. This season I will probably get Diamond if I continue to win with 55% on my main champion, some of my other champions do not have that kind of win rate, but that is okay.

Like many others have noted, it is my belief that game knowledge and strategy are more important than mechanical skill. I frequently see players with excellent mechanical skills stuck in various ranks. Generally, what they lack are that knowledge and strategy, with which they would rise far beyond their peers. Mechanically speaking many champions are limited in how far you can take them, but we see challenger players on basically any champion, because more of the game is based on other elements that mechanical prowess.

Most low ELO games have that kind of good mechanical skills, but terrible strategic planning due to a lack of game knowledge. It is more frequent that advice is given to positioning and objectives than, "practice your mechanical skills because they are lacking" and with good reason. One can always improve their game knowledge, but your mechanical skills will only shift very subtly after a certain point.

Meira31 votes

Good shit man, i specially like the part when you talk about you can't blame your teammates if you lose, i totally agree.

JRobin311 votes

Thanks! I feel it's important to understand that when players are complaining about their teammates, there is often a grain of truth buried in their mountain of salt. I hope that some players will read and be a little less mountain of salt and a little more grain of truth. They are right! But they are also wrong! At the same time! Math is beautiful!

AlienPrimate1 votes

You sort of explained it, but forgot to actually do the math for one thing.

Randoms on my team? 4 Randoms on enemy team? 5 Me's on my team? 1 Randoms in my game? 9

Chances of trolls/feeders/AFK's on my team? 4/9 Chances of trolls/feeders/AFK's on enemy team? 5/9

You are 11.1% more likely to have a troll/feeder/AFK on the enemy team as you are on your team if you are never that person.

JRobin313 votes

Okay. Math is hard. So let me fix things up for you:

You should've said that is the chance of getting a bad player is, p, then the chance of getting at least one bad player on your team (besides yourself of course) is: :( = [1 - (1-p)^4] And the probability of getting at least one bad player on the enemy team is: :) = [1 - (1-p)^5]

So... how likely is this? Let's suppose that half of all players besides yourself are bad If p = .5, then :( = 15/16 or 93.75% and :) = 31/32 or 96.875%. Is that really all that significant a difference? Let's look at the difference in general which would be: D = :) - :( = [1 - (1-p)^5] - [1 - (1-p)^4] = (1-p)^4 - (1-p)^5 solving 0 = D' = 5*(1-p)^4 - 4*(1-p)^3 yields p = 1 or p = 1/5 at p= 1 we have a minimum difference of 0 at p = 0, we also have a difference of 0, at p = 1/5 we have a maximum

At p = 1/5, :) = 2101/3125 or 67.232%, :( = 369/625 or 59.04%, and Maximum D = :) - :( = 256/3125 or 8.192%

But how useful is this result? Is it realistic to say that one out of every five players is a troll/AFK/feeder? Because as we can see if more than one out of every five players is a troll/AFK/feeder, then the difference between the chance for your team versus the chance for the enemy team actually goes down! We all desperately want to believe that if you aren't a troll/AFK/feeder, then you will be able to make the difference in your games. But the reality is actually closer to this: you don't really have the amount of control over your games that you think you have (or think you should have) and that's the reason you blame teammates to begin with. If you actually had the control over most of your games, then you would mostly have to blame yourself for any losses. This is an unrealistic view.

**TL;DR Saying there are five randoms on the enemy team and four randoms on your team is just the tip of a massive iceberg. It's fairly meaningless to argue that because you are the one non-random variable, you have significantly more control over your games. If anything the nine random variables show that you have little control over your games. ** If you want to have more control over your games, then you need to improve. It doesn't matter if you are correct that trolls/AFKs/feeders determined the outcomes of most of your games. RIOT can't make the players you are matched with better players.

Zerodusk1 votes

This was an awesome post, i really enjoyed/agreed with everything said here. I still find myself getting into arguements with people though... got to find that mute button quicker lmao.

More from Gameplay

More by JRobin31