Greedy Goblin

Thursday, January 26, 2017

League of Legends is rigged

Since the Warwick update ended my data collection, here is the report. Comment on this post.

Also, I made a dumb mistake. I should have used an anonymous LoL account to play and test and then I could publish any statistics and theories when I got them without fear of devs messing with it, since they couldn't figure out which account is me (of course I had to not mention Warwick or exact games, just aggregate statistics until I finally quit the game). Anyway, that can't be fixed now, I won't get a new computer and a new ISP just to continue testing a game that I know to be rigged (they obviously have some alt detection mechanism that uses machine IDs and IPs).

I didn't uninstall yet, I played some normals with the new WW. The new E and W are great. New Q is somewhat annoying. New R is very annoying, but I guess that's what the "skillz" kids like.

29 comments:

Baelnor said...

Great result and probably one of the reasons I stopped playing too. The matchmaker is just that, making matches with predetermined outcomes

Gurkengelee said...

Great research Gevlon. I will now go and "peer review" it by using your formular and your exploit mode. I will notify you if I can reproduce your results.

Gurkengelee said...

Follow up question: You only classify "New" players into those with a new champion. But the main selling point of League are skins. So would'nt you have to account for these buyers seperatly? You cannot check recent skin purchases on op.gg though.

Doktor Jeep said...

I am willing to bet this algorithm is common in F2P/P2W games. I have "felt" the effect described.

Let's face it: There is no trust in any such game system like this. Once there is monetary motive to rig the game, all bets are off.

"Gold-ammo'ed" is a thing.

Raziel Walker said...

I think most online games are rigged in a similar manner.
World of warships I get the same, after a few losses with moronic teammates you get a few wins because the enemy team is filled with potatoes.

Gevlon said...

@Gurkengelee: I have no data to determine who is a recent skin buyer. But champion usage correlates with that too. After elementalist Lux was out, Lux was everywhere.

Anonymous said...

Haven't read yet the whole post. You didn't describe how you define whether a champion is new.
Op.gg provides also champion stats for normal games. Ie. if a player has played 20 normal games with a champion and then plays ranked one, he is surely in a different position to one who has played no normals with that champ.

Gevlon said...

Yes I did: if op.gg returns nothing or only a few games (including normals and custom). It might surprise you, as it seems obvious to first play normals with a new champion, but most people don't do that. They just try their new champ on ranked.

Gurkengelee said...

First impression: This works like a charm. Besides a couple of fair games I also got a game where our Swain had no games with his champion. But our jungle, top lane and ADC were considered "good". While Swain played very careful and went 0/1 we stomped the enemies on all other lane. They surrendered at 20.

Afterwards I got a jungle main who had 60% win rate in the jungle. But he was autofilled to mid and complained loudly. Jungle wouldn't swap and he locked in Lee. Naturally my team started tilting in champ select. Lee was also up against a 70% win rate Yasuo mid main. So I locked in Blitz support and took every chance to roam to mid like a retard. Between being camped by the jungler and myself yasuo ended the game at 3/13 and we won handesomly with a 20 kill lead. My team kept spamming surrender throughout the whole game though.

maxim said...

So i've read your report and i still think you are assuming malice and greed in situations that can be easily explained without either.

Suppose i were to write a team-matchmaker and were to make choices on how it is supposed to put teams together. Initially, i'd write a simple dumb version that would simply put people together by their ELO. Then i would add a simple algorithm on top of that, that would try to balance cumulative team ratings. And then i'd see an issue - people who switch champions (for whatever reason) are not performing on their ELO level, and thus not only suffer themselves, but cause their teammates to suffer, too. So, what do i do?

Given what i already have in place, a simple solution requiring neither greed, nor malice, presents itself: compensate for fresh-hero skill-reduction with slightly stronger teammates.

I then run into yet another weird problem. As it turns out, on some levels of the ladder, players with slightly higher ELO are significantly better, due to some quirk or another of how ladders have chaotically formed. As a result, Gevlon's phenomenon where good and average players become feeders appears. However, i now reason that
1) introducing even more crutches into the system is bound to create even more problems like that, so i just don't see the profit of messing with the system further and leave it be
2) good and average players could stand to lose a little to superior players in somewhat controlled low-headcount fights and would only benefit from the experience before getting in real 5v5 smackdowns of higher leagues

As a result, the system remains as is.

Your advice on how to play against teams that have 2 noobs and 2 gosu that out-ELO all of yours is valuable. However, i could do without all the poorly substantiated "game is rigged" bitterness.

maxim said...

All of the above being said, i'd like to point out that the most interesting thing here is how the optimal play against "sure-win" teams goes against the established metagame. If my goal was to really help out the lowbie players, this established metagame is exactly the kind of target i'd go after.

Anonymous said...

In op.gg you can go to champions tab, and then normal. It seems to provide all historical stats. Ie. it should be quite usable in determining who is actually new with champ.

Anonymous said...

I'm curious as to what percent of your games had both a good and a new player. It's quite possible that account sharing is happening with a higher ranked friend or ELO booster. Smurfs that are good enough will play a variety of champions to keep away the boredom. In such a case, the smurf plus one good player could greatly increase the win rate for your team. Do you have some data on the KDA and CS of the new players in your easy win scenario?

Skeddar said...

"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. [...] Force teamfights!"

My main question is if you tried this 'exploit mode' in several games in a row no matter what your formula says. You once described it as 'a totally retarded way to play', and in case of a support WW as double jungle I might agree with you. But abusing a number advantage on a lane or in a teamfight is usually a good tactic, and pushing together to earn fast turret gold also seems reasonable.
So... I wonder if you managed to ship around Warwick's weakness, the low early game impact (as most Warwicks just farm till they got their ultimate) with behing quite active in the early game (invading, early ganks, early pushing), and your 'exploit mode' is just a better way of playing Silver elo games.

Gamjatang said...

I'm curious as to what percent of your games had both a good and a new player. It's quite possible that account sharing is happening with a higher ranked friend or ELO booster. Smurfs that are good enough will play a variety of champions to keep away the boredom. In such a case, the smurf plus one good player could greatly increase the win rate for your team. Do you have some data on the KDA and CS of the new players in your easy win scenario?

Of a similar vein, it may simply be that a good player and a new player have something in common. They are less likely to feed. I think a new player is less likely to go for the kill when laning. That puts the chance of a feeder to only the two remaining players. The probability of a feeder on the other team is thus 2.5x higher. Thus, if feeders determined the outcome of the game then you should win much more often. The remaining wins could be from smurfs in duos or account sharing.

All we need is some data on the KDA and CS on the players playing a new champion for the first time in ranked.

Gevlon said...

@Maxim: giving stronger teammates to players who don't perform well is called rigging. If Joe isn't performing on silver level for whatever reason, it's the NORMAL thing to let them fall to bronze. There is absolutely no more reason to save them from defeat than to save the guy who just plays badly because he's drunk.

But here is the twist: How does the matchmaker knows that the player will pick a new champion? You are first placed into a team and pick champion later. The matchmaker can't know if the next game I'll play WW or not! The ONLY info Riot has that predicts upcoming new champion usage is buying the champion or buying skin for them. So even if you're right and they just wanted to help people with more champions, their algorithm is "if recent_purchase then easy_win"

Finally: Even if they would want to "help a bit" to new champion users, don't you think that 77% winrate is a bit off?

@Anon: about 30% Easy Win, 30% Sure loss, 40% Fair game

@Skeddar: the normal setup is normal because it's optimal. Early game you can't dive tower, if you attack 2v1 at a tower and the enemy isn't an idiot, you'll likely won't kill him or just at the cost of a death. In the meantime your original lane enemy gets fed. I mean if mid goes second top, then enemy mid can easily lasthit all the minions while his minions die without giving gold or XP to your team. He can also break your undefended tower faster you brake theirs 2v1. This strategy is outright retarded and that's why not used.

Slawomir Chmielewski said...

I think it can't be avoided in this type of games (ie, games where microtransactions purchase new game assets).
It is obvious that any new asset will be losing games. Without practice people are worse, they have no experience with new assets. Thus, a high chance of loss.
If people who purchase new items lose their first games they will instantly recognise the pattern "new hero = loss". Even if they don't realise it consciously, they will stop buying new heroes to avoid losses. Very important is the timing - the shorter time span between events, the higher chance of association. Which is why any game immediately after purchasing should be a win.
Therefore any game with this sales style who does not introduce some form of rigging for purchasers will inevitably fall short in sales and lose the competition to games that do.
The fact that people openly despise p2w games (only because they are poors, most people like rigged games as long as they are rigged in their favour) means game makers can't just sell wins.
So, they can't sell wins, they can't sell new characters without rigging and they can't sell games themselves (cause nobody wants to pay subscriptions, the market has been broken by f2p). Very few sane people will spend money on skins only, so they can't rely on income from this alone.
They have to sell something after all. The only thing left is new items plus rigging the game to avoid immediate negative association with purchasing items.

It seems that we are facing a logical problem. It is NOT Riot or other game makers being dishonest. It is natural selection killing all games that aren't rigged. Our brains work like that and nobody can help it. Actually, if people could not play a newly bought character for 24hours after purchase, it would mostly remove the negative associations, I wonder if this alone would allow for a fair system.

New, constructive task for GreedyGoblin: design a game payment system that isn't rigged, isn't subscription (cause people hate it), isn't expensive (cause nobody pays £30 for games made by unknown publishers), does not charge for DLCs and expansions (as this is a form of subscription) and isn't blatantly p2w. Repeat: not a game design, payment design. Gl hf.

maxim said...

@Gevlon
A game where 5 okayish players go up against 3 okayish players and 2 fresh-champion players would have 90%+ winrate for the first team. Compared to that, 70% is really not all that bad. And i bet it only comes up at certain points in the ladder, where - for some reason or the other - individual rating points correspond to much larger improvements in skill than normally.

And if having games with more even winrates requires me as a game designer to do some matchmaking decisions based on recent purchases, i'd gladly do so.

The rest comes across to me as just your ideology on how everything should always be 100% fair all the time. Some things must be 100% fair, but Silver League LoL is not one of those things.

Alessandro said...

Play another MOBA, I recommend blizzard "Heroes of The Storm".

The question is... Does other MOBAs also rig the match-making system? Blizzard also sells champions and skins...

Steel H. said...

"It needed a Support Hecarim (post tomorrow) to wake up and realize how far I went from my original". What's the story about this one, you didn't mention it in the post.

Anonymous said...

Those defending RIOT here are so frustrating to witness. I have friends in high mmr (challenger) and they noticed another factor to predetermination of outcome and that is several people queuing at the same time (odd number for friends) and on the team that gets less friends, the friends troll. This is called win trading and high win rate high challengers are low diamond trash that exploit to climb. To climb to challenger the abuse duo with smurfs of entire league mmr gap and at challenger where you cannot duo, they win trade by queuing at same times and ensuring an average superior to 50%. This has been brought to attention in Europe for a while now and now even in NA. Riot ignores such claims and silences those players on reddit and league official forums. Good work gob!

Gevlon said...

@Maxim, Slawomir: normal games, custom games are a thing. New champion users could simply play some normals. Hell, they could be required to do so before going ranked. I played 4 normals with the new Warwick and I still make stupid mistakes and rightfully would be in Bronze.

@Slawomir: people don't hate subscriptions, devs hate it because they want idiots pay $10000 instead of $15. I don't question your logic that rigging is the result of people wanting to buy wins without the stigma of "wallet warriors". It's just LoL openly claimed to be a fair e-sport instead of yet another p2w crap.

@Maxim: A game is fair or rigged. I'm not claiming that League of Legends should be banned and its devs imprisoned for rigging Silver games. But I do say that if you are in Silver, Bronze, Low-Gold (80% of the players), the matchmaker is screwing with you and you shouldn't play. Also, if someone rigs Silver, what makes you think they don't rig platinum? For example by giving free wins if you get a second smurf account to practice.

@Alessandro: I generally believe that a game is either
- everyone pays equally (buy box or subscription)
- openly P2W (gold ammo)
- rigged
So without checking I would say HotS is rigged. It would take some time to prove it. Maybe the devs are smarter than me and I can't prove it, but I'm just as sure it's rigged as I'm sure that aircraft trails don't contain mind controlling chemicals.

@Steel H: sorry, I totally forgot Support Hecarim. The Warwick patch forced me to publish the results, I wanted to climb to gold originally. Don't worry, the half written draft is in the blog engine, he will come.

maxim said...

@Gevlon
I am not a fan of forcing people into various game modes. Everyone can play ranked, with the implication that before a certain level you are "playing for fun" and after a certain level it gets "serious".

The "who is to say" argument is easily reversed. Who is to say that the developer necessarily will rig the game above "serious" level? Last time I checked, presumption of innocence is still a thing. Also, there are actual business repercussions for screwing over eSport oriented players after overpromising.

The smurf is a separate issue. However, as a general rule, developers don't like smurfs, but are forced to implement them due to various cultural conventions regarding player account anonymyty and ability to restart the game "from scratch". As a result, everyone i know of that runs a smurfable ladder wants to disincentivise smurfs as much as possible and make them do as little damage to matchmaking as possible in the process of rising up. The only game business model in which incentivising smurfs makes sense is the subscription-based one, which has now become a sort of a privilege only select game studios are allowed to enjoy.

Gevlon said...

@maxim: they are indeed can play ranked. But why are they entitled to win? Besides because they paid. Which is the literal "pay to win".

Presumption of innocence after proven guilty isn't working. You are saying "yes they are rigging, but we should believe they are not rigging against me, just against everyone else". Good luck with that!

Anonymous said...

I say again, correlation does not imply causation. Good work on analyzing games. Got some questions. How much LP you gain and lose on your first placed ranked games? How much LP you get and lose now? I mean, if I would rig the matchmaking system, i would keep ELO system intact, because its very good way to measure players who play. Abusing LP gains and rank placement is way easier to develop and maintain.

Let me tell you my story, just to give some insight what might help. Ive done similar experiment with sona (very underplayed and underused champ like warwick) on twisted treeline. I got more of a extrem case. First, twisted treeline ranked opened recently, so most of the players dont got any elo rating. Those who got elo, where team ranked players from last season. And second, ive played not a single ranked game until twisted treeline solo ranked opened. Ive not bought anything with money, but i got all champs unlocked while playing aram and normal twisted treeline, so alot games played over years. After placement games, firstly i was placed silver 2. Within a few days, i climbed quickly to gold 1. Then they reseted the ranking system and i had to remake placement again. I forced to lose more games, if im not mistaken, lost 8/10 matches and got placed on bronze 2. Again, i quickly started to rise. Because every player got more or less same ELO rating, every victory gived 23-26LP(yes, 4 wins in row was enough to get promotion) and every defeat removed 14! LP points. More you played, faster you would rank up, there was no way to lose ranks. After i got to gold 1 again, and read your blog, i decided to play only sona, just to prove that you can climb with any champion, even with sona. Worked better then expected. Because i played alot, i was first on every new league that i entered. I was matched up against players who are not in high skill, but who just have played alot. Unconventional champion pick and lack on knowledge gived huge advantage and i carried or let be carried on lot of games. Literally, i was 1 win away from entering diamond while only playing sona. Im still plat though. Then real life happened and i was forced to make a 3 week pause on lol. Alot happened on that time. Skilled players slowly but surely climbed up. After coming back, sona wont work at all anymore. 100% of times counterpicked and overplayed. I losed 2 ranks in plat then i choose to stop play for a while sona as ranked. Reason was teammates. After all ranked play, there where afkers from start, feeders trollers, happens, just keep playing. Then suddenly, alot players afk after 1-2 deaths. Game was still winnable and very even, they just wait to 15 min in base to surrender. I admit, playing sona feels like i am trolling, but i play good as i can, often from behind. Thats enough reason that teammates take more risk on early and when they dont pay off, afk and blame me for trollpick. After 5 games in row and losing rank, i stopped sona experiment. I point out, when i was entering plat, all games was played out til the end, noone where afk. Before i stopped on plat, every win was 23-26 LP and defeat 14-16 LP. After 3 weeks, Wins/defeats where equal around 20 points. What it means, the main mass leveled up to plat and they know how to play against sona effectively.

maxim said...

@Gevlon
70% win rate is not "entitled to win".
On the flipside scenario, in which new-hero people go against old-hero people without help and lose near 100% of the time, what makes the "old-hero" people entitled to win?

Your position on "guilty until proven innocent" is akin to wearing a gas mask while smelling the roses just because you ran into a fart previously. Sure, it saves you from bad experience. It also makes you miss out on all the good ones.

It is a safe position to take, but i just can't see it as a worthwhile one.

Gevlon said...

@Maxim: the player is in full control of his actions. If he CHOOSES to pick a champion he sucks with, he deserves to lose. He is free to learn the new champion in normal games.

maxim said...

@Gevlon
Maybe, but why do other players deserve to win then? Are you building a stale game that locks everyone into their chosen champion?

The only thing the player "deserves" is interesting games. Games where you are 100% losing / winning are less interesting than games when you are 70% losing / winning, even if 70% is still far from perfect (and i don't know of a way to solve this particular matchmaking problem without creating even more issues elsewhere).

Oddly enough, the existence of your strategy of playing such matches could potentially help bring that ratio from 70% o 50%. However, you yourself are the biggest obstacle towards its widespread adoption, because you tie this strategy with the notion that the game is not worth playing to begin with.

ANE said...

You could try atlas reactor as its free to play and I don't see it as pay to win.
Aside from that archage that has been mentioned already