Greedy Goblin

Do not play League of Legends!

I started playing League of Legends believing that it's a skill game and wanted to do a project that focuses on strategic decisions instead. I used the undertuned and no-skill old Warwick to exclude personal skill from the research. It didn't work because I bumped into what most players call "ELO Hell". Games seemed to be decided outside of my actions. I mean before the 10th minute a teammate or an opponent got 8 kills and no deaths thanks to the feeder in the opposing lane. In itself it didn't seem a problem, it should be a random effect with your skill being the deciding factor (the enemy has 5/4 more chance to get a moron feeder), but I started to notice that I get more than my fair share of feeder idiots.

Then I realized that the system is rigged and developed a counter that allowed me to could climb to Gold (top 1/4 of the ranked players) playing only old Warwick and then Nunu when WW was patched while being lowest damage of the team most of the games

Finding the rigging started by analyzing the teammates on, trying to catch possible feeders to dodge the game. I assumed that players who have a history of defeats are bad and I should just dodge. To my surprise I've found that many-many players have no history with their chosen champion at all. This is my history with Aatrox:

Yep, nothing, since I've never played Aatrox. I've seen this lot of times. So I started dodging games where many of the teammates were like this or with horrible winrate with their champion. It's common sense, if I'd pull Aatrox at the first time, I'd suck horribly. Despite dodging 3/4 of the games, my winrate didn't get much better and I didn't get free of idiot feeders. Seeing that I'm missing something fundamental, I calculated the winrates of my games if the various kinds of teammates were missing or only 1 was present versus if there were 2 or more:
What the hell? I'm winning more games when I have baddies? And when average players are present, my winrate drops? Note: Good is 60%+ winrate with chosen champion, Bad is below 40% or no history, Average is everything else. I stopped dodging games to have more baddies in team and plotted my results versus "bad - average" (which can be between all Bad +4 and all Average -4):

This was crazy enough. Why would having a teammate with history of defeats be a predictor of wins? So came the final piece of the puzzle: separating Bads from New champion users (who had no games with that champion like me with Aatrox or only 1-2, not older than 2 days). I assumed that if someone playing a champion for the first time, he'll suck, so belongs to baddies. After the separation the Bads predicted defeats, as they should. On the other hand the presence of teammates who just pulled a new champion is the best predictor of a win. Why? How? What can I do about this?

The first question is the easiest to answer: the "new champion user" is the buyer. If you purchased a new champion or a skin to a champion you haven't used for long, you want to play this champion! Sure, it's not a 1:1 connection, you can get new champions for free from shards and you can buy skins for your main. But we can agree that someone who just pulled Aatrox for the first time has more chance of a recent purchase (likely buying Aatrox) than someone who played Aatrox 50 times before this game. The company wants to reward paying players. If they win, they are happy and more likely continue to play and continue to pay.

"How?" is trickier. There is a simple but limited way: give the other team AFK-ers, DC-ers and toxic players combined with kids who go troll when insulted. This is a sure win for the buyer team, this is why Riot refuses to ban or "prisoner island" them, they are great tool to rig games. Here are some nice examples. However such players are in short supply outside of Bronze and low silver (hint: half of the players are in this range). The troublemaker himself loses a lot and placing Bronzes into a high Silver or Gold team would be quite suspicious. There is a mask for that: Unranked. Players during their placement games can show up anywhere without being obvious, but they are in limited supply compared to all players. Still, I initiated over 20 Remakes and the enemy initiated 8 during my play, so it's clear that AFK-ers are more likely placed to my team than to the enemy (since I'm not a buyer).

But the above dirty tricks account for only a small part of the games. Most of the rigged games contain normal players and equal rating on both sides. The buyers just pulled a new champion they have zero experience with, so they will definitely play below their usual performance. How can their team win? At first let's see the winner side. This chart is one of my latest, created after I fitted formulas over the lase 298 games with no exceptions. I've found that if (New - Average > 1) or ((New > 0) and (Good - Bad > 1)) then the games have high winrate. I called these games "Easy Wins". Let's see how the player distribution looks like (scale is number of teammates I've met):

It's pretty obvious what's going on here: in Easy Wins the team is buyers and boosters. Before you'd ask, the winrate of games which have Good players but not classified as "Easy Win" is only 50%, so Good without New is not predicting a win. And this is after I figured out to beat the opposite side. Back when I played normally, Good players outside of Easy wins had below 40% winrate.

Now let's see the opposing side, because the game doesn't just set up a buyers+boosters team and throws them into the wind. Remember that buyers are bad, so they are prone to to feed. The game needs to guarantee not only that they can be carried but also that they don't have a frustrating game: winning because a teammate is 20/0/0 while you are 0/10/0 and being called "feeder idiot" isn't the experience Riot wants for the buyers. So meet with the "Sure Loss" team formula: (Bad - Average/2 - Good/2 - New >-1) or ((Good > 1) and (not Easy_Win)). Let's see the player distribution and for comparison the fair games (which aren't Easy Wins or Sure Losses):

The surprising thing here is that Sure Loss games are not dominated by baddies, most baddies are placed in fair games! Both Goods and especially Averages has bigger than fair share in Sure Loss games. Remember, without exploit mode, these games are 80% lost. It's not by mistake, the baddie is not a feeder, because he is aware of his noobness and plays defensively. He is usually placed in a lane facing the the buyer, to guarantee that the buyer won't be farmed. The Good and Average players are the feeders. The matchmaker sets up a game where they are facing a better player, whose main counters their main. Believing in their l33tness, they engage again and again and - hello 1/10/3 Yasuo! The point is that the new champion users can easily get 1-2 kills behind their lane opponents, it doesn't matter. When the laning phase ends, the fed booster smashes the opposing team. I noticed that when I'm in such games, I often face junglers who play way better than me and when I was playing normally, often I was the feeder. Not anymore since I realized how to break these games.

One particularly nasty form of the above is Autofill. The player has proper MMR, but he doesn't play his lane well at all. It doesn't matter if he is Gold with his top and mid lane performance when he is filled to bot or support! Please note how many times someone complains about his lane and someone swaps with him. If the matchmaker wasn't screwing with the team on purpose, they would have been swapped automatically. Also, there are good industry practices for role unbalances: WoW gives out goodie bag if you queue up as tank, Riot could do that too. There could also be different queue lengths for lanes, if someone insists to go mid, he could wait half an hour. Or simply rework support to make it more fun: for example let them lasthit and their support item just redirects 80% of the gold to the nearest teammate, so a good supp could double kill and not steal farm from the ADC. They keep support hated on purpose so they can use autofill to give wins to buyers.

How do people climb out from ELO Hell? One way is by being very good, this case they win fair games by a large enough margin to compensate for rigged defeats. Please note that this needs 70+% "fair" winrate, so you don't just have to be better than people in your league, you must be much better. The other way is taking the advice of "pros" and learning lot of champions. Then you'll win and believe you are winning because you learned lot of champions. The truth is that you are winning because you purchased lot of champions. Remember: you win when you are New champion user. After you learned the champion, you become Good and no longer have good winrate without exploit.

How can you break this? Well, if you got DC-er, AFK-er, unranked, toxic+snowflake combo, you usually can't. But that's the minority of the set up defeats and they are not so common in the upper half (Silver 2 and above). In most set up defeats your team is better than the enemy. You are stacked with mediocre players who are facing with boosted wallet warriors who are literally playing their first game with their champion. You should be smashing them and you usually are ... except in one or rarely two lanes, but in that lane your teammate feeds like hell. The point of exploit mode is to break laning phase and force teamfights as soon as possible. How? Suggest invading enemy jungle at start and gank + push a lane which is not yours! If you are jungler, pick one lane and babysit+push. If you are laner, roam to other lanes as many times as you can to gank. Suggest dragon fights. It doesn't matter if you lose your lane, if you are 40 minions behind, if everyone calls you an idiot. Every minute your team spends with normal laning is another kill for that Katarina or Ahri or Annie against your retarded Yasuo. Force teamfights! Then the 2-3 yet-not-fed booster can't carry 2-3 clueless wallet warriors. Important: the role of "1/10/3 Yasuo" can be cast on you by the matchmaker. If your laning enemy outplays you once, note this and avoid further engagement. Buy armor/resist or simply run from him. He is selected to be better than you, you won't beat him!

I tested exploit mode in a very aggressive manner: queued in as support and picked Warwick, smite and went for second jungler. Of course I lost almost all "easy win" and "fair" games. But I won most of the "Sure loss" games since everyone was running around as headless chicken and fights happened in random locations where my average teammates defeated the wallet warriors.

So the checklist of climbing fast:
  1. If any of your teammates whines about being autofilled and not helped out by someone swapping, consider it a Sure Loss
  2. Check every teammate on
  3. If you see an unranked and you aren't in Bronze where unranked belongs, consider it a Sure Loss
  4. Classify the teammates into New (no games with selected champion or just 1-2 not older than 2 days), Good (over 60% winrate), Bad (below 40% winrate) and Average (everyone else).
  5. If (New - Average > 1) or ((New > 0) and (Good - Bad > 1)) then the game is Easy Win (unless lane-whining or unranked teammate), enjoy. Focus on your lane and smash your enemy, he is selected to be countered by you.
  6. If (Bad - Average/2 - Good/2 - New >-1) or ((Good > 1) and (not Easy_Win)) than the game is Sure Loss
  7. If the game is Sure Loss, do everything to shorten the laning phase. Leave your lane and push another.
  8. If the game is neither Sure Loss, nor Easy Win, it's fair, just play
However my advice is simpler: just uninstall this damn rigged cashgrab! If free players like you don't give free wins to buyers, either because you quit or because you are exploiting, the system will collapse.

Finally, the nastiest thing: the outcome of your next game should not be affected by your last. OK, if you got mad because of a loss, that increases the chance of chainloss. But I've found the opposite:
The reason is that the matchmaker wants to prevent excessive frustration, so after a defeat it's more likely to give an Easy Win and after a win it's more likely to give a Sure Loss. This way the rigging is less observable to the human eye (you don't get an AFK-er or a 0/10/0 mid game after game) but statistically detectable. Why is it particularly evil? Because if you win with your skill, the game throws more Sure Losses than Free Wins at your way, pulling you back. If you lose because you suck, it throws you more Easy Wins to push you up. It's not just that buyers get free wins and everyone else get free losses. Everyone are pushed to the median. This is the ELO Hell.

Little extra: if you are still in doubt and don't want to run extensive research, just do one thing: at every champion selection toss a coin and if it's head, dodge the game. This should do nothing besides getting you 30 mins lockouts and -10LPs. In reality, your winrate will tank, because the matchmaker is also used to punish dodgers and toxic players. The dodger will believe that he loses because his dodging formula is wrong, the toxic will think "players who insult teammates lose 16% more games" and stops dodging/toxic. Which in itself isn't a bad goal. Just another evidence that the matchmaker is anything but fair.

You can comment on the announcing post.