Greedy Goblin

Tuesday, November 8, 2016

Possible LoL matchmaking rigging (or major error)

I'm back in Silver 3 after a series of "unlucky" games. As I mentioned many times, I queue dodge when teammates pick champions they have horrible winrate or too few games played. But this didn't work on Sunday, because several of my teammates had less than 10 games combined. I had such players before and assumed (bad habit) that due to long waiting times, bronze newbies are matched into my game. I ignored this as I believed that the other team also has a similar newbies to provide equal team MMR.

But this was backward. On Sunday more people play and indeed queues were shorter. So why I didn't get teams of similar level instead of several "bronze noobs"? Then I remembered the Silver 2 promotion screenshot:
There are 13 Silver 2 players listed. Out of them, 4 (including me) had more than 10 ranked wins. The rest were newbies. Not "noobs", bad players, but genuine newbies who just started ranked games and try to wrap their head around "why there are two LeBlancs here?" and "how did Ashe kill me, she was on the other end of the map?". If you are in Silver, you can check your own leagues in the client. According to the statistics Silver 2 is above the median. So these players were ranked "better than most". Which is obviously wrong, not just because I saw them being completely clueless, but because all players get better with experience. Sure, the M&S won't be great, but will be better than they were. If these newbies are so talented that they deserve to be Silver 2 after one-digit games, they should all be in Gold in a month, which is impossible. The average veteran must be average, and the average newbie must be way below that.

The above is a fact and we can say without doubt than newbie placement on ladders is completely wrong in League of Legends. Now the conspiracy theory about why. Riot makes money from selling champions and their skins. Their problem is that if Joe buys a new champion, he'll suck with it compared to his old champions with multiple games with. So if he is not already in Bronze, he'll likely chain-feed with his new champion until he learns to play. Punishing your customers with series of defeats and teammates abusing them for purchase is not the best marketing. Unlike other pay-to-win games, LoL can't just sell raw power, as it markets itself as an e-sport. If they would give buffs to newly purchased champions for a few games, some math wizard would catch them and there would be a scandal.

So their method of boosting paying players is likely done with these "misplaced" newbies. The paying customers are given wins by placing these "misplaced" newbies to the opposing team. Imagine a game where 2 paying customers and 3 Silver II players with 100+ games are facing 5 Silver II newbies with 4-5 games each. While formally everything is fine (MMR of the teams are equal, the champions and items are equal in-battle), the second team will be obliterated, even if the paying players are useless.

How can they do it? Because LoL is a free to play game, so newbies are in infinite supply. Since "leveling" is pretty long if you don't pay, Riot can be pretty sure who is a paying player and who is a freeloader by the time ranked play starts. So the freeloader newbier can be placed to any MMR as fodder. Sure, they'll chain-lose until they either learns to play or reach their rightful place in Bronze, maybe quitting because of this, but who cares. For every loss they suffered, a paying player gets a win.

As I mentioned, this is a theory with no proof possible without internal leak. But I see no other reason why total newbies with one-digit games are in Silver 2, above the average.

Now what? Shall I just make a "don't play League of Legends" page and move on? Not yet! I will try to outsmart the rigging which is an implicit proof of its existence. Remember that my World of Tanks project concluded when I found a way to earn XP very fast by playing badly. I probably do that in LoL as ultimate proof that they rig their games.

38 comments:

Marek Zaborowski said...

Thats some serious tinfoil there Gevlon. As long as You cant proof it by hard data, I dont think anyone can support You there.

Anonymous said...

No reason? Or no plausible reason?
What if existing players created a new account? (Smurfing, a good number of established players have more then one account)
What if a player from another MOBA started playing League?

Gevlon said...

@Marek: do you have better explanation for newbies placed above the average?

@Anon: and how would the system know after 4 games that he is a smurf?
More importantly: we aren't seeing a handful "newbies" in Silver 2. Most players in Silver 2 are newbies. Are they all Smurfs?

Anonymous said...

Is that only wins for this season? Could they have got a higher rank last season and then returned? I assume there is a seeding factor that affects placement based on results from last season.

Marek Zaborowski said...

Well, I am newbie myself (started playing this summer, didnt play ranked until last month) and I was put into Silver 2. I didnt pay a dim to Riot, before purchuasing some skins -after- I was placed there.

So I think the players You meet this day, was "oh no, I am gonna miss my gold bonuses, better go ranked now". Specially that only like 40% of players plays ranked anyway? And since You had winning streak and Your hidden MMR was through the roof, matchmaking gives You -really- hard matchups where You HAD to carry Yourself. Thats how I see it.

There is no real point in rigging matchmaking though, cause main audience of LOL is playing draft. So its better to focus on more skins, marketing and "friendliness" than rigging anything to make cash. Riot wins for being "friendly".

Anonymous said...

Are you a paying customer, Gevlon? If so, shouldn't YOU be boosted?

Gevlon said...

@Anon: wins for this season. But do you think that suddenly all the "newbies" are returning, outnumbering the veterans 1:4?!

@Marek: and did you have business in Silver 2 back then? Or did you chain-lose your way down to bronze? Matchmaking gave me teammates who literally didn't know what lasthitting is AFTER my MMR went out of the roof.

The point is to reward those who buy skins with the good feeling of winning.

@Anon: no I'm not. I paid them $10 a two years ago when made the account, stopped playing and returned a month ago. Haven't paid since then.

Marek Zaborowski said...

Thats a lot of hassle for little to no gain Gevlon. They risk that Lol would be assiocated with P2W and social dosent likes that A LOT. I mean, You buy skins to make Your peers jealous. If skins would be assiocated with buying power, people would look on skin-users as they are BAD. And that is complete opposite of what makes people buy them.

Remember, that people willing to buy "power" in LOL already have alternative, they are paying good players too "boost" their accounts, so they can show off to their friends. Riot cant really competate here.

Anonymous said...

As with everything the placement system works on 'average'.
Some things that affect a lot: experience from other games, how much one has concentrated on learning vs having fun when leveling to level 30.

From my personal story, the placements put me in my first season to silver 4. After ~18 games I was in silver 2. Did I improve dramatically in those ~18 ranked games to deserve a leap from 4 to 2? Hell no, the placement just put me bit too low.

Yes, there are players who get to s2 after placements and have no business there. However there also people who are placed to b5 and clearly should be much higher. It is just 10 games, there is a big element of random factor. It is meant to work on average.

For some alternative perspective you may google for placement stories from the beginning of this season. Fun stories of Season 5 gold 2 characters going to bronze 4 after promos in Season 6 etc... The beginning of the new season and the placement madness there should give you some material for future posts.

Gevlon said...

@Marek: the gain is "fun" for the buyers. Winning is fun, losing is rage. Their goal is to make buyers enjoy the game and keep playing - therefore keep buying. It's not a rational process "I bought skin and I now won and buy more skins" but an emotional "I won, I love this game, I'll play this game, I buy a skin to look cool in the game I love and play"

@Anonymous: the average is average. My point is that average NEWBIES are placed above the average of the total playerbase which includes veterans. Do you claim that these players on the screenshot will all be in Gold and Platinum after they actually learned to play? Then who'll be in Bronze and Silver?

Also, I wouldn't make a fuss about a handful of newbies in S2 which can be explained by placement error (he got chain luck in placement games) or smurf. I'm making a fuss about the fact that 3 out of 4 players in S2 were newbies while many veterans with hundreds of games are in S4-5. Do you claim that MOST newbies are that good?

Finally: I don't question that season start can be a mess. But it's season END. The leagues should be polished and finalized instead of riddled with clueless newbies.

Pheredhel said...

I think you should just stop your metric of "how many games with this champion".
Why?

- you loose LP, so not worth it
- if you raise your MMR by carrying them, you will see it less, or people more skilled even with new champions
- it is very hard to claim a system is rigged if it gives you odd results while trying to game the system. you are trying to get odd results in the first place! This essentially puts your post as "The system does not give me the advantages I try to gain by gaming the system so it is rigged". The "rigged" may very well be Riots system that prevents you boosting your rank by "gaming the system".

And yes, Riot is easily capable of that kind of protection system. Riot invests a LOT in research to make the game and the players better. Even psychological priming. There is a nice talk about the hints they display before / at the start of the game, and the text and color choices (can't find the video right now, but google "riot priming color text" for a lot of articles on that).

So before calling foul, check your own results by playing by the rules/restrict your strategies to in-game. I am convinced that your in-game rules you have for yourself work. Your out of game systems may just be screwing you over.








Gevlon said...

@Pheredhel: I did nothing to place other people to silver 2. My machinations explain my placement, not everyone else's. Any anti-gaming trick could only explain why am I not in Platinum yet and not why they are in Silver 2.

Anonymous said...

Continuation to last, I don't say that all of them will be in Gold and Platinum. However I think that on average ;) they have higher chance of being placed to Gold+ after they have learned the game.

In my opinion most of Lol players never go much higher than they were initially placed. This means that odds are that if a player was initially placed to bronze, he will never get much higher. Many players have attitude issues which will prevent them from going from bronze to any higher even if they have thousands of games. There are many youtube videos with 'highest mastery points here'. Many of those players with thousands of games with particular champion are in bronze/silver. Thus yes, many 'veteran' players never go any higher.

Assuming there is a problem with having players with less than 10 games in the same rank as players with 200+ games assumes that loads of games => Improvement in LoL. Unfortunately I don't think is the case. Some people never learn nor do they wish to learn.

Gevlon said...

@Anon: nonsense. LoL is mostly a twitch game. No way in hell someone won't get better in lasthitting over time. Sure, I fully agree that someone with bad attitude, who "plays for fun", who starts flamewar in chat, who "I want to play mid so I go second mid" will never climb out of Bronze. But he'll definitely be better than he was a month ago when he was an equally useless moron AND he couldn't even lasthit.

Caldazar said...

Queue dodging is also a way of matchmaking dodging. I wonder if the system is taking it into account. If it is (which is by no means impossible.), all your conclusions will be invalid, or at least not applicable to non queue dodgers.

Gevlon said...

@Caldazar: yes, my queue dodging placed a bunch of newbies to Silver 2. All my fault.

Anonymous said...

Last hitting requires effort to improve on. If you play without trying to improve last hitting odds are you won't really increase your last hitting performance.

As an example, let's take a look at players at
http://www.lolking.net/leaderboards#/eune/150000 (bronze 4) and players at
http://www.lolking.net/leaderboards#/eune/140000 (bronze 3) and
http://www.lolking.net/leaderboards#/eune/110000 (silver 5)

By your theory there should be a clear difference in the number of games played
and players in silver 5 should have on average much greater number of games.
I don't see it in those numbers / players.

Thus based on the numbers (though not representing all) I can only conclude that at least when it comes to Bronze 5 - Silver 5 the number of games played does not predict at all how good you are.

Caldazar said...

@Gevlon: noone said that. But maybe the newbies are queue dodging smurfs, or who knows. But just offhandedly dismissing the queue dodging as a possible factor is bad science.

Gevlon said...

@Anon: Check http://www.lolking.net/leaderboards#/eune/10000 (Platinum 4)
The lowest on the page has 50 wins, half people have 100+ wins. It seems that the top players have lot of plays (as common sense says). We should be seeing Silver 2 halfway between Bronze and Platinum and we don't!

@Caldazar: again: I don't question that smurfs and whatnot could explain a few % newbies in Silver 2. But we are talking about the majority here! Do everyone cheats and queue dodges and such?

Marek Zaborowski said...

I think we would have to have access to Riot marketing data, to check if Your theory makes sense.

Rigging matchmaking in ranked favor of paying customers would make sense only if there is strong correlation between winning games and buying skins. And I doubt there is.

From my experience with LoL social, they are not playing to win (duuuuh, thats the problem with them, isnt it?). They play to be "cool". So when they are carried, but they had shitty game nonetheless (feeded as hell f.e) they are NOT happy. So giving them carriers, dosent make senses. Giving them newbs, they can stomp, does. But, You cant really stomp if You lack skill. Dont get me wrong, but even when enemy team is weak, being a carrier demands basics skills. And if You posses these, there is a good chance You are average silver player. Cause this is silver, the middle-class of the game. Bronze is for players that are below-average.

So no, sorry Gevlon. I dont see a point in rigging matchmaking in such a manner. Socials dont buy skins because they are happy with game, they buy skins because they project -how good- they are in the game. Even more, making social -lose- with champion that he bought tons of skins, makes him think "this champion sucks, I need new one!" (he cant see, that HE is the fault) so he goes and buy new champ and new skins to him.

Its not like I dont suspect Riot for milking money from customers, I just think they have soooo much more powerfull tool than rigging matchmaking to favour paying customers, that is pointless from their POV.

Anonymous said...

Looking at players from http://www.lolking.net/leaderboards#/eune/110000 we can see the following changes in mmr from season begin

2 * same
increase of 100
2 * increase of 200
drop of 100
drop of 150
drop of 250
minor drop
minor increase

=> Thus taking the players who are now in that list and their initial placements (some didn't play ranked until last month), we can say that on average they were in the correct place

If you want it would be very interesting to see what happens statistically to all people who were initially placed in silver 2. Based on that very very narrow sample I have for that silver 5 group, the answer could be that on average they were placed in correct place.

Gevlon said...

@Marek: in order to buy and see/parade a skin, you have to buy and play a champion. What's the point of an Annie skin if you don't play Annie?! New skins are bound to only a few champions. For example the recent and hyped "Devil Teemo" skin needs you to play Teemo. I guess that batch had a Witch Morgana and one more champion. So you have 3 "cool, hyped" champions.

ArthasDKlol buys Devil Teemo skin and wants to show it off. But he never played Teemo - and barely played midlane. So now he queues as mid and picks Teemo. He will horribly suck against a player equal to himself on main. He will feed as hell. The "fair" outcome would be 1/10 wins in the next ten games, 3/10 on the next ten until he finally learns to use Teemo. Instead, he'll get 5/10 wins by placed against newbies, so he gets a boost to be carried over the first hard games.

Gevlon said...

@Anonymous: half of the http://www.lolking.net/leaderboards#/eune/110000 players have less than 10 games. I don't understand how do you define "now in that list and their initial placements" for them.

Anonymous said...

What I mean is that you can click each of the players in the list and look at their performance chart. (Ignore everything before feb 22 as it's season 5 before that, yes I made the mistake in previous comment). They all have more than 10 games played, not 10 wins, but 10 games. But that's beside the point.

You can see how their performance has changed from their starting mmr of the season.
If your theory of low game amount correlating to being overranked is true, you should see a sudden drop in the start of graphic in the performance of people. Also those people who have hundreds of game.

Anonymous said...

The answer to all of this is simple: Monday was the last day of this season and everyone and his mother wanted hastily attain some rank - any rank. This makes huge disparity between skill levels at this time and the whole B5 - P3 seeding area is a big mess.

Again, having very few games in ranked does not make you a worse player than some battle-hardened veteran with lots of games, quite the opposite. It means the system is not quite sure where you belong and you can prove to be way better (or worse) player than the system thinks you are.

Wander said...

Most people think that if they play their skill will improve when actually research on games like tennis shows the opposite. If you are not deliberately trying to get better and focusing on it you simply develope more lazy habits and get Worse with more games. Now I think tennis is WAY more of a twitch game than LoL.

For those interested in research on developing skill there is a book called "Peak: Secrets from the science of expertise". The basic gist is that practise makes you better but simply doing something does not make you better.

For example new inexperienced doctors often preform better than experienced ones.

Gevlon said...

@Wander: so you claim that if I'd go and play Tennis first time, I'd be just as good than after months of practice?! That's crazy.

I fully understand - and myself preach - that the morons and slackers with experience can easily be surpassed by goal oriented players. I myself got more ISK 3 months into EVE than most veterans see after 5 years.

But it doesn't change the fact that everyone improves from his own past performance. ArthasDKlol will suck even after a year, but will clearly suck less than he did on day one.

Also what the hell Riot uses for a new player to estimate his rank? How fast he killed botmatches? How did he fare against other newbies on lvl 15 games?

Samus said...

The main problem I have with this theory is that LoL MMR is trying to give every player a 50% winrate, regardless of skill. If you are better, they put you against harder players. If you are worse, they put you against worse players. If they are "trying to make you lose," it is because your winrate (especially recent winrate) is above 50%.

From that perspective, the MMR is working "correctly" in your case. The MMR is still trying to give you a 50% winrate, and your previous win streak set off flags that you were far better than the players around you. You WERE being put in matches that were more difficult to win, with the assumption that you would be rewarded 20 LP for a win and only lose 10 LP for a loss (or something like that). The game assumes you will move up through points, not more wins.

Also, I must point out that the game almost never sees win streaks like yours that are due to playing smart and doing the right things, rather than flat mechanical skill. When the game sees a player win 15 of 20 games, it is almost always going to be because they are mechanically far above their current opponents. You are an outlier. If you were winning games because you were getting 20 kills on Katarina, you would still have been about to carry your team to wins 50% of the time even with your newfound retarded teammates.

Silver 2 is right at the median. Where else would the game put new players who seem competent? They have only seen a handful of ranked games from these players, but they have to put them somewhere. They could be a Bronze 5 failure or a Diamond 1 smurf, the game has to figure that out with very little information. They do take into account CS numbers, KDA, etc when deciding where to place. But it isn't hard to see how a player who should be in Silver 5 could get lucky and win 8 of his 10 placement matches, so the game wrongly puts him in Silver 2.

99smite said...

If someone wants to statistically prove that a game mechanic works as intended or is rigged, he would need a large data base, which Gevlon does not have at the moment.
As long as the databse is not large enough, any random result might look suspicious.

So, I wish Gevlon will get the amount of data to be able to come to valid conclusions.

At the moment, I kind of doubt that there should be anything wrong with the team-building procedure... Since a lot of players play LoL, I am sure that other people would have already found out if anything was rigged.

nightgerbil said...

"But it doesn't change the fact that everyone improves from his own past performance. ArthasDKlol will suck even after a year, but will clearly suck less than he did on day one.
"
RE your reply to wander, If you aren't trying to improve or if you reached your skill cap you won't gain anything from repetition thats his point and hes right. I'll give an example I have been making cups of tea all my life. Some cups I drink taste awful, some taste good. Depends a lot on the teabag and the quailty of the milk. If I went and made another 1000 cups I wouldn't suddenly be able to make better cups of tea. I'm skill capped.

In your tennis example if you went and played it for the first time and then spent the next 3 months just whacking the ball around the court at random, without thinking about it or bothering to learn tactics or play (and learn from) anyone with a clue about how the game is played, then you would still be just as "good" ie bad. The only thing you would have gained would be basic mechanical skills (ie hitting the ball), but you would still be thrashed by anyone who understood the game. Arthasdklol will always suck at the game. and he will suck after ten years at the exact level he sucked after the 1 month it took him to figure out how to hand in a quest, equip a piece of gear, kill a mob and spam /trade with "gold plz lol"

Gevlon said...
@Caldazar: yes, my queue dodging placed a bunch of newbies to Silver 2. All my fault. : well it might be. Maybe the game is seeing you queue dodge and is putting you with the newbies in the silver 2 bracket. Maybe thats a thing to stop your queue dodging antics? As a scientist to prove anything you have to remove all other variables from an experiment to test something. Thats the only way to get a defintive answer as to wether this drug cures this illness yes/no. You don't find out by giving the patient ten drugs then wondering which one cured him and which one made him go blind.

Maybe try getting back to silver 2 the non queue dodging way and see if its still putting you in with the clueless? if it does you will be able to say defentively that your queue dodging isn't to blame.

Unknown said...

" @Anon: wins for this season. But do you think that suddenly all the "newbies" are returning, outnumbering the veterans 1:4?! "

Yes. End of season attracts hordes of newbies who hope to get gold for the challenger skin (or any other reward), especially as Riot advertises end-of-season rewards in the client.

Smokeman said...

I gotta say, this whole thing looks crooked.

Why would you feel compelled to place people in matches where they have a 50% win rate? If you're awesome, and you start low, you can expect to win a lot until you get to your median point where you would naturally have a 50% rate...

Then, how would it even be possible to leap to the 50% point after 4 wins? The way to run a ladder is to have everyone start at the bottom at the beginning of the season and then let the cream roar to the top. Not just arbitrarily seed people all over the map.

I don't care how well you did last season, how can you get to 50% with 4 wins?

Oh wait. They get paid when you buy skins. And you don't sell skins to the majority of players (That statistically will suck.) be letting them be destroyed in the rankings. People buy skins when they feel awesome because they're winners. "If I just buy that new skin, I'll stop losing.." said no one ever.

That said, I'm going to assume that there is a level where it stops being rigged, and Gevlon just isn't up to that point. Anything below that point is the "All rigged, all the time" show designed to sell skins.

Anonymous said...

A significant number of bronze/silver players play only a few games in a season.

I used to play normals with 2 friends (all of us silver), and we got 1-2 plat players on the enemy team in a large percentage of our games. I knew something was off because we should have gotten a lot less plat+ players if matches were fully random (normal mmr!= ranked mmr) according to statistics. Except those stats show every player who played the 10 games to get a rank. I randomly came across another site with some simple stats but those were very different from the ones on the bigger sites like op.gg. It only showed players who played in the last 2 weeks (or was it 20 days?) and plat+ players were almost the top 20% there instead of the usual top 10%. This was around early summer and forgot the name of the site, but you should be able to check when the low match count people played last. So the median of all ACTIVE players is around silver 1, you should stop encountering newbies mid to high gold.

Marek Zaborowski said...

Gevlon You asked me, what is the point of buying Annie skin and not useing. The answer is: there is none. But since when socials follows reason? Rigged matchmaking is clearly to manipulate lv1 and lv2 players, since lv 3 and lv 4 players thrives for competition that is fair and based on what they do.

Anonymous said...

Educational psychologist here.

First, keep up the content, Gevlon. Love your perspective.

Second, repeating the same task will improve efficacy, not skill.

If you have a poor vocabulary and low fluency, many will tell you to read more.
However, you will only noticeably increase your vocabulary if you are aware of the purpose of your reading. You are noting words you dont understand and looking up their definition and rereading its use in the source material for appropriate context. Or maybe you are spending time to interrogate meaning through the context if the sentence, using syntax as clues, etc.

Many just read. If they skipped over words before, they will continue to do that in each subsequent title. Their fluency will improve at the level they are at, but their actual lexile progression will not.

If you play LoL one way, and you are convinced you are paying right, and blame your failure on external forces without any introspection, you will not get better at last hitting, unless you know it's important and are trying to improve in that domain. You won't get better at positioning or timing unless you are willing to look at why you are failing and how to improve.

You need to recognize that your serve sucks, and play more games focusing on getting the ball over the net, in the box, with speed and spin, before you start acing anyone.

The problems is the M's too quickly call out rules and mechanics being unbalanced and unfair and repeatedly deploy the same failing tactics time and again with observable improvement that could be argued as an increase in luck instead of skill.

The kid struggling to read might not get a better comprehension from the text, buy he will get faster at recognizing words he already knows, so he reads more, faster. His efficiency has increased, but his level remains the same.

LazinessEvolved said...

All joking aside, I played Teemo mid, full AP, as a shock pick in low gold. It works surprisingly well. Most AP mids you face have no idea how to handle it. Once you get shrooms, their pouncy ults become suicide missions.

Anonymous said...

Unrelated game, but recently read about how the game Battlerite deals with the MMR, which is quite interesting.

Rather than just have a flat number, it has a "Mean" and a "Variance", which loosely meet the meaning of those words in statistics.

The number rank you are shown (has visible MMR number) is Mean-3*Variance, and when you have a low volume of games, the variance can be quite high.

This does have some side effects, where if you win a game that you were expected to lose, while your Mean will go up, the Variance might as well, and if 3*variance > Mean change, then your visible number actually goes down.

Overall though it means the more games you play, the more 'sure' the system is of your position.

If LoL is using a similar system (and battlerite didn't invent it, just have explained it well), then this may indicate those low volume of game players have a high variance, so while they are sitting in silver II, they may actually be 'good enough' to be in Gold, but also may be terrible and belong in Bronze, but because they haven't played enough its not really sure.

Low sample volume = high variance, while the high volume = low variance, so actually people with lots of games are going to be closer to Silver II skill, while the people with very few games may actually be smurfs and really good, or the complete brickheads who don't know what items are

Wander said...

Sorry for the late reply.
Everyone when first learning something puts in a lot of effort at improvement but usually they quickly hit a "good enough" point and don't need much effort to get by. That's when people stagnate and think they are getting better when they are not.
When ArthasDKlol first goes raiding he will get better at raiding but after the first two times will cease improvement. Take LegolasXX who is more dedicated and keeps improving for longer. Eventually he will also reach a point where his skills cap out. This is not how good he could be but rather where he needs more focused pratice to improve.
In both cases neither will realize when their skills plateaued. If they care about getting better, a time may come when they realize they are not improving but they often do not know how to improve further.
I'll flesh out the doctor comment. Most doctors have little feedback on if they made the right call. They care about improving but need objective feedback to see what to do differently. The exception is surgeons who DO improve over time because they get feedback. Surgeries are recorded and they watch themselves and their colleagues and can spot mistakes without much trouble and learn from them.
I think the same applies to league. I've often heard people say you should record your games to watch and learn from if you wish to improve. I haven't tried that yet but if I get serious about improvement I will.