Elo Hell: Testing Moba 5v5 Rating Systems. Do they work?

[@vvvortic]

i deserve challenjour but ranking system holding me back per say, gg

[@CarnivorePlays]

Stop getting lazy....

[@ExpiredOatmeal]

I've been saying this for ages, and I always get told that I'm just crying about my elo. I play this game for fun, not to climb, so I'm not bitching about my elo, that being said, the reason I don't play ranked to actively climb latter is because the system is so flawed. When you're judging one person solely off of whether his/her team wins or loses and how many games they've played, you're not going to be accurate unless that person has played an absurd amount of games. How do you have the most played game on Earth that's heavily competitive and lack any form of complex ranking system? You either have to be one of the best/worst players in the world, or play 500-1000+ games a season to be accurately ranked, and that's unacceptable. What in the hell made them think it was okay to leave individual contribution out of an individual's rank??? It took someone explaining how the entire equation is flawed before most players could understand that the system is in fact inaccurate, when it should be common sense.

[@david21686]

Have you considered using a model that estimates a team's strength as a sum of the individual player's strengths, and creating a regression model to estimate the strength of each individual based on what teams won? For instance, let's say that the strength of team 1 (T1) is equal to the strength of players 1 through 5 (P1 + P2 + P3 + P4 + P5). The strength of team 2 (T2) is equal to the strength of players 6-10 (P6 + P7 + P8 + P9 + P10). Skipping forward a bit, let's say that the strength of team 100 (T100) is equal to the strength of players 1,3,5,7, and 9 (P1 + P3 + P5 + P7 + P9). Let's say that there were 50 games with 100 different teams, and only 10 players. Then you have 50 equations and 10 variables; you can use some sort of linear regression analysis to come to a good estimate about the relative strengths of each player. Bump that up to 600 players, 1000 games, and several hundred teams, and I'm certain you can get a better estimate of relative strengths with fewer games.

[@NoOne3234]

david21686 I have thought of doing something like that (It's very common among college football analysts to keep a list of teams played and then run the games over and over again so that the ratings converge on a result), but I decided to keep it to something I was sure would scale well to a league of several million players that is constantly playing. I'll probably look into it again.

[@BrograckgroMalog]

Rabbit on Da Moon Have you looked into using a logistic regression. Where Y either 1 or 0, 1 being a win zero being a loss? And running a ranking system like that. I know the we could use a LRMC (logistic regression Markov chain) this could give the team or each player a better rating based off of KDA and length of game to determine which persons are better. But then again I don't have the data to see if this theory would hold, maybe my model would not have the right functional form.

[@KowalSonn]

BrograckgroMalog I have no idea about those terms you are using but taking into account KDA would make people from losing teams just go afk to don't die and keep their pretty nice KDA after winning their lane. Take into consideration that supports and toplaners can affect game really hard while having poor KDA (in close games supports and tanks die a lot) so that would make higher tiers like diamond+ just flood with adc and mid mains which in most of games get best kda (like season 2 diamond full of adcs).

[@VopleForce]

This was really interesting, have you considered how much a "players skill value" changes when they are playing their main role/champion compared to a less familiar one. The amount of games to reach that sweet spot i would expect to be far higher given how much this value can change. It will be interesting to see what riots team builder ranked queue rating are like when people can play at a more consistent personal skill level.

[@SherlockHolmes000]

Meta, character balance, and mechanics themselves also play a huge role in this. In Dota 2, at certain MMRs (or Elo), certain characters are abused repeatedly because they are substantially overpowered in that particular MMR rating, such as Slark in 2-3k MMR. These types of characters are force multipliers for the player, and require more teamwork of the opposing team to defeat, than it does of skill for the player to use. As a player in a public match, with randomly assigned teammates, you may be the best player in the game, but your opponents may use strategies/characters that you can't defeat alone, and at that point, you are forced to rely on team cooperation in an MMR where team cooperation doesn't happen. So what overall happens, is that players who learn to abuse the meta of the ranking, end up floating to the top, whereas players that do not, stay in that rating.

[@HOLYPEPSI]

Yeah many people don't really understand the learning curve of league. I'm a Platinum player who has 2 smurf accounts in gold. Every time I did placements I was instantly placed in silver 1 or 2 on brand new accounts. If you're at a higher skill level, you will consistently carry your team, that is a fact. By no means does that mean every game is winnable but you will on average win more games. Just saying, I could probably go on a 20+ win streak going Vayne or Ezreal mid in bronze elo while watching anime.(While leveling up my smurfs I kinda watched Attack On Titan and won mid almost every time level 20 - 30) Another great example is, if you put a bronze or silver level player in Gold, they will get absolutely trashed because they aren't at that skill level. But according to many people 5v5 ranking systems don't work, so it should be almost impossible for me to have 2 accounts in the top 20% of the community and 1 account in the top 10% of the community. PS: only took me 25 games on bother smurfs to get them to Gold from unranked never played ranked before, thus proving the elo system works. -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- 2nd PS: 90% of the reason why you are stuck in Bronze & Silver is because of lack of knowledge of champions and correct build paths / build orders. Pretty critical for getting into the higher leagues.

[@DemolitionTurtle]

I loved this video! I've always been interested in how ranking systems work and stuff like that from the variety of competitive games I've played, but I never considered testing a custom ranking system with simulated players and games. I think I'm going to have to have a go at this, as it sounds really interesting! I hope you decide to make more videos on this topic or something similar, as I really enjoyed this one! :D

[@blackchoas]

Thank you very much for this video, I had never put much thought into how the ELO system actually works, this piece was simple but informative very interesting,

[@BahamutPrime]

While I don't know how you could model this it makes me think of two issues in the lol when compared to your model 1. People have not played the same amount of games 2. Players skill vary (game by game, by role and champion and also ideally they get better over time as well) Very Interesting to see how much varying the amount of LP you gain per game to account for player strength makes a difference in the accuracy of the model though.

[@tashurastogi]

@rabbit on da moon why did you stop making videos? will you ever start again?

[@robbie9230]

+zeldakartds I'm talking real world, meaning that specific players have different learning curves

[@mcgsanchez]

vvvortic should explain us how this really works, gg

[@Styhn]

Could you elaborate on your monte carlo approach? Cool video!

[@br3ak_]

I am plat 3 on my main account, but I have over 80 games on my smurf and I still can't get out of silver 2-silver 1.

[@Landibert]

I like this but there is one important thing missing I think: The (new) experience per game. It's really hard to do I would guess but each player would also need a random number (let's just say from 1 to 10, but don't pinpoint me on these values, they are quite random!) of experience per game. Random because people learn in very different speeds. Means with every game (win and lose!) a player gains his specifc number of XP which passively increases his MMR to reflect that he is getting better (on average). Is it possible to include this? Does it even make sense to include it? P.S.: I'm actually horrible at math but I really like statistics and numbers. ;) If this method seems bad, I'm sorry.

[@kotiebean]

what about people that are duoed? Can you fit into where your strength wants you to be fast with a partner?

[@Tjlo1339]

How do you even dictate how much each individual put into winning a game? That in itself is almost impossible to calculate. For example a top laner that puts so much pressure on the map that forces the jungler top and freeing up the other for members of the team to win 4v3.

[@StickInMudd]

Thank you for this thoughtful analysis. I know this one player who has played over a thousand ranked games (for the past few years combined), and is consistently in Bronze 4 and 5. He is so awful at this game, I really don't know how someone can be so bad. So people who are in those low elos deserve to be there. As for the other elos, however, this game has so many more variables than the ones just based on player skill. There is team composition of champions, toxicity of players at a given time, is the player having an 'off' game, random disconnects from the internet, and of course lag. Although the above list of variables are important, the most important, imo, is the one about champion pool for a player. Each player has a limited champion pool, and if they are try harding, they will use their best, which is probably around 5 or 6 champions. These champions will most likely be similar in play style, such as hard engagers, or burst mages, etc, to fit that player's preferred stlye. Sometimes, just randomly, your preferred play style can be countered pretty hard, just due to meta changes, forcing you to use champions you are not familiar with, thus lowering your chances of winning. TL;DR: Basically, you should add the variables of champion strength and champion counterpick, to the player strength variable. Instead of one number the whole data set, for all 1000 games, that number should vary between the different champion strengths, and sometimes that player should be countered altogether. IMO, this would provide a more interesting, and closer to reality, graph of numbers to look at.

[@Kramillion]

But the important things you can't calculate is how the individual feels/acts. Whether he is on tilt or is inspired. Mathematical equations can only take you so far.

[@FinalBossWTMN]

In a 5v5 game, where the 4 other players are random and not constant, a ranking based on win/loss only, CANNOT work. If one player is powerful enough to win the game (which, would ruin a 5v5 period) then it would be accurate, but that is not the case, so it's not. If it was accurate, when a Challenger level player creates a smurf, he should literally not ever lose a single game, at least up until high plat/diamond. But they do lose. A lot. Many of them, can't even get their smurfs up to their mains rank, or even get out of gold without a second high ranked smurf queing with them. In the closed beta, Riot constantly said, the current ranking system, didn't work for solo que, and it was a temporary placeholder until they could make a better one. But the game exploded, became popular, made huge amounts of $$ anyways, so they now have no reason to bother. Why spend money fixing it, if they won't get more money out of it? I used to keep data to prove how this system is ineffective, at one point, had Riot backed into a corner in a huge forum post, with absolute loads of data from lost games where the players did nearly everything they could, but had leavers/feeders/awful team comps/etcetc. Finally the Riot member said "Elo works. If you're good enough, one person CAN carry a team, and you should be able to carry yourself to where you belong." So, I then took a post from a different Riot person, whom was responding to a player saying it's not fair for two people to que for "solo que" and face solo players. Riots response to that was "it's a 5v5 TEAM game, 1 player doesn't make a big enough difference to sway the game that heavily in most case." BOTH of those statements CANNOT be true at the same time, it has to be one or the other, that's it. Bottom line, Elo is a system made that can ONLY WORK ACCURATELY, when the 2 teams are consistent. When gameplay is NOT taken into effect whatsoever, and an 0/20/0 and a leaver, gain/lose the same amount of elo, as the person who goes 20/0 every game and carries/tries to carry, it's not going to work well. Basically, it CAN work eventually, if you don't have awful luck. But it takes A LOT LONGER. Also, generally, when you get closer to where you "should be" it's going to get much harder, and advance slower. If you are say, plat skill level, it will be fairly easy for you to dominate more bronze/silver games. But once you reach gold, it's going to slow down, as the gap between your skill and the other 9 players lessens, and at that point you have to pretty much ride it out in a wave of a huge amount of played games, and hope you end up where you should be.

[@krokso519]

But you did not take into consideration the promos bur you still did a reaaaaly good job

[@thies3817]

Im not 100% sure but would´nt it be impossible as a support main to get some elo in the second system because u always have a negative score?

[@Frikgeek]

+Thies Fronius You misunderstand how the 2nd system works. It still doesn't record anything aside from game outcome but it adjusts gains/losses for each player varying by their Elo. So if higher Elo player duos(or in his case gets randomly matched) with a lower Elo player the higher Elo player would gain more Elo for a win and lose more for a loss while his teammates would both gain/lose less.

[@robbie9230]

Also, in the real world after 10,000 games a player would be expected to improve alot over the given period of time right? So wouldn't this make the curve even more scattered?

[@TheIsaia]

Kishi Rno I dont belive this would have anything to do with this since in the simulation every player played an equal amount of gains, (and improved at the same rate)

[@gamekiller0123]

ZeldaKartDs Do all players improve at the same rate though? There is also the case of skill dropping in periods where you don't play and new players join all the time.

[@LogicIsLogical]

ZeldaKartDs But players dont improve at the same rate.Theres people with 2k games (in total) that are diamond and then there are people woth 2k games that are still gold.

[@LogicIsLogical]

LogicIsLogical I'm not saying elo hell exists.But there should be(and I think riot does have it in place from what I've read) greater emphasis on new data than historical data.

[@valoreclipse-samael4706]

So this mean lol ranking system actualy not accurately measure ur skill?

[@lXBlackWolfXl]

This is a nice video with plenty of mathematics to explain the results. But there's one issue: you're assuming everyone uses the ELO rating the same. They do not. This is evident in smite. People noticed playing ranked (where you can actually see each player's ratings) that the system regularly put people at the lowest ranks against opponents with the highest rank. Worse yet, the rank of the players within each time varied widely. You could see games where one gold player would be thrown into a match with 4 bronze allies, and all gold and silver opponents. Its clear that the system wasn't matching players up based on skill. And you know what? Hi-rez confessed to this. Their system does NOT pair you up with allies and opponents of the same rank as you. In reality, the system is designed to keep your winrate at 50%. If you go above that, the system will pair you up with increasingly bad allies and give you opponents that are increasingly higher rank than you. However, you to start to fall below 50%, then the system gives you allies that are higher rank than you, and opponents that are below you. They claim to did this to boost player confidence, because it ensures that everyone has a winrate of 50% regardless of their actual skill. And they reported that the system was working EXACTLY as they intended; everyone's winrate hovered around 50%, with even the best players not managing to get above 55%. This kind of system creates a whole host of problems that you could write an essay about. 1. Games are ALWAYS one-sided. Playing smite for years, I only ever got TWO games that were actually a fair fight (and I remember them because they were so fair, which was unusual for the game). Every game is made to be one-sided to keep everyone's win rates at 50% 2. Keeping everyone's winrate at 50% makes everyone believe they know exactly what they're doing, regardless of whether they do or not. This results in a culture of arrogance, where everyone believes they're a pro, and because the system often pairs you with allies that are below you in ranking, it IS 100% true that if you lose a match, chances are it wasn't directly your fault, but who were you paired with. 3. Since everyone thinks they know what they're doing, the overall skill level is low. This can really be seen in the different versions of Smite. PC players were clearly better than ps4 players (on average, anyway). However, even on ps4 everyone was convinced that they knew what they were doing because their winrate was 50%, and even if they moved over to pc and got owned hard game after game, their winrate would still be 50%. 4. The 'rewards' are completely backwards. The game literally punishes you if you win, and rewards you if you lose. I actually started intentionally feeding for a while in smite. And you know what? My winrate didn't go down one bit! In fact, I had far more pleasant games when I deliberately tried to sabotage the game than I did when I actually made an effort. I didn't know this at the time, but what was happening was that the system was putting me in teams that could win despite my behavior because I was doing so poorly. So yeah, I got rewarded for intentionally feeding, and punished for actually trying. Wtf? 5. Rigged games. Who enjoys that? Your own skill in such a system literally means NOTHING, because every game you play is rigged to either make you win or lose, depending on what would keep your winrate at 50%. All these reasons (and the fact that the game is COMPLETELY un-moderated, and admittedly so, oh and the game seemed to have NO concept of balance, because the developers claim that even they couldn't play their own game, seriously) is why I left. More recently though, I've heard that dota 2 literally does the exact same thing! If this is true, then how common is it? I always thought it was unique to Smite and its obviously inept development team, but if dota 2 does it also, who knows who else? I don't know if LoL does this (haven't played it in years), but I don't recall the matchmaking system feeling like I was always playing one-sided games. Then again, I did seem to alternate back and forth between winning and losing streaks for some reason; I would literally win 4-5 games in a row, then lose 4-5 games in a row back and forth. How common is this? And we clearly shouldn't be trusting corporations, that are just trying to milk people out of as much money as possible, to make games that are enjoyable and fair. We don't know what's going on behind the scenes, we don't know how the matchmaking or how some aspects of the game actually work. As for dota, there is evidence to back this up. The top players of dota 1 can never get a winrate of over 55% in dota 2, no matter what. It does seem like the system is pairing them up with bad allies to try and keep their winrate down. Even miracle and other top dota 2 players have released videos where they're carrying teams that are clearly far worse than they are. So it IS perfectly possible that this IS the case in dota 2. You don't see this one-sided nonsense in dota 1, though in dota 1, new players are virtually non-existent. As a consequence, the entire player base is made up of people who have been playing a long long time, thus there isn't much variation in skill level. Long story short: just because a game has a ranking system, doesn't mean that its fair. They may not be matching you up in the way that you think they are.

[@TrueHimbo]

For the elo equation when you add in your opponents elo do you take the opponents new elo since winning or losing that game or do you take their elo before that game is played?

[@Frikgeek]

+Bryce Johnson How the fuck would you use the opponents new Elo if that equation is needed to calculate his new Elo?

[@ElephantDickGaming]

How did you make your graphs?

[@JEFF4507]

The varying volume takes away from what is otherwise an interesting video.

[@ImBoredAlot64]

I think Riot uses a system closer to the "trueskill" system microsoft crafted than the ELO system of chess (although, the trueskill system is similar to elo). Check it out here- research.microsoft.com/en-us/projects/trueskill/details.aspx . I'm really curious your thoughts on this system, and if it shortens the # of games needed for a "pretty" curve.

[@NoOne3234]

ImBoredAlot64 "Trueskill"=Team Glicko with a gaussian rather than logistic distribution. Valve does something similar so Riot probably does, too. Glicko itself is modified elo that incorporates uncertainty into the predicted outcome term and does a better job of incorporating it in the K term as well. Many of the gains are with introducing new players and reintroducing old players who have been inactive for a while into an existing league. It should not dramatically improve convergence in my simulation, but Glicko should perform better than Elo in real life. What are telltale signs of Glicko? Glicko matchmakers will try to match you with people of similar rating and uncertainty. If you haven't played a LOL game mode for a while, you'll be placed with players with big error bars near their ratings. You'll get a large range in player skill during your first few games, which will decrease as you reduce your uncertainty term.

[@whyando]

what value of k do you use?

[@Jastro88]

Having to play thousands of games in order to get your correct ELO rating is a bit disheartening, but it makes sense. I wonder how many games the average Challenger level player has played?

[@bastiendetaille]

Jastro88 Most of them get Challenger after 150/200 games. When playing in Silver/Gold/Plat they got like 95%winrate

[@NoOne3234]

Jastro88 The number of necessary games depends on how accurate you want it to be. When playing in an established league, it's on the order of hundreds (~500 is a good number for convergence). Furthermore, (and I learned this from Phreak's Reddit post) Riot seems to start seeding you for ranked when you fill out the survey when you create your account. This implies that they also use some information from normal MMR and under lvl 30 performance when seeding you for ranked. Depending on how they do it, they could shave off hundreds of ranked games using this info.

[@hapeenes2927]

Jastro88 im pretty sure that pro players with a lot of experience like Dyrus, Froggen or Mata/Madlife who are playing since season 1 have played more than 10k games already. I would like to say more. This system is only viable if someone has never improved or if everyone is improving at the same rate which seems like bullshit to me. Play and enjoy the game and eventually you will reach plat/diamond as they are not that hard to reach actually.

[@Jastro88]

Rabbit on Da Moon It makes sense. Even when I play a BOT game I am often matched with players who are in the same league/tier as me. (I'm S2 currently.) Normals seems to be the same way. As you stated, it saves time.

[@Jastro88]

Aced Positive Seems like a good approach. I do enjoy the game, and if I stop having fun while playing I take a break. It certainly does seem to largely be about grinding out games. (With the caveat that you should be actively playing your best/trying to improve, of course; auto-piloting through a grind session doesn't seem productive.) It's the same way with any discipline, really. If you want to be really good at something you have to do it a lot.

[@GremiasV]

LOOK OUT FOR PANTHEONS BIG ASS FOOT.

[@jackwood7691]

What song, Is it disney? can someone tell me

[@NoOne3234]

Jack Wood Tchaikovsky -Waltz of the Flowers.

[@jackwood7691]

Rabbit on Da Moon Thank you !!

[@mathiens8666]

+Jack Wood Darude Sandstorm

[@smipth]

There is only one data point can accurately measure the skill of a league player over a set of matchmaking games. Whether or not he picked teemo. Just kidding! ... it is whether his team took out the enemy nexus or if the opposing team took out theirs. At least when league was first released this was the case and indeed riot would defend different ways of playing the game back then. In recent years, as the ways to play league have diminished, then well yea fuck it I'm sure individual contributions would be a great way to determine elo.

[@coda72]

This is garbage there is a chance where u can go go bronze 5 with challenger skill a 0,001 but is my case >-

[@bassisku]

It's more like 0,00000000000000000000000000000000000000000000001

[@cmck362]

Buble There are multiple examples of high ranked streamers doing bronze to diamond runs without getting stuck in bronze. There is also examples of riot giving pro players bronze mmr accounts to help test their ranking system and booster detection. Theres also way back in the day hotshotgg (or was it saint idk pretty sure its one of the two) getting his account hacked to 0 elo and climbing back out. So no its impossible to get stuck in bronze if you're gold or better. Silver might get stuck for a while, but thats the same thing as diamond smurfs having some trouble getting out of plat. Its impossible to get stuck anywhere significantly below where you belong. I'm low gold and I took my forgotten bronze 3 smurf out with about a 30 game win streak playing nothing but graves at the end of season 3.

[@loominato5035]

Doesn't the fact that you are "simulating" the effects of "player strength" impact the final result? Wouldn't the results be different if skill of the players impacted the outcome of the game differently? Sorry if it's a stupid question.

[@ecyor0]

loominato That's a terminology thing - he's using "player strength" as a generic measure of skill. It exists so we can quantifiably say "these players here are objectively better than those players there" and then see if the elo system consistently puts the better players on top. In real life, that value isn't directly measurable, and the elo system is trying to sort the players based on that invisible, abstract value that's the sum total of all their skill and teamwork.

[@loominato5035]

Wandering Bishop I get the terminology, what I'm (poorly) saying is: while player strength exists, doesn't the fact that you can't implement how player strength **actually** affects a game in reality affect the outcome of the situation? The "game outcome function" with arguments "player strength" is not the same in reality and in his simulation. I think this makes his simulation not accurate at all.

[@jelk1]

loominato You are right. Just like mentioned in the video the higher skilled players would be likely to have a higher of a impact in a game than the lower skilled. Imagine 4 bronze V + 1 diamond I versus 5 silver V. Let's assume bronze V is about 600 ELO, silver V is 1200 ELO and Diamond 1 is 2200 ELO. Then the team ELOs would be 4600 / 5 vs 6000 / 5 and if I understood right how it was calculated in the model the five silver V's team would be likely to win. Yet in real life that Diamond would probably just carry the game single handedly by first getting fed and then roaming all over the map. So in real life better people would really be winning way more than in this model yet this model already is already enough to show that Elo hell isn't real. And to all those suggesting improvement in players' strengths it wouldn't probably change anything regarding this video but might be interesting to see how people would rise in ladder if they would improve their strength faster than anyone else.

[@breakfreakmofo]

Sounds like HotshotGG

[@ShappySDJ]

CSGO does there ranking system much better then League does IMO but definitely both could be better

[@nickaclark]

The problem with comparing LoL vs CS:GO is that CS:GO calculates your KDA and league doesn't. If you are good in CS:GO no matter where you play you will get a lot of kills. On the other hand if you are a good support in league you won't have any kills.

[@hapeenes2927]

Nick Clark well considering how much statistics current match history has i think there is a possibility of making matchmaking better as well as ranking system. But to be honest i think it's not that bad. I mean we don't see pro players stuck in gold elo even thou we can see their capabilities of playing this game. Challenger players can reach masters/diamond within like 100 games every time on any smurf which i think pretty much says it all.

[@Frikgeek]

+Nick Clark CS:GO doesn't actually adjust your rating based on KDA, that's a myth. What it actually records are rounds won/lost and MVP stars. So if you're getting a 4K every round but never winning any you still won't rank up. It's also why you can sometimes rank up after a tie or if you lose 14-16 but were the MVP in a lot of those rounds.

[@jojopojo333]

If ur stuck in low elo, ur bad. Simple as that

[@barondryer4972]

well done.

[@iwillnoteatzebugs]

NO THEY DON T !!! ELO IS MADE FOR 1V1 LIKE CHESS NOT FREAKING 5V5 GAMES

[@drakestevenson6710]

Yes they do

[@Dewillo344]

I disagree. 70 games or so is enough to put you in the division you belong give or take 1.

[@NoOne3234]

Dewillough If a player is entering an established league with veteran players, 70 games might very well do the trick. However, things would get very bad if everyone only had 70 games.

[@tomascoelhopinto]

you can't get shit with only 70 games I'm sorry. You may very well go on a lucky winstreak for 30 games (let's say you win 25 out of the 30) and that's over 40% of the total games played. After those initial 70 games you may very well rise in elo or drop. I was Gold with 70 games. I kept playing now I'm close to Diamond, with a Diamond IV MMR. So you do need to keep playing.

[@NoOne3234]

I retract my previous statement after doing the calculations to see what my model actually predicts. A new player enters a simulated league where the average player has 1000 games. With only 200 games, the mean value for a good player will be shifted below equivalent players by ~200 elo. It takes around 600 games to converge on the actual strength. Assuming league of legends rating has the same performance, the model agrees with Tomas Coelho Pinto.

[@GamingVidsPs3]

Theres always one guy who comes on to a video full of logic and data analysis and says ''no i think its that.'' nothing to base it on, no logic, nothing. Just a stab in the dark.

[@agentcyde5770]

You talk to soft. I watched 30 seconds and stopped because I couldnt hear,

[@xThizu]

elo hell exists

[@unsubscriben0w]

Found the bronze

[@xThizu]

midder i could explain but u wouldnt care

[@11Brawlman]

xThizu i care explain

[@aozora7]

Julian Simon It's hard to win more than you lose when in most games there are feeders/AFKers that make outcomes of matches fairly random for their teammates. Once you get out of feeder/AFKer hell, you stat to climb much faster, since your skill has higher impact on games, even though average difference in skill is less.

[@xThizu]

Julian Simon Basicly elo hell is not being stuck in low elo. But the grind itself to get out. That is hell. You will lose quite some games trying to for example get Silver if you are bronze 5. Even though you are maybe currently already above the skill level of bronze. And thus those games are a pure hell to play out. Unless you are like diamond 1 and rofl stomp everything there will be some form of grind which definitely can be a hell.

[@Zorgloth]

wow you're not teaching anything, everything is obvious