Obviously massive rank disparities, and the result was exactly as those massive rank disparities would have predicted. Are we happy with this result? Does anyone consider this a good job?
Eybdoog11 votesNope. Game is broken
Kai Guy9 votesLets take a look at the match averages for the players in recent games to see if this was a gross mismatch.
Silver 4-3.
Silver 3.
Silver 4 - 2
(you.) Silver 3 - 1
S4-3
Vs.
S3-g4. Duoed W Diana.
Bronze 2- Silver 3. Duoed with Morde.
S3-s2
S3-s1
S4-s3.
All in all pretty similar, outliers being the duo on the enemy with the lowest and highest MMR in the game. Annie having low Data is a high uncertainty account but Dianna looks to be lower MMR then her.So lets try another site. One with Estimates. https://na.whatismymmr.com/
823±2, 826 ±22, 927 ±26, 760 ±35,797 ±37. VS 807±20, 811 ±10, 864±30(you), 823±21, 823±2 (your "bronze" teammate)
Averaged out? 826.6 Vs 825.6 Looks more or less about right ball park, Some Noise from further games played impacting accounts MMR and the fact that loss would have changed results. Nothing looks outrageous thou.
Look, MM has to balance out Duos. So seeing a game that's around the Mean value between Bronze and gold seems about right.
I have been telling boards all season that Duoing implementation really needs to be updated to function off Assigned hidden MMR not Tiers because the Range gaps create Mismatches.
Hopefully you will lean that Titles are a bad metric to use to gauge skill.
Nea10412 votes[{quoted}](name=Kai Guy,realm=NA,application-id=3ErqAdtq,discussion-id=KEFFbONc,comment-id=0001,timestamp=2019-10-03T21:13:33.056+0000)
You are just trying to explain how a flawed system works. Even if you're correct, the system still is not fine.
It has been a disgusting stomp, and everyone could have told that just by looking at the initial team rosters. Since Riot is not fixing anything, and since this kind of one-sided matches are so frequent, it's all intentional.
Kai Guy4 votesYou blend high uncertainty with low to speed up the time to place. This will generate mismatches. You use Randomized distribution because it's fair, not because it can magically make close games.
Mmr is the best mm model for a skill ladder. Riot has features tacked on which are for popularity and retention. Unlike folks who just rage about teammates I took the time to study Elo's work. So my issues e riots extra bells and whistles can be conveyed via logic not just feelings.
Op likes to play jinx mid crit build and still did poorly vs the weakest player in the match. This post is them looking for a way to excuse a sub 20% wr. It's easyer to blame mm the be accountable for personal impact.
I have no patience for folks who expect perfect mm. Humans are inconsistent, misplaced accounts are an expectation, data is needed to judge where on the curve some on should be, and there's not a fucking infinante # of players waiting in the que of identical skill To build close range even games 24/7.
A realistic goal for mm is to have the worst player in a match have their good games be able to beat the best player on am off game. Mmr is built aroumd this and despite the expectation it will generate some trashed matches.
Study up on the topic. I'm not the one with unrealistic expectations.
Sit In Fountain5 votesI solo killed Diana three times in lane and left lane somewhere between 50-70 cs ahead. How is that doing poorly? Also, where is the sub 20% wr?
Kai Guy1 votesLast 20 games. Mid role win rate.
Edit. The call out on my end was not very fair. Sorry, your on the wrong end of my frustration with the folks who participate in spreading misinformation. It was more personal then needed and if you want I will eddit It out.
haaaaaaalp3 votesYou are just trying to explain how a flawed system works. Even if you're correct, the system still is not fine.
- That game ended 11 - 29 , with a 10k+ gold difference (25 minutes, btw).
It has been a disgusting stomp, and everyone could have told that just by looking at the initial team rosters. Since Riot is not fixing anything, and since this kind of one-sided matches are so frequent, it's all intentional.
That isn’t the matchmaking system. Every game in this meta is a huge stomp.
Nea1042 votesThat isn’t the matchmaking system. Every game in this meta is a huge stomp.
It is, the moment you could have told that just by looking at the 2 team rosters, from the start. It's not (just) about snowballing.
Then, of course a better team will snowball, but it's a consequence of an uneven, unfair, flawed and probably rigged matchmaking.
Starcraft243ver1 votesIt is, the moment you could have told that just by looking at the 2 team rosters, from the start. It's not (just) about snowballing.
Then, of course a better team will snowball, but it's a consequence of an uneven, unfair, flawed and probably rigged matchmaking.
Hum. No. You are looking at one game that validates your theory and you say "look ! i'm right", that's called a confirmation bias.
I can design a game in which every single game will be a stomp despite a perfect matchmaking if you want, it's very easy.
Take lol, make the aram map 1v1, allow buying, each kill gives 15 000 gold. Now even if you play against your doppelganger this will be a stomp.
You see ? Perfect matchmaking doesnt mean your game can't be a stomp. This is the situation we have now in league, you overextend top a bit, it's a small mistake, you get killed, your jungler gets killed, lose scuttle, 2 plates, 2 waves you are 1500 gold and 2 level behind. maybe your opponent did the same exact mistake 2 minut earlier but as your jungler was not there to punish it didnt matter.
Dope Solo2 votesOther people have made similar threads with multiple games looking identical.
Starcraft243ver1 votes[{quoted}](name=Dope Solo,realm=NA,application-id=3ErqAdtq,discussion-id=KEFFbONc,comment-id=000100000001000000000000,timestamp=2019-10-04T19:25:58.014+0000)
Other people have made similar threads with multiple games looking identical.
it's easy when you pick the games that validates your bias. A game has only 2 outcomes, so being right 50% of the time is no different than randomly picking without even looking at the ranks.
Dope Solo1 votesIts even easier when 80% of all your games look the type.
Nea1042 votesHum. No. You are looking at one game that validates your theory and you say "look ! i'm right", that's called a confirmation bias.
Hahaha, nice one. Just have a look at anyone's match history. Stomps are the majority of the games, and this is not statistically explainable.
I can design a game in which every single game will be a stomp despite a perfect matchmaking if you want, it's very easy.
Take lol, make the aram map 1v1, allow buying, each kill gives 15 000 gold. Now even if you play against your doppelganger this will be a stomp. [...]
Sadly, you missed the point. The problem is that you can tell who is gonna win since the beginning.
Try again, I've been reading this kind of poor excuses for so long ;)
Starcraft243ver1 votesNo you can't. You can do it 50% of the time, but since a game has only 2 outcomes it's no different than a random guess.
Now if you tell me that, looking at the champions picked, the bans and the stats of the champions for each player in the game, you can tell who has the best odds to win at the BEGINING OF THE GAME (and not at the begining of the lobby) I totaly agree with you.
But you can't blame matchmaking for this, because bans and champions picks happens AFTER matchmaking has already matched players together.
Also as a matter of fact, I never ever dodge the games and even when we sometimes have very weird compositions (I even got a no smite jungler in diamond because he tough he was top), we managed to pull a win in half the case.
Now the problem is if at the start of your game you act like your game is already lost, you bias the outcome as well.
The Kombinator1 votes{quoted} Also as a matter of fact, I never ever dodge the games and even when we sometimes have very weird compositions (I even got a no smite jungler in diamond because he tough he was top), we managed to pull a win in half the case.
One of the reason the weird compisition can win is, because nobody prepared for it. At least the great majority of players on lower ranks have no idea what to do.
Starcraft243ver1 votes[{quoted}](name=The Kombinator,realm=EUNE,application-id=3ErqAdtq,discussion-id=KEFFbONc,comment-id=00010000000100000000000100000000,timestamp=2019-10-07T10:15:56.034+0000)
One of the reason the weird compisition can win is, because nobody prepared for it. At least the great majority of players on lower ranks have no idea what to do.
Diamond elo, so I will not pretend we fully understand the game, but it's not a complete zoo either (well, it can be sometimes).
My point was most of the time people are likely to be like "omg we got an yvern adc we've lost", when it's not necessary true.
Kai Guy1 votesYOU CANT USE MMR TO RIG MMR.
It violates the math. These systems require a naturally occurring average to function. Its on a curve. You cant use probability math with faked probabilities because they stop being what's naturally the most likely event and become a unnaturally forced result.
Put 100 random people in a room. You can sort that room by the average Hight. You can mark everybody's Hight on a wall. Then look at those marks and correct judge what the average height is, then you can point out people who are above and below the average height. Force everybody to wear shoes and Stilts to be equal height. You can no longer sort them accurately by the average height because that data is corrupted by rigging them to be incorrect heights. Because everybody being forced to be the same average removes the ability to compare it as above or below average. Rigged MMR destroys the metric used to grade skill that would be required to accurately rig games.
Nea1042 votes[{quoted}](name=Kai Guy,realm=NA,application-id=3ErqAdtq,discussion-id=KEFFbONc,comment-id=00010000000100000001,timestamp=2019-10-04T20:54:00.233+0000)
YOU CANT USE MMR TO RIG MMR.
You can use other ways, especially if you hide everyone's (actual) MMR because of... reasons.
Force everybody to wear shoes and Stilts to be equal height. You can no longer sort them accurately by the average height because that data is corrupted by rigging them to be incorrect heights.
You can just take into account what shoes you gave to everyone, and use that new (hidden) parameter as you want. Nobody else will know, but eventually someone (wise enough) will notice that somehow you still know who is actually taller.
Again, you have no explanation why games/team comps are so often one-sided.
It's not a statistical error of any kind. There's no way a gaussian curve (peaked on teams being even), not even a very broad one; you kinda have 2 peaks where you should instead have the tails (you team is far better - the other team is far better).
Something it's just wrong, it's just evident, and it's pretty clear why it happens (= you gotta play more).
Kai Guy1 votesMMR because of... reasons.
People are easily distracted. Tiers and Divisions work as a shiny distraction for folks who otherwise would only ever lose MMR all season. I don't like it being hidden but its not a mystery why. Player retention + popularity of the que.
You can just take into account what shoes you gave to everyone, and use that new (hidden) parameter as you want. Nobody else will know, but eventually someone (wise enough) will notice that somehow you still know who is actually taller.
Cant, Apply context of MMR. Your using the marks on the wall to judge height not the Hidden knowledge of everybody's exact height. MMR does not know hidden perimeters for skill magically at the start. It builds off natural averages and comparisons to the average. It needs to be able to go around and look at each mark on the wall, if you make every mark the same height it has no way to find that Hidden perimeter. If it cant find hidden individual skill it cant use hidden individual skill to rig results. You cant force results to a point it represents a majority, the more you do it the less accurate skill becomes measured. The measuring point starts equal for everybody, so the average will pool around that and above and below will move at a rate built off how much they stand out.
Also there's Zero reason to Force results. Your time to climb for the average player by definition takes eternity. So... cant really out grind that. The level of complexity to build a good Rigged system is far higher then functional MMR.
Again, you have no explanation why games/team comps are so often one-sided
Look at attitudes and tell me why would it not be frequently one sided? Players Give the fuck up at 2 kills, give up If they lose lane, give up in champion select. Players don't play to win and roll over to move on to next. Moba gameplay frequently is snowball as all hell. Dota, Hots, Smite. Not a league exclusive issue. The # of folks who just stop playing to win is stupid high and I would honestly rather they AFK over soft int all match.
There's no way a gaussian curve (peaked on teams being even), not even a very broad one; you kinda have 2 peaks where you should instead have the tails (you team is far better - the other team is far better).
You cant use a single game and binary results to build a gaussian curve. MMR wont care if its 1 sided or not. Its + or -. Curve is built by the starting point and distributes around that. If gameplay is snowball games are snowball. That should be self evident. As MMR does not remotely give a dam about how snowball gameplay is, trying to build a curve around gameplay and how 1 sided its perceived by you does not have any relation to the system.
If riot wants to force more # games they can run low K factors, instant and 100% consistent impact for the entire ranked ladder. Cant say I am seeing that. https://imgur.com/a/ZvlF5Fo I did my best to sink this alt account to low MMR. Placed Bronze 1 by abusing Remake counting as a loss and their new "predictive" placements. As I never played to int my MMR remains above the alt accounts Title and The system wont let me stay low, skipped promo to Silver 4. Skipped from silver 4 right to Silver 2 with 1 win. It very clearly not trying to make my experience a grind and it wants to place me into the right Rank.
So, Please understand that Standard MMR systems wont check gameplay, all it sees is a Win or loss, so gameplay as proof is inherently flawed as its not related. Universal system from probability and averages, how is this not getting thru to folks? yall don't like snowball gameplay then look at gameplay, Is it behavior or Balance? Even Performance influenced MMR systems still use W and L as the metrics for ladder movement and personal stats just add or lower the speed for that.
Study the system. Read the wiki. Start With Elo, Then move onto MMR. Take a look at Microsoft Trueskill if you want to see the mathematics used to adapt Elo into MMR.
I don't know what else to do besides be blunt and honest and try to explain the concepts used to folks who are clearly unwilling to learn the functions of something they wanna rage about.
The Kombinator1 votes[{quoted}](name=Kai Guy,realm=NA,application-id=3ErqAdtq,discussion-id=KEFFbONc,comment-id=0001,timestamp=2019-10-03T21:13:33.056+0000)
Lets take a look at the match averages for the players in recent games to see if this was a gross mismatch.
Vs.
All in all pretty similar, outliers being the duo on the enemy with the lowest and highest MMR in the game. Annie having low Data is a high uncertainty account but Dianna looks to be lower MMR then her.So lets try another site. One with Estimates. https://na.whatismymmr.com/
823±2, 826 ±22, 927 ±26, 760 ±35,797 ±37. VS 807±20, 811 ±10, 864±30(you), 823±21, 823±2 (your "bronze" teammate)
Averaged out? 826.6 Vs 825.6 Looks more or less about right ball park, Some Noise from further games played impacting accounts MMR and the fact that loss would have changed results. Nothing looks outrageous thou.
Look, MM has to balance out Duos. So seeing a game that's around the Mean value between Bronze and gold seems about right.
I have been telling boards all season that Duoing implementation really needs to be updated to function off Assigned hidden MMR not Tiers because the Range gaps create Mismatches.Hopefully you will lean that Titles are a bad metric to use to gauge skill.
You greatly overestimate the value of s8. rank. There are 2 possibilities.
For easier calculation let's pretend, that Bronze4 is the bottom.
(2+5+6+7+8)/5=5,6
(5+3*6+9)/5=6,4
The average difference is 0,8 division.
Starcraft243ver1 votesYou do not seem to understand that ranks are irrevelant and are not even looked at by the matchmaking algorithm.
420 grams4 votesNo
KVbqbFsC8e3 votesWhere do you see any "massive" rank disparties?
A Gold 4 with 500+ games and Silver MMR?
The Bronze 3 Annie with just 9 games played and two previous Seasons at Silver?
Everyone in that game is pretty much Silver.
Noodles9122 votesThats probably because the team with higher elo has lower MMR, so the lower elo team gets matched with them. Also the game cares about fast matchmaking only, so the odds of it being efficient is rare, but there is that OP -3lp dodge everyone should abuse as a last resort.
BackdoorPrincess2 votesHonestly even the winrates don't match up, even if the "mmr" does (it doesn't). An entire team with negative winrates shouldnt be up against a team with positive winrates especially when they all have hundreds of games, making them not outlier numbers. Combine that with the higher elo person having one of the positive winrates, when the entire tier lower team has losing rates, this game, like many others, was loaded beyond belief.
I recently dipped back into ranked to get my rewards for the first time in forever, and I have a 63% wr with only ~30 games, but when you look at the games that I won/lost, the difference is almost always a person that has a huge winrate for the elo they're in, put directly opposite someone that has a sub-50 winrate (with a ton of games played). These aren't "oh this toplane is gold versus gold but the rest of the teams are silver, with this adc being upper bronze", it's "hey this gold is directly opposite this bronze, and this entire team is, just in general, a more winning team".
Some of those wins I have that were against those stacked teams are straight up from refusing to surrender and just catching the fed enemy making a mistake, then winning the game from it at 30+ minutes, because damage at this point allows a push to pretty much win at that point, as the enemies will be frantic enough to rush in 1 or 2 at a time to try and stop it. Granted, this is at silver/gold/plat elo, so catching those mistakes is fairly easy as game mechanics haven't really taken hold, and meta--game certainly isn't a thing, but still these are wins that statistically wouldn't have happened if not for direct action I took, creating a small counterwave that grows to a mega-push, sacrificing a wind drake for 2 top towers, sneaking rift on botlane while my team starts another losing towerdivefight near top, flashing to stun the fed vayne that stepped out a little too far, forcing a skirmish/standoff at baron while they have a bot inhib down that takes 1/both home towers.
These games are not games, they are babysitting where you hope your team will not throw a big enough tantrum to lose, while you try to calm people down and politely suggest trying X instead of Y. If you learn to play the bigger picture, those kill disparities won't matter as much in the lower elo, which by the time you reach upper elo's, your teammates shouldn't be dying as much to cause that big of a difference in gold. You see all the Challenger/Diamonds do iron-to-diamond streams/series, and it's either from one-tricking(sometimes) or it's from knowing the meta-game or knowing the importance of cs/wave manipulation/map pressure.
The amount of times I take tier 2-3 tower as Illaoi because no one wants to fuck with me when i'm as ahead as I am, because they either have to a)contest dragon/herald, or b)save that tower/laner, that is the choice you want to force on them. If you figure out when it's best to stay top and push/resist, compared to when it's best to join your team to 5v4 while they try to, you will do much better.
Big tickets to win : Play around objectives, make a huge mess toplane when your jungler wants dragon, attract as much attention as you can to make it safer for them. Your team wants to secure baron? Shove the everloving shit out of botlane, and shake that ass hard enough that 2-3 come to kill you while your team grabs it. Don't towerdive for that kill if it puts you <50% hp, instead ask jungle for pressure or keep them zoned from cs. Successful gank? You bet your ass you should grab at least one plating before you leave (unless all super low and someone is coming or whatnot). Kill stealing doesn't exist. Period. This is not challenger/masters/lcs where everyone is the best, do not assume your vayne will get that last auto if the kill is right there, the jungler could be waiting, there could be a bait, whatever it is, remove them from the map and move to an objective. FOCUS ON CS. If you have to take minion aggro to harass, it ain't worth it. Your 100 damage ability is not worth taking 8*20 damage procs from ranged, let alone if the melees get you. A kill is on average worth about 16cs (without shutdowns), learn which champs suck at waveclearing, which will shove the lane passively by trying to harass, and which thrive under-tower farming, and use it. Communicate. Tell your jungler before game starts, or during the first 30 "hey this is a nasus, gank top a ton and we'll make him useless", "hey they counter me, so I'd appreciate help", "I counter them hard and this is my main, focus on bot/mid, i'll say if i get too much pressure from their jungle", "hey drags up and their mid is dead, i'll shove top to get their lee to come up, secure that shit" BUY A FUCKING PINK I dont care what lane, what role, whatever, anytime you have an extra 75 gold you should pick one up if you haven't already. Instead of buying that longsword on your first back, grab boots and that pink. Got a kill? Grab a refillable and a pink, boots and a pink, whatever. Can't finish that item? Grab a pink also. Are you ahead? Buy a pink, starve their jungler. Are you behind? Grab a pink, prevent a dive, give your jungler a reason to come up there. If both an adc and support get one on the first back, you can ward your jungle bush and your tri/behind turret bush. Top can do the same, mid can pick for a river or defensive bush ward. If you spend 75 gold and the jungler goes to that bush and finds it, congratulations, you spent 75 gold to not blow sums/die. Vision wins games, period. If you are losing, please don't pink river, it is literally a waste since theyll clear it the second they get a chance. Play to your teams strengths, do you have a lot of dive characters? Do you have a kai'sa/kog and a lot of tanks? Do you only have long range poke? Figure out which and you know how long you have before your team falls off, and to not dive if you have a braum and darius when you're talon.
Kai Guy1 votesI got to say, i don't understand. What's the end goal of splitting ladder by rank by WR as well as MMRs?
You have either An fucking awful experience of MM trying to balance teams to have 50%, which punishes you for having a good Win rate with awfull teammates to counter balance that. OR give you great teammates all the time if your awful. Not ideal if you want to sort ranked ladder by individual skill is it?
The other alternative I see is you break every Tier and division into a Bad WR, average WR, and high WR que.... which seems pointless because weak WR should sink out of a division and Ranking and High WR should climb so why the fuck would it be better to divide that up into multiple skill levels for each rank rather then just move individuals away from the locations they have high WR or low WR in?
Or am I missing something? How do you use individuals WR with MM besides one of the two above?
BackdoorPrincess1 votes[{quoted}](name=Kai Guy,realm=NA,application-id=3ErqAdtq,discussion-id=KEFFbONc,comment-id=00050000,timestamp=2019-10-04T10:19:26.190+0000)
I got to say, i don't understand. What's the end goal of splitting ladder by rank by WR as well as MMRs?
You have either An fucking awful experience of MM trying to balance teams to have 50%, which punishes you for having a good Win rate with awfull teammates to counter balance that. OR give you great teammates all the time if your awful. Not ideal if you want to sort ranked ladder by individual skill is it?
The other alternative I see is you break every Tier and division into a Bad WR, average WR, and high WR que.... which seems pointless because weak WR should sink out of a division and Ranking and High WR should climb so why the fuck would it be better to divide that up into multiple skill levels for each rank rather then just move individuals away from the locations they have high WR or low WR in?
Or am I missing something? How do you use individuals WR with MM besides one of the two above?
Same players in game, different teams set up. There is no reason for one team to have entirely all positive WR's while the other doesnt have one, its yolo/duo so swap some of those teams up so that both teams have positive and negative WR's. WR's are not always entirely reflective of a champion being too strong or weak (according to based rito), but it does reflect a players overall skill with said champ and ability to adapt to situations. If I see someone below a 45% WR you best believe I'm gonna camp the everloving shit out of them, either tilting them beyond belief or straight up getting them to quit, and if you're entire team has winrates similar to that, they're probably gonna give up as well.
It would be entirely possible (and not even a lot of work) to have it balance around winrate secondarily, first sort it by the mmr, then have it position so WR's are a little more even on both sides. Swap the toplaners or jungles or mids, whatever. Duo's can be accounted for by placing the higher WR's, if any, on the opposing team since the duo's would have more communication (in the case of only one set of course). There's already the existence of autofill that would make this even worse, if you have an already losing winrate just to get thrown into a position you're unfamiliar/uncomfortable with.
Your assertions are bad and are filled with fallacies, and the entire Elo system originates in chess where there is exactly one person on each side of the board, not 5 on 5. There's no reason to "split the ladder", but there is a reason to not stack teams by giving one team players that are definitively better at adjusting than another. There are such things as unwinnable games, lower elo does tend to turn into stomps more than anything, but those are also the games where it's easier to catch a mistake and turn it back around, but a player has to have the knowledge how to, which just isn't the case down there. Metagame mechanics are weak below diamond, and winning a game because your team can outrotate, focus objectives better, or catch a single fed person out, when you're massively down, just feels weak and like a luck victory, not anything you actually learned from.
Kai Guy1 votesits yolo/duo so swap some of those teams up so that both teams have positive and negative WR
So you harm good players with Weak players to counter balance the teams WR out eh... Gee that sounds a bit like "forced 50%" and "Rigged MM" RNG is not always fun but its Fair.
Elo's system has nothing to do with chess any one who argues that its limited because of the game it was developed for has clearly not studied the topic.
Rn=Ro+K(W-We)
The only single variable that has any relevance to chess is W. W is built from Win, Loss, or Draw. So it turns out Wins losses and Draws are not at all exclusive to chess. Please don't ever use that argument again.
I own Arpad Elos book on the topic here is a Direct quote from Professor Elo " Although the system herein described was originally developed for rating chess players, its application is by no means limited to chess. It is infact a universal system, applicable to any type of competitive activity in which individuals or teams engage in pairwise competition."
MMR is the adaptation to account for individuals separately from a team. Elo works just fine for a team game with a set rooster but to make it function for indvidual raiting as a portion of a team there are steps needed to account for that. At the most basic its done by accounting for K variables on a individual level so that consistency in play despite RNG numbers as trends in play are consistent via an individuals skill and can be measured as an assumption of their average.
So to be very clear with what your asking for. You want to punish Players with positive WR by placing Negative WR to balance out teams rather then leave it to RNG that treats everybody equally and also functions with the mathematical laws of Averages and Large numbers to brute force luck out of the equation over time.
I have spent a good deal of time studying rating systems and the concepts behind them. If you haven't even glanced at a wikipage you probably should do so before replying. Microsoft Trueskill shows the math for a MMR system if you don't know where to find one.
Forceing teams by WR is stupid. # of games impacts WR change, it does not add quality to MM, and there's no gains in sorting skill.
Niggel Wessom1 votesThat's basically bronze you should just carry. There is something you are doing or forgetting to do that's keeping you there and it isnt your team.
Dope Solo1 votesAn entire team of negative win rates should not be matched against a team with positive win rates- period. This is, statistically speaking, guaranteed loss matchmaking. A phrase that...apparently will also get your thread deleted if you dare to bring it up.
For those people that think win/loss ratio is the only thing that goes into your mmr, here's some more evidence to suggest you're just totally ducking (and tragically, cluelessly) wrong:
Riot is actively calculating your probability of losing at every stage of the game.
Subdue2 votes[{quoted}](name=Dope Solo,realm=NA,application-id=3ErqAdtq,discussion-id=KEFFbONc,comment-id=0007,timestamp=2019-10-04T17:53:11.997+0000)
An entire team of negative win rates should not be matched against a team with positive win rates- period. This is, statistically speaking, guaranteed loss matchmaking. A phrase that...apparently will also get your thread deleted if you dare to bring it up.
For those people that think win/loss ratio is the only thing that goes into your mmr, here's some more evidence to suggest you're just totally ducking (and tragically, cluelessly) wrong:
Riot is actively calculating your probability of losing at every stage of the game.
First of all, win/loss IS the only thing that factors into your MMR.
Second of all, your "evidence" isn't evidence of what you're suggesting. There are plenty of reasons for Riot to calculate win probability based on variables, such as balancing the game.
Finally, win rate is a useless statistic for matchmaking. If Faker has a 65% winrate against Challenger+ players, and a random Platinum smurf pulls an 85% winrate in Iron, would you rather have the random Platinum smurf on your team than Faker? Probably not. In the same token, someone who was very lucky in their provisionals and had their MMR escalated above where it should be would eventually have a negative winrate as his MMR came back down. Likewise, someone who was very unlucky in the provisionals would have a positive winrate as his MMR climbed up. The two players could be exactly equally skilled, but due to the circumstances of their early games, have different winrates. This is further affected by MMR carrying over from season to season.
Dope Solo1 votesSorry man, but the evidence points directly to the contrary. Matches would never be set up with all (+) vs all (-) win rates if you were correct, and mismatches (total stomps) happen a lot lately. There is a whole heapload of data being generated every match by each player, all signs suggestedly point towards (this is called interpolation and extrapolation using available information for your future reference) riot actively using that data, and in all probability one of those uses is to generate matches - they just aren't that good at what they do. I don't blame you if it isn't obvious, I had a graduate school science education.
Subdue1 votes[{quoted}](name=Dope Solo,realm=NA,application-id=3ErqAdtq,discussion-id=KEFFbONc,comment-id=000700000000,timestamp=2019-10-04T23:47:07.785+0000)
Sorry man, but the evidence points directly to the contrary. Matches would never be set up with all (+) vs all (-) win rates if you were correct, and mismatches (total stomps) happen a lot lately. There is a whole heapload of data being generated every match by each player, all signs suggestedly point towards (this is called interpolation and extrapolation using available information for your future reference) riot actively using that data, and in all probability one of those uses is to generate matches - they just aren't that good at what they do. I don't blame you if it isn't obvious, I had a graduate school science education.
I literally explained WHY the winrate value didn't mean anything to matchmaking, and your response is "they wouldn't do that..." There's no reason NOT to do that, because THAT doesn't mean anything.
Dope Solo1 votes"First of all, win/loss IS the only thing that factors into your MMR."
This is wrong.
Starcraft243ver1 votesImagine you start a new account and get a 500 lose streak, you will hit the bottom of the mmr range. Then, since you were a challenger smurf, you go on a 100 win streak, you'll hit master without a doubt, yet you have a 500L-100W record, which is around 17% winrate.
It doesnt make your matches not balanced.
Dope Solo1 votesTrue enough, but matches are imbalanced a significant portion of the time now, and in stark contrast to last season.
The Kombinator1 votesImagine you start a new account and get a 500 lose streak, you will hit the bottom of the mmr range. Then, since you were a challenger smurf, you go on a 100 win streak, you'll hit master without a doubt, yet you have a 500L-100W record, which is around 17% winrate.
It doesnt make your matches not balanced.
Smurfing, and int lose is not something any matchmaking could possibly handle.
Starcraft243ver1 votes[{quoted}](name=The Kombinator,realm=EUNE,application-id=3ErqAdtq,discussion-id=KEFFbONc,comment-id=000700010001,timestamp=2019-10-09T17:47:33.589+0000)
Smurfing, and int lose is not something any matchmaking could possibly handle.
I was just pointing out that win ratio alone can't be a good factor for matchmaking.
Starcraft243ver1 votesObviously massive rank disparities
Ranks are not considered by the matchmaking algorithm, so why do you look at the ranks ?
The Kombinator1 votesRanks are not considered by the matchmaking algorithm, so why do you look at the ranks ?
Ranks should reflect the MMR above a certain number of matches. A relatively low number for true accounts, and a higher number for smurfs. Assuming, that the player doesn't throw matches intentionally.
Starcraft243ver1 votes[{quoted}](name=The Kombinator,realm=EUNE,application-id=3ErqAdtq,discussion-id=KEFFbONc,comment-id=00080000,timestamp=2019-10-09T07:30:15.822+0000)
Ranks should reflect the MMR above a certain number of matches. A relatively low number for true accounts, and a higher number for smurfs. Assuming, that the player doesn't throw matches intentionally.
Assuming the player doesnt change his champion pool Assuming the player doesnt decay due to inactivity Assuming this is the same player on the account
The Kombinator1 votesAssuming the player doesnt change his champion pool Assuming the player doesnt decay due to inactivity Assuming this is the same player on the account
In other words. The player is normal, and active.
Champion pool change doesn't matter unless he is the "ranked exist only" guy. Draft pick doesn't affect ranked mmr, and you shouldn't play champions on ranked without practice on normals first.
Starcraft243ver1 votes[{quoted}](name=The Kombinator,realm=EUNE,application-id=3ErqAdtq,discussion-id=KEFFbONc,comment-id=0008000000000000,timestamp=2019-10-09T09:37:59.260+0000)
In other words. The player is normal, and active.
Champion pool change doesn't matter unless he is the "ranked exist only" guy. Draft pick doesn't affect ranked mmr, and you shouldn't play champions on ranked without practice on normals first.
I've changed my champion pool a few time over the years and everytime i dropped a full tier before slowly getting back to my former level.
The Kombinator1 votesI've changed my champion pool a few time over the years and everytime i dropped a full tier before slowly getting back to my former level.
That's why you play normals first, and not start ranked with first timer champions. You still get less efficiency of course, but not bronze level, if you are in gold otherwise. Unless you intentionally reduce your normal MMR drastically, but as told.
Int losings, smurfs, and e-bay accs. are not something any matchmaking system could handle.
FioraWillCarry1 votesLooks pretty normal to me. Don't see a problem here. It's functioning correctly.