: The infrastructure required to run the extra code just to _calculate_ all of the extra info needed to forcibly give people winning or losing matches would be far too expensive to maintain as a business. For a single player it wouldn't weigh on the system too much to run the few extra lines of code, but multiply that by every user concurrently logged in and playing on your region's server at any given time. That's a lot of times to re-run the same lines of code. And besides, it would double queue times because the pools of players available to matchmaking would have to be split in two (people owed wins and people owed losses). 50% win rate is the outcome of the system, a byproduct. It's not a setpoint that Riot uses. With setpoints, you can tell the system what you want, in this case, to match players so that everyone has a 50% chance of winning, but it leaves out all sorts of important factors. So it fails to be accurate, and tends to over or under estimate players' skill levels. This leads to unintended imbalances in matchmaking, aka win and loss streaks. If you're playing at your actual skill level with people similar to you, you _will_ have a 50% win rate because your chances of winning against people of the same skill level as you are 50/50. It's the same as a control system, the system attempts to correct for the changes in your MMR by matching you with people it thinks are at your new skill level, but because your MMR is still changing (be it from a win streak, or a loss streak), it messes up and can over/under shoot, and then wind up putting you in a match you don't belong in. Now multiply that by 10 because there are 10 different players who's MMR may or may not accurately reflect their true skill level. To be honest, I'm surprised matchmaking works at all since things can get ten times as wonky. Riot should be using PID control to more accurately predict people's skill level instead of whatever they're doing with MMR. See here for an example of what your MMR would look like compared to your actual skill level. The colorful lines are examples of MMR as it changes over time, and the flat reference line is the setpoint (your new skill level / MMR) after a change has occurred. In reality your skill level is constantly changing and it won't look as flat, but the idea is to show how a system's outputs oscillate when responding to a change. https://upload.wikimedia.org/wikipedia/commons/thumb/a/a3/PID_varyingP.jpg/320px-PID_varyingP.jpg
What a perfect description; thank you. The over or under-estimation of skill is exactly what I feel during games. I will think of this when I play: The oscillation, as you describe it. It's just so frustrating that I can almost forecast my losses after my wins... That's just a system that is broken in my opinion. If you're right and there is no algorithm, just an MMR that makes up the whole team, then I think it would highly benefit if there was an algorithm that looks at time played, champion mastery, and game ratings (S, A, B, C, etc.), and determine what each players skill is. This math could be done passively on the account (like every 24 hours). Until then, I dispute my right to be furious at this 50/50 nonsensical system, forcing people to play like madmen and snowball or never climb the ranks.
: so question... if people like TF BLADES , and other high high elo players can get to master/challenger again and again regardless of their teammates and this so called "forced 50% win/loss algorithm " then why do you think it exist?
Players get placed in a similar elo that they had last year.
: 50/50 is something that will come to every single type of matchmaking system as long as you play enough games. You have to be either way better or worst than everyone for you to achieve anything but 50% winrate over a long period of time
I hear you, but in professional sports, that's not how teams are formed. They don't look at your wins and say "Hey, he's really good, we'll have to put him in a team that sucks to balance the game." We don't know exactly how the algorithm works, but why can't silver players just play with all silver players, bronze with all bronze, etc. ?
Rioter Comments
Rioter Comments
: A Gargoyle Galio skin would be cool, maybe for the Harrowing with a Victorian gothic style? https://img0.etsystatic.com/033/1/6611438/il_340x270.569992952_n1ol.jpg
Rioter Comments
: All 3 of the intentional AFKers in my ranked games last night are still playing
I propose an idea: If a player hasn't reached a certain rock-bottom standard, they are booted from the game. I would rather fight a 4v5.
: "Gothic Galio" would be a cool skin too and could visually function within the same space as Traditional Galio. Would certainly be a way of giving him a more traditional gargoyle-like appearance/fantasy.
That would be pretty cool. I'd like a stone gargoyle with red eyes!
Rioter Comments
: So a champion requires a brain to play is "unplayable"? Okay then.
I never said she was too complex... I'm saying she's weak. Too weak to be viable.
Leo1024 (NA)
: new champion
Friend of the forest, suggests that he can take their buffs, exp, and gold without killing them. Perhaps the camps will follow him once they're defeated and act like pets.
: > [{quoted}](name=Sharpsterman,realm=NA,application-id=3ErqAdtq,discussion-id=exwoz8YN,comment-id=,timestamp=2016-09-16T01:59:29.083+0000) > > {{champion:429}} was undeservingly nerfed earlier this year. No, she wasnt. Kalista is a mistake champion that should never have been released, let alone be even remotely viable. She needs a rework, not buffs Also, comparing Kalista to Kog'Maw, of all champions, lol? They are completely different.
It's my understanding that Kalista's niche is that she can engage with 1 or more enemies and at the exact moment they realize they can't win, Rend takes their lives for epic multi-kills. That's why I choose her over others. Also the ability to dodge and dance around the battlefield is an added bonus.
Rioter Comments


Level 143 (NA)
Lifetime Upvotes
Create a Discussion