Enhanced ranked matchmaking

DarthArchon·Jul 9, 2019·11 votes·7,083 views

It feel to me riot is very prone to implement measure about chat but still do nothing to improve matchmaking after 9 years! Only real change was position matchmaking which was a good idea that honestly took too long to be implemented.

The fact is there are 15 different things that could be done to improve matchmaking, like what kind of player are you and you are looking for. E.g I'm a very aggro player, i make my own opportunity and i take some risk and i know from experience that if you put me with someone who only play defensively and care about his KDA, neither of us gonna have fun, cause i'm gonna die trying to last hit a guys when he walk around behind me cause he wasn't full hp for that one. My style of play is incompatible with a defensive player.

After 9 years you still do virtually nothing about trolls, griever and those who play champs they have no idea how to play. like ranking recently release champ and do whatever.

Gotta be honest i got banned for raging on players, but you really make me feel like it's the only way to make some player not play ranked when they're not ready, since nothing you do goes along making the ranked experience better and more competitive. It should be ok and conceivable to ban people from ranked because they played poorly.

Realistically LOL as stopped growing and you guys don't realize that competitive player, like me, are going away because you don't show that you care about them, it's hurting your bottom line without you realizing it. I'll star buying skins and emote again when the game will look sustainable to play for me. currently i win 30 lp and lose 12, meaning i'm way under ranked and yet ranking feel grindy AF and every of the games where i got a player i need to carry all the time are psychologically straining and What's in there for me? Basically i just need to wait for player with about the same skills as me to have an easy win. If i don't i'm gonna have to carry someone to a probable lost anyway... un fun, make me wanna never play this game again, specially now that you've given me 14 days to think about other games.

Banning people for chat related offences just make people stop using chat so why don't you remove it altogether and make better matchmaking?

24 Comments

UrarakaJul 9, 20193 votes

Because if matchmaking is made better, players won't play double or triple amount of games for climbing. I agree that chat based offenses needs to be punished as it makes the game toxic, but sadly it seems that they don't even care for anything other than skins and banning for toxicity.

Kai GuyJul 9, 20191 votes

Literally could just lower LP gains to lower value by lowering MMR gains. Also Your arguments kinda stupid.

Time to reach the top of an Elo system for the average player is expected to take eternity so your "its to make things grindy" Argument is pretty weak.

Yes... Riot wants to add more time to infinity. Brilliant deduction there Watson.

DarthArchonJul 9, 20191 votes

I understand literally nothing of what you're saying and i don't even see how it relate to what i'm saying.

Care to rephrase what you've said to make it more tangible?

Kai GuyJul 10, 20191 votes

I was not looking to address what you said. It was to Maayon. So, My comment to Maayonmayyan is to point out the flaw in his claim. That riot wants bad MM to create a longer grind for players. You can simply have lower Point gains with a functional system to accomplish this and enjoy good MM. The benefit of a weaker MM system does not inherently guaranty an increase in player time investment to achieved end results... you have no way to really predict the outcome for all accounts accurately when you just throw randomness into it and at the same time actively looking to weaken your MM harms your game.

MMR systems can be tested to check for accuracy and set system expectations for # of games for accuracy. Meaning you have the tools to adjust how long it takes to become an accurate value. Its insanely easy to have a functional system with decent MM and just make it grindy AF with value adjustments.

As you seem interested, Let me directly address your post. Your MMR is higher then your assigned Tier and it looks like its a solid 2 divisions above. You have a net balance of 2 wins. Your sitting at 32W v 30 L. My guess is this is likely due to K variuable but might be WE depending on if you duoed a lot early on. I don't want to lie or mislead you this is just my guess. I don't have riots adjustment formula, I also don't have the desire to go over and check your entire season, or find out exactly what your preseason MMR was.

My assumption is that 2 divisions does not make over a class interval so there is no reason to play catch up. I can tell you that Riots system does play catchup because I tried to abuse it in the past by dodging promo games to enable me to que with lower MMR friends as it stands to big of a gap and you skip not just divisions but you skip promotional matches completely. What you should expect is if you sit at a 50% win rate currently that you will end up as Gold 1 or Gold 2 owing to that being the MMR value your account currently seems to have. Your LP gains are by your own claim slanted to help move you to that range and even if baddies flop promos your gonna get a freeby so 50% will promote you. I am willing to go so far as to say that If you can maintain a higher % your going to make it to plat ranking on your account. However, at 52% its gonna be a bit of a grind and your unlikely to see a large enough gap to ever skip divisions. Again, not my system don't have their data so don't discount that I may be wrong but I feel like this is a reasonable expectation for me to try and set for you..

Back to addressing Maayonmayyan . Throwing shitty MM at players does not make it more grindy remotely as reliably as a very simple and reliable method achieved by setting lower point values per match. Because lower gains = a consistent increase in time for any and all accounts in any and all directions on the ranked ladder. If I halved the MMR gains for all accounts then it would consistently take 2 times longer to reach the correct ranks 100% of the time for 100% of players. Basic math.

Pretend its my game, As the dev I can set exactly how many games my system would take any player to reach any rank because in MMR/Elo systems the point values are arbitrary values, they behave consistently and in relation to probability theory and distribution models. What You can change the value of K... which impacts the weight early games carry and reduces over time and in relation to ladder position. What you can change is how many points are generated out of (W-We.) Hell you can make your C variable be any value you want and adjust your We expectations accordingly.

I will leave a data dump with information about Systems at the bottom. The exact opposite of a TLDR I suppose. Skip to the Bolded "VARIABLE TRANSLATION" if you want to understand the variables I mentioned in the last paragraph.

So as the systems are made of probability theory, a normal player will never dramatically climb the ladder. No matter how many games are played the expectation is that once they are in the correct location they wont climb or sink and just float around the Range the belong. MMR and Elo Points are not an absolute value but represent the Mean average of a players performance. Its built with context of who you play in addition to the results of the games.

Context of who you play matters. If your playing to win, then you see a pretty dramatic difference in your results in a game you vs a 3 month old infant or say Faker in lane, one of the reasons "why not use Stats" fails as MM so badly. Not to mention how its consistently easyer to pad stats then play to win. Karthus who hits R every time its off CD has more damage from that ability then the one who saves it for kills or to help teammates, easy example to make.

** Variable Translation guide.** Context. Elo's adjustment. Rn=Ro+K(W-We). In English. Your new raiting after a game (Rn) Is your Old Rating (Ro) Plus the result of the match (W) adjusted to account for system expectations (We. We being a more complex variable that's made of a probability model that adjusts for gaps between players Ro values.) Which is multiplied system uncertainty so that it moves players more rapidly early on to correct ranges (K. uncertainty variable. Higher with lower # of games, decreases traditionally when you have higher # of games and also at benchmarks on the ladder to prevent luck boosting some one to a range they get dumpstered at.)

C is the variable used for Class interval. This is a basic expectation for wins or losses between a gap between a players raiting (R). In chess using Arpad Elos system C is 200 points of E (E = 1 elo point.) Various distribution models can be used and they impact system expectations differently. For example the odds for various modles at 1 gap of C with Elo. Normal distribution for elo would be 3.17 odds. Logistic model would be 3.16 odds And linear would be 3.00 odds.

It gets dramatically higher in a gap of 2C (400 point gap). Odds look like this. Normal =11.95 Logistical =9.99 Linear =15.10

Of them... Logistical is the best thou normal distribution is workable. I assume Riot is using some kind of Logistical modle due to the fact it works well with automation.

The numbers I provided and algorithms are taken directly from Arpad Elo's Book. "The Raiting of Chess players Past and Present." First printed in 1978

UrarakaJul 9, 20194 votes

currently i win 30 lp and lose 12, meaning i'm way under ranked and yet ranking feel grindy AF and every of the games where i got a player i need to carry all the time are psychologically straining and What's in there for me? Basically i just need to wait for player with about the same skills as me to have an easy win. If i don't i'm gonna have to carry someone to a probable lost anyway

What the op said clearly means that their MMR is above current rank, probably one or two ranks higher. They're placed in a lower rank because matchmaking isn't good and be forced to play more than the required games to get that rank; MMR should be rank directly (as in many games) and when it isn't the matching should not force players to play more games to reach their rank.

Idk why you even consider this as climbing to top or something related to that, maybe you're just shit posting or just too stupid to even understand something so basic.

Kai GuyJul 10, 20191 votes

You wish to talk about Basics?

Setting lower MMR gains is a pretty clear cut and basic way for a dev to 100% accurately impact their time for accounts to climb. That's kind of the body of my comment in addition to pointing out that MMR expects the climb to be an eternity for the average player. Its not hard. lets say your trying to reach 100$ starting from 0$. As your boss I pick your pay grade. So at any point If I decided to pay you 2$ an hour instead of 4$ then it takes 2X longer and as I can do basic fucking math I can tell you this is 25 hours of labor VS 50 hours of labor . Hours can be replaced with # of games pretty easy when you shift the context to climbing to a rank from the Starting MMR or shit even the soft resets which...you know... the devs get to set the fucking values for?

All im saying is Its honestly that easy to adjust your MMR value changes cuz its an arbitrary number. Riot could literally just decided to halve K variable values across the board to do this. Very simple change they could make any time they want. There's a negative to it thou, it Would make smurfs/trolls more obnoxious.

Guess you missed this basic point I wanted to make. that's ok. Happens, I don't always convey information to others very well. So, here is more clarity on the topic so please contemplate it now that I have made an effort to draw your attention to it.

OP has a net total of 2 wins over losses over 62 matches. Yes going off his OP.GG and also his LP gains the fair statement is that he is 2 divisions ahead. No... that's not really a significant amount. The dude should end up gold 2-1 even if he only ever goes win loss for the rest of the season barring dramatic duoing or player population MM issues generating massive WE variable impact. Assuming Tiers = class intervals that's something like 100 so …. no idea WTF riot uses but in a Elo system that would be 64% to 46% if the OP was VS some one at his assigned tier that's 100% accurately placed.

The main advantage for riots LP system is you can set more dramatic MMR gains that don't provide a incorrect title at a high K value. this helps deal with smurfs and Trolls. There are also gains in your ability to punish cherry picking behaviors for folks trying to game lobbys and prevent dodgeing from turning any account into a lower MMR value.

Remember, MMR is a range. Your expected to be inconsistent and dude its not supposed to EVER lock to a single value and it never will because the concept is to constantly be self correcting.

So no, i'm not shit posting your comment was ignorant so I wanted to explain some basic issues with it. I don't mind I am always happy to educate folks. Cuz then I can get yall to focus on the shit Riot Royally FUCKS UP. like Duoing. Its shit in how they implemented it. The game is better if that gets addressed.

Elo and MMR are proven and well documented and have been peer reviewed for years. I personally consider any one with 0 studying on the topic to claim its a failure to be an idiot similar to antivaxxers or flat earthers. It frustrates me that people don't get the basics. in a PvP game. PLAYERS VS PLAYERS! Gameplay is directly impacted by Player behaviors because that shit is what drives gameplay.

I also don't want to ever give riot a free pass on leaving bad systems in place. 1. Duoing needs fixing. 2. Player behavior systems needs to be a high priority always. I cant tell you what Red has that job... Lyte used to but now a days? No clue.

Let me know if anything is confusing or unclear.

UrarakaJul 10, 20192 votes

The problem is that even after 60 games,a player isn't in the rank he is supposed to be. Check other games which uses elo rating to see whether the players reach their rank faster than this or not. I have played chess and multiplayer games where this is used and few mins games, with ratings being transparent and mostly mmr/elo rating directly translates to rank.

Only in LoL i find that players are placed in ranks way below their rating and even after 100 games one can be in this situation; did the player improve with every game, to have such a situation? Skipping promos or skipping a division is good enough for a reward, but after that matchmaking screws up things and makes the climb slower.

Changing K value is one way and together with it there should be hard penalties for smurfs and better smurf detection; changing matchmaking is another way. Yes, for duo it is worse than the solo and my point is entire thing should be fixed.

tbh, they have come up with the auto ban system for toxicity and so there is no need to concern about it. Because of players understanding that they're still not in the rank their rating is, they are forced to play more games to climb and at one point it becomes simply frustrating and not fun to play.

Worst part is above this, matchmaking is inclined more towards giving games faster to have low queue times and makes many games unfair; this adds up to the number of games to be played while reducing the fun even more.

Kai GuyJul 11, 20192 votes

The problem is that even after 60 games,a player isn't in the rank he is supposed to be

Again. 60 games is nothing. Chess is 1v1. Less variables. Accuracy requires a lower # of games. As I own Arpad Elos book on the topic I can happily share his system expectations for correctly placing a player to the right class interval with you. 25 games = 92%. 30=95% 60G=99%
100 = 99.9%

Now lets move on to MMR. Microsoft publishes their TrueSkill system and its a commonly used MMR system so I tend to use it for my discussions about MMR. 2 Teams/4 Players Per Team 46 2 Teams/8 Players Per Team 91 ok.. That's not bad right? Oh... wait "The actual number of games per gamer can be up to three times higher depending on several factors"

So we can make it into a Range then. 2 Teams/4 Players Per Team 46-138 2 Teams/8 Players Per Team 91-273

Pretty clear that more players = more time to correctly locate the appropriate class interval for a player. Even if we used the 4v4 model then its clear that the OP sitting at 62 games is well with in the correct range for class interval calibration.

Feel free to request any game you play with a MM system if you want me to take a look at what those devs do and what their system expectations are. If the publish the information then I tend to be pretty good at finding it.

Lets throw in the context of riots LP system, they are seeing slanted gains due to account MMR not reflecting the assigned Title. "30 lp and lose 12" So win loss win loss pattern would result in net positive of 18 LP every 2 games. resulting in his MMR and Rank collapsing into the correct representation.

The advantage of LP system is you can have more dramatic MMR gains early on to reduce how long accounts spend time mismatched in the incorrect class intervals that you can not do with a "purist" Elo/MMR system. The down side is that inside of the correct interval takes longer to calibrate to high accuracy for titles when some one is in the proper range.

As mentioned. Id just have titles locked so that player need to participate in the bare minimum for average expectations of accuracy and completely rework how Riot runs Duo queing.

Worst part is above this, matchmaking is inclined more towards giving games faster to have low queue times and makes many games unfair; this adds up to the number of games to be played while reducing the fun even more.

Part of the problem is most players only care whats fun for themselves with no consideration to other players. I also am inclined to agree that I feel riots MM does generate games to rapidly at a cost of quality. I want to be clear that's just an oppenion of mine that I cant prove.

Changing K value is one way and together with it there should be hard penalties for smurfs and better smurf detection; changing matchmaking is another way. Yes, for duo it is worse than the solo and my point is entire thing should be fixed.

I hate riots implementation of duoing. It wont break the ladder. It does make mismatched games of low quality. Its left alone because its popular. LP is a system to deal with smurfs/trolls/newbies by allowing higher Weight on early games to have trends of play more rapidly adjust location. Admittedly you can also accomplish this with Provisional matches. Also... despite the fact this is in theory possible with out seeing exact numbers I cant in good faith tell you this is done. Its just theoretically possible.

Pub stomping is old AF gamer behavior and typically pretty selfish. Same with Greifing. Its Difficult to manage in PvP games but is a worthwhile attempt.

Overall. My personal solution is to encourage full premade teams via Clan/guild systems because a consistent Rooster of players results in a team being a singular entity for the purpose of rating skill. I believe Leauge has the playerbase population to support this kind of system.

tbh, they have come up with the auto ban system for toxicity and so there is no need to concern about it.

Nah theres always a reason to be concerned for it. Toxic behaviors more then flameing and running it down. Building player culture is worthwhile and makes the game better. Always be looking to improve and lower toxicity in the games player base is a huge deal. It shows up in games very fast due to population size and the fact there's 10 individuals in each match.

More then anything else riot can do, maintaining healthy players maintains healthy gameplay. Gameplay is impacted most by players, more then balance or directional decisions riot makes. How people play the game is the single largest impact on how players experience the game.

I think I addressed everything. If you feel I skiped something relevant let me know Via a box quote and I will get back to you.

I take the time to educate folks about MMR so they can make reasonable assumptions. 60 games really Is nothing, folks need to be willing to put in 100+ games for a ranked season. Thinking its magic and can accurately place accounts with 0 data just makes frustration in the player base which often results in toxic behaviors. I get being competitive but man I just wish folks could keep in mind a simple fact. This is a game and it should be something you enjoy spending your time on.

UrarakaJul 12, 20191 votes

Wow, thanks for explaining everything in detail and sorry for misunderstanding your first comment(s).

Regarding rank in team games and elo:

Yes, the variables in team game is a mess and since performance doesn't affect MMR, there is this delay in reaching the rank; most of the games I play are 1v1 except Dota 2, so didn't think of 1v1 vs team ratings. But still, like Dota 2 has there can be something like MMR (range)= rank(+tiers)-maybe idk if there is hidden reasons for not having it; mmr is also visible there.

Regarding faster queue times instead of fairer games:

Well maybe it is a thing for support players, I hardly get queue times greater than 10seconds-I can take snapshots of every single game in this acc and if needed on a smurf acc to show this, as well as why it is unfair especially to a support player. my win/loss greatly depends on adc and my team overall, games in which my team has unfair matchups [especially when it is like silver 1 supp-me(I was silver 1 for many games)+adc-silver 4 or bronze 1 vs silver 2 adc+silver 2 supp] it is like 75% loss.

Regarding flaming and gameplay:

Though 60 games is a less count and I agree with an increased game count to get an accurate rank for the mmr with the current system, imo it simply becomes demotivating at one point when one has the rating to climb-but just loses because of bad matchmaking. What's more, it is one of the reasons for players to flame as they get stomped and the players(in and around silver, myself included sometimes) write more than "play safe" to ignite them. Frustrated with losing lane, they start wasting time in chat and once it starts, it just grows as many players just don't mute or ignore them. This quickly transforms to inting and funniest thing is they find ways to flame without getting caught, by writing stuff like fck you or fcuk you in place of fu*k you and ofcourse the automated warning/ban system for flaming doesn't work for this and it needs improvement over time.

Improving matchmaking is in RIOT's hands and decreasing toxicity is (actually) in players' hands; while we have an option to mute a toxic player and escape from it, we simply don't have an option to escape from the bad matchmaking. Imo after a point, it feels more of a chore than an enjoyment as a player plays good, has good mmr-but needs to play that extra 40-50 games to reach to his rank(by looking at Lp gain/loss). And this is what makes the players to go for boosting services (which ruins evreything big time), as they don't have to play more games or just buy an account with the rank they desire (high elo players buying low elo accounts is even worse than low elo player buying high elo account).

Kai GuyJul 13, 20192 votes

Wow, thanks for explaining everything in detail and sorry for misunderstanding your first comment(s). No problem. I typicaly consider it my fault when misunderstandings happen.

so didn't think of 1v1 vs team ratings Don't trip it. That's pretty common. Another common error that gets made is from the fact that Elo systems for 1v1 tend to be set as Zero sum. MMR systems are not because uncertainty needs to be tracked on a player to player biases. Its a common error to make mistakes when comparing the two systems. Now you can treat a team as 1v1 for an Elo system, but the precondition is that the team has a set rooster of players. When its RNG teamamtes they need individual tracking so individual probability expectations can be made to use for MM.

But still, like Dota 2 has there can be something like MMR (range)= rank(+tiers)-maybe idk if there is hidden reasons for not having it; mmr is also visible there.

Basically, it was more popular with players. The Tier division thing caught on really fast and its kinda standard in most games now even with MM systems that don't inherently use MMR or Elo. Month for month ranked participation increased as much as 2x over the same time periods last year while dodge rates dropped to about half. The number of leavers / afks in these games also remained well below most other queues. In other words the number of new ranked players and existing players who participate in ranked games is greater than ever before and the quality of the games improved. Ultimately we think this is the best measure of how successful a game type is

The above is 2013. Also more Info from Devs 2018. Using MMR as the sole mark of achievement in League punishes half of the playerbase as their MMR will decline over the course of the season, which sucks because most of them are gradually getting better at the game—but so is everyone else around them. Ranked tiers also provide contextual progression and status. Knowing you’re “a Gold player” as opposed to “a 1650 MMR player” or “120,353 on the server” gives you clearer targets to work towards. Moving from 1595 to 1600 MMR is probably not that compelling, but promoting from Silver I to Gold V should give you the knowledge that you’re truly improving. The Leagues system also gives you a bit of protection from losing a bunch of games in a row and having your MMR plummet as a result.

Personally I like Visible MMR scores, but the exact Value showing up helps folks figure out what K factor slowdown is set to which gives Smurfs and Trolls more opportunity to game the system.

Well maybe it is a thing for support players, I hardly get queue times greater than 10seconds-I can take snapshots of every single game in this acc and if needed on a smurf acc to show this, as well as why it is unfair especially to a support player

No need, I am also a support main. I don't quite agree with how much agency your giving ADCs, but I also tend to spend a good amount of time working on trying to get other roles up to par so most the ranked season when I do start to grind my overall position is lower then what I usualy can achieve making it a bit lopsided for my climb.

especially when it is like silver 1 supp-me(I was silver 1 for many games)+adc-silver 4 or bronze 1 vs silver 2 adc+silver 2 supp] it is like 75% loss.

You might find this interesting. Assuming class interval (200 points for Elo in chess with his OG system.) is a full Rank and using Arpad Elos Probability table for logistical distribution. His odds to win for that skill gap at low uncertianty are 76% to 24% for the higher rated player. So pretty much if your similar skill to the enemy support and your ADC is a full Rank lower then the enemy adc you actually do expect to see that number you listed of 3 to 1 odds. There is a reason I am vocally opposed to Duoing being left alone as Tiers and divisions, it needs to be off the same class interval MM considers "fair" To lower mismatched games.

we simply don't have an option to escape from the bad matchmaking Yea. But MM cant do anything about player behaviors. Even 10% toxic playerbase is a massive amount. As the gameplay is PvP gameplay is massively impacted by Players. 1 in 10 players being toxic, shows up almost every game due to the fact there are 10 players every match. Theres only so much that can be done in advanced as you cant accurately predict what some one is gonna do prior to the match being played, and at that point its to late. That's why I am always looking to get players to consider the importance of player behavior systems and the high value that removing toxic asshats from the game provides.

MMR is a bit rough for your average player, but its honestly rough for any one stuck at any MMR. The average player just ends up stuck basicaly right from the get go and does not feel good about it. Knowing that improving results in an ability to climb takes the edge off but trying to convince people to set aside their personal observation and experience Even when your providing the math, is difficult. Like you said, at times its honestly demotivating, which lowers effort, which lowers impact, which keeps folks stuck or turns some players toxic.

Now.. Statistics to impact players for variables like damage cc kda etc. Those have some issues but this post is pretty long so unless your actively curious why a stat system fails harder for MM over a Elo derivative ima wrap this conversation up.

Thanks for your time and I hope ya found it helpful.

KVbqbFsC8eJul 10, 20192 votes

The one thing that could drastically improve matchmaking is the one thing that will never happen. Removing duo que. I have had far more games ruined by duo que than I have with trolls/afkers.

Kai GuyJul 10, 20192 votes

Even if its not removed a fantastic QOL change is to Force Duoing to be restricted to the same class interval that MM uses to generate a fair match. MMR has a range it wants to use for quality between the lowest rated players and highest rated players. That's why you don't see a iron 4 and Diamond 1 player randomly show up on 1 team in say silver/gold MMR when that's the que they have those ratings in. (normals are discounted as there is no restriction for duoing so the two could be premade.)

An average team MMR is not just balance the two teams to the same average. It needs to have a range it wants to pull from, tighter is better. 50 is the average between 0 and 100 but its also 45 and 55.

So making Duo restricted of MMR values... not fucking Tiers and divisions would increase quality. Because silver 4 can have bronze MMR and Gold 1 can have plat MMR the two can duo and if you look at player distribution that ends up representing a skill gap of about 80% of the entire games player base. That's garbage!

I love elo and MMR systems. Duoing is one of Riots biggest fuck ups.... it wont break the ladder at the upper end and its left alone because its popular with players. So fine... But you want to up match quality. Remove it, or restrict it as a reasonable compromise.

You wanna discuss issues with the frindge ends of ladder? That's fine. Set a reasonable assumption value for outliers that's a hard cap. At max some one is treated as say 2600 for what they can que with. Even if they are 2900.... treat them as 2600. Do the same for the bottom end, issue solved for the 0.001% (<- this is a made up number. its honestly a small amount thou guys. Top of challenger and bottom of iron.) of folks that would impact.

KnKitsuneJul 10, 20191 votes

If you are winning 30 LP and losing 12, I really envy you. Used to get and lose around 20LP, but after countless loss streaks over things completely out of my control, I'm stuck getting 10-12 LP and losing at least 20.

To me, it feels like trolls/inters/griefers have become too common this season. On my team and the enemy's.

Kai GuyJul 10, 20191 votes

Some day I hope I start to see widespread support from boards users when I go on a rant about the massive impact player behavior systems make on quality of gameplay.

DarthArchonJul 10, 20191 votes

[{quoted}](name=Kai Guy,realm=NA,application-id=3ErqAdtq,discussion-id=xcv4Qrev,comment-id=00010000,timestamp=2019-07-10T09:16:30.636+0000)

Some day I hope I start to see widespread support from boards users when I go on a rant about the massive impact player behavior systems make on quality of gameplay.

What do you mean? I know that having seen other post of this kind saying that winning to much in a row and you start getting trolls and inter and it feel true to me.

I guess that's a problem with algorithms that even the programmers don't understand.

SubdueJul 10, 20191 votes

{quoted}

What do you mean? I know that having seen other post of this kind saying that winning to much in a row and you start getting trolls and inter and it feel true to me.

I guess that's a problem with algorithms that even the programmers don't understand.

I think Kai explained it extensively, but it's actually pretty easy to understand in concept, if you understand how it works.

The players in your games generally have similar MMRs.
Players with higher MMR are generally better than players with lower MMR.
As you win, your MMR goes up, and as you lose, your MMR goes down.
Because of #2 and #3 above, as you win, the players you are playing with and against are better and better.
Because of #4 above, your overall impact on the game decreases. Imagine you're playing basketball with 9 3rd graders. Now imagine you're playing basketball with 9 18 year olds.
Because of #4 and #5 above, eventually the players you're playing with and against are as good or better than you are. Then, your ability to win is equal to or even less than your opponents, and you start to lose.

The net effect of this is that players will rubberband between the MMR where they are so much better than the players in the game that they win most of their games and the MMR where they are so much worse than the players in the game that they lose most of their games. The range between these two values depends largely on the player, their playstyle, their temperment and so on. A consistent player has a much smaller range, while a more feast or famine player might have a larger one.

DarthArchonJul 10, 20191 votes

{quoted}

I think Kai explained it extensively, but it's actually pretty easy to understand in concept, if you understand how it works.

The players in your games generally have similar MMRs.

Players with higher MMR are generally better than players with lower MMR.

As you win, your MMR goes up, and as you lose, your MMR goes down.

Because of #2 and #3 above, as you win, the players you are playing with and against are better and better.

Because of #4 above, your overall impact on the game decreases. Imagine you're playing basketball with 9 3rd graders. Now imagine you're playing basketball with 9 18 year olds.

Because of #4 and #5 above, eventually the players you're playing with and against are as good or better than you are. Then, your ability to win is equal to or even less than your opponents, and you start to lose.

The net effect of this is that players will rubberband between the MMR where they are so much better than the players in the game that they win most of their games and the MMR where they are so much worse than the players in the game that they lose most of their games. The range between these two values depends largely on the player, their playstyle, their temperment and so on. A consistent player has a much smaller range, while a more feast or famine player might have a larger one.

Right my point is that i know i'm plat lvl since i've been there every season i played enought to be there, yet when i ranked and get placed silver 2, it can take 2-3 weeks just to get out of silver, most of the time because promos, specially silver 1 is filled with bad player. It shouldn't take me 3 month to go back when i know i'm plat. I don't play lol all that much and don

SubdueJul 10, 20191 votes

{quoted}

Right my point is that i know i'm plat lvl since i've been there every season i played enought to be there, yet when i ranked and get placed silver 2, it can take 2-3 weeks just to get out of silver, most of the time because promos, specially silver 1 is filled with bad player. It shouldn't take me 3 month to go back when i know i'm plat. I don't play lol all that much and don

The ranked season is a year long. If it takes 2-3 weeks to climb 2 divisions, and you belong in plat and are starting in silver 2, that's about half a season to get back to plat, which doesn't seem excessive to me.

That being said, Op.gg says you haven't ended a season higher than Gold 4 so...

Kai GuyJul 10, 20191 votes

Also gonna chime in and point out the issue how riot freely gives titles instantly because its a feel good system for folks to see quick gratification. For most the ladder bottom end Demotions take a MASSIVE mmr loss so its hard to fall out of a Tier. 0 MM impact outside of duoing which riot does weakly (for the same player perception reasons they do Demotion protection so dramatically for divisions.)

Personally if I wanted to have a system that instilled confidence in players Id require it to have a minimum # of games that it considers to achieve workable accuracy for placement and no title or division would be provided to players. Thing is people are idiots who don't like or want to understand math and would rather see quick rewards. So Riots system is filled with player perception altering systems that don't do much besides make people feel better. Adding. Iron for example is likely a 0 MMR impact attempt to try and manipulate perceptions to get players to make more effort for the mid levels of the ladder. Redistributing the ladder and using soft resets doing this made some really NASTY issues this season.

Id just let idiots sink and lose rank if they start playing halfassed and kick them from ranked ques if they actively look to throw games to spite teammates /other toxic behaviors. No demotion protection for an entire MMR division that's likely used to set class intervals in riots system. No Gold 4 sitting around at Silver 3 mmr. that dudes instantly silver again the moment he drops out of my probability models range of expectations.

Probaly a good thing for most folks that I am not a dev and I doubt id be popular with players owing to how blunt I approach things.

Inari Fox OrrionJul 10, 20191 votes

[{quoted}](name=Kai Guy,realm=NA,application-id=3ErqAdtq,discussion-id=xcv4Qrev,comment-id=0001000000000000000000000000,timestamp=2019-07-10T18:39:52.743+0000)

Also gonna chime in and point out the issue how riot freely gives titles instantly because its a feel good system for folks to see quick gratification. For most the ladder bottom end Demotions take a MASSIVE mmr loss so its hard to fall out of a Tier. 0 MM impact outside of duoing which riot does weakly (for the same player perception reasons they do Demotion protection so dramatically for divisions.)

Personally if I wanted to have a system that instilled confidence in players Id require it to have a minimum # of games that it considers to achieve workable accuracy for placement and no title or division would be provided to players. Thing is people are idiots who don't like or want to understand math and would rather see quick rewards. So Riots system is filled with player perception altering systems that don't do much besides make people feel better. Adding. Iron for example is likely a 0 MMR impact attempt to try and manipulate perceptions to get players to make more effort for the mid levels of the ladder. Redistributing the ladder and using soft resets doing this made some really NASTY issues this season.

Id just let idiots sink and lose rank if they start playing halfassed and kick them from ranked ques if they actively look to throw games to spite teammates /other toxic behaviors. No demotion protection for an entire MMR division that's likely used to set class intervals in riots system. No Gold 4 sitting around at Silver 3 mmr. that dudes instantly silver again the moment he drops out of my probability models range of expectations.

Probaly a good thing for most folks that I am not a dev and I doubt id be popular with players owing to how blunt I approach things.

Consider the effect it'd have on players in a low-impact role and/or the fact that most people play to rank up with a friend or two. It wouldn't be popular because it wouldn't be fun, not because it wouldn't be stroking the ego or something.

Their current system is a weird, ineffectual version of a mid-variance system where you're allowed to queue up with a high-MMR player within the same MMR bracket, but your combined MMR will be higher and therfor you'll need to work your ass off to win. The problem is... it uses ranks and divisions, and not MMR for this system. Your friend can have Diamond 2 MMR with a Silver ranking and you can be queued up with him up against Plat 1's and you'll get dumpstered+carried. An alternate system would be to individually take MMR and match you up against people with similar MMR values, 1:1. Your friend is Platinum-MMR? You'll get a Platimum-MMR opponent. Too bad both systems have issues thanks to auto-fill, trolls, ragers, AFKers, etc.

I have to ask, Kai, how would you handle outliers like me whom have Gold-tier prowess but are only ever able to get out of Bronze and Silver with the help of someone they can consistently rely on to follow up on their setups, to use the vision they provide, and to take the aid offered? "Oh well, as an outlier, you don't matter" is about as good as the current system because we are all outliers within it.

Kai GuyJul 11, 20191 votes

The problem is... it uses ranks and divisions, and not MMR for this system.

^ correct. Duoing is something you will see me complain about. I strongly dislike Riots implementation.

Your friend can have Diamond 2 MMR with a Silver ranking and you can be queued up with him up against Plat 1's Incorrect. The system plays catchup. You will not just skip divisions but you skip promos entirely when your MMR to title gap gets to larger. I have tried to abuse it before by dodging promo games in the lobby to keep my rank low. 1. Because I thought it would be funny to have a really low rank compared to the games I get put into to freak teammates and enemies out. 2. Because I abuse duoing to show my friends what its like to play vs higher MMR. A pretty good way to get them to improve. The larger the average skill gap the more dramatic it is when they are outclassed. Also a good way to check how much personal impact I can make on a champion.

I have to ask, Kai, how would you handle outliers like me whom have Gold-tier prowess but are only ever able to get out of Bronze and Silver with the help of someone they can consistently rely on to follow up on their setups, to use the vision they provide, and to take the aid offered? "Oh well, as an outlier, you don't matter"

Your not an outlier. You have literally the same tools available to you as any player. The "have to be a hard carry to climb" Is a BS argument.

Here is a Q and A from Trueskill for ya. Q: You are saying that the TrueSkill ranking system assumes that the skill of a team is the sum of the skills of its players. I think this model is not appropriate: I am usually playing much better with people from my friends list rather than with random players. Will this assumption lead to incorrect rankings? A: The assumption that the team skill is the sum of the skills of its players is exactly that: an assumption. The TrueSkill ranking system will use the assumption to adopt the skill points of individual players such that the team outcome can be best predicted based on the additive assumptions of the skills. Provided that you and your friends also play team games with other players now and then, the TrueSkill ranking system will assign you a skill belief that is somewhere between the skill when you are playing with your friends and the skill when you are playing as an individual. So, in the worst case, every other game is not with your friends: then you are slightly ranked too high when you play with random team players and slightly ranked too low when you play with your friends. But, if you mostly play with your friends only the system will identify your skill correctly for most of your games.

Again.... your not an outlier. You have the exact same access to victory as any other given player.

"is about as good as the current system because we are all outliers within it."

By definition you don't have everything in a system as "outliers" so your misusing the concept if you claim everybody is one.

Imo. Keep playing with your friend in ranked if you want to climb. It works for you. You have good synergy. Some times 1 + 1 = 5 so to speak. Playing solo you don't seem to be able to make it. Nothing wrong with that. There's no shame in being lower MMR or in the zone representative of a majority of players. You can adjust your playstyle or don't. Look its your call but your not an outlier mate.

Kai GuyJul 10, 20191 votes

https://www.microsoft.com/en-us/research/project/trueskill-ranking-system/

https://www.microsoft.com/en-us/research/publication/trueskilltm-a-bayesian-skill-rating-system/

https://en.wikipedia.org/wiki/Elo_rating_system

https://books.google.com/books/about/The_Rating_of_Chess_Players_Past_and_Pre.html?id=syjcPQAACAAJ

I guess that's a problem with algorithms that even the programmers don't understand.

No... its understood quite well. There are well documented and have been peer reviewed for years as well as a massive amount of data that's generated by the Trillions of games played. Riots system CAN BE DONE BETTER! But its not broken and any one telling you it is full of shit.

Players are constantly wrong and don't even take a second to research the topic before spewing BS random guesses all over the boards to justify their raiting not matching their personal self expectation. Most players consider their "skill level" to only be their best games. In form, when they play well. Its commonly assumed as the mean average for their top end performances, when a MMR system is the mean average for ALL performances.

Toxicicty is a issue for a competitive ladder, behaviors that throw losses into the game will in MMR just impact that accounts average performance and lower them on the ladder. Player behavior systems are what get used to remove a player from the que, bans and so on. MMR is objective and looks at the result of a match. Player systems will look at the individual details of a match, chat logs, if some ones running it down mid, if some ones refusing to play as a teammate to win, some one rage quit or afked, sold off all their items. ETC. Many times you need a subjective observation to really track down an issue case as automation for subjective judgment is far more complex then a MMR system.

Players consistently consider 3 games a streak. Iits biased and terrible logic. https://www.random.org/coins/?num=100&cur=60-usd.0025c-ct Flip 100 coins. Your legendary Fair "50% odds". Gee golly gosh monkeys! a streak starting at 3 for heads or tails shows up ALL THE TIME. But a streak of 15? Almost never see that.

Lets look at some Binary results of W or L. So 3 for a streak is just a weak number being something like 2^3 = 8. That's what 7 to 1 odds to see if your looking at literally just 3 games? There 1 possible outcome out of all 8 possible combinations you can make that's gonna be 3 wins. This gets exponetionally smaller as you up the # of possible results, for example 2^4, one more game? For a 3 streak would be what. 16 ways to make combinations. WWWW.WWWl. LWWW. So... look for a streak of 3 in a set of 4 games and you end up with... i think 4.333 to 1 odds?

I am bad with factorials and struggle with statistics so I may have provided incorrect Odds.

and MMR systems don't care about order? WWlW is the same as WWWL unless you hit a K factor slowdown. +5+5+5-5= 10. -5+5+5+5=10. +5+5-5+5 = 10. Its basic addition for your Rating point changes, Yea WE and K impact it but the rules are the same for everbody. 100 random players with a raiting of 1200 with K at X value vs 100 1400 with K at Y will generate the same result every single time for the adjustment, the math has rules it obeys the same way every time.

Also no shit you see a streak end when you have a awfull teammate. Your more likely to lose when you have crappy teammates. Should be self evident? Like OMG we by far where the better team and stomped them super hard. Thats why we lost? Vs. Some one sucked it up, behaved inmautrly and I had a crappy game because of it. That's why we lost.

No idea why players think, oh my team of 5 and their team of 5... my team better be ALL better players then the other 5. That's fair? A reasonable expectation. 1-2 folks on your team are gonna be outclassed. when its 3-4 its usually GG.

And again back to my comment about importance of player behavior systems. Look at what a 10% player population looks like in contect of 5v5 games. 10 slots. your 1. 9 open slots. Lets measure 10% toxicity. If only 1 in 10 players is a negtivive teammate. https://www.wizards.com/dnd/dice/dice.htm Go to D 10. Set it to be 9 dice, this represents all the other players in a game. Pick 1 number. I am going to use 1 for mine. 1 has 10% odds of RNG being rolled on a D10.

Roll(9d10)+0: 1,5,6,4,2,1,8,4,2,+0 Total:33

Roll(9d10)+0: 8,6,10,2,6,3,8,9,4,+0 Total:56

Roll(9d10)+0: 9,10,4,4,10,10,8,1,10,+0 Total:66

Roll(9d10)+0: 4,9,5,6,1,10,6,10,8,+0 Total:59

Roll(9d10)+0: 9,7,1,5,1,2,8,6,5,+0 Total:44

4 out of 5 games had toxic players at a rate as low as 10%.

Many players would think oh 1 in 10 games might see a troll. when... well 10% is a massive number your gonna find in your games almost always. So... player behavior needs to be a constant focus and consistently negative players have to be removed for competitive integrity.

Also Darth. Nothing personal. I have no beef with you. Please take any harsh language to never be directed personally at you, I'm just venting a bit.

I just like numbers and dislike how fucking difficult it is to try and convince players to care about the behavior systems riot uses because MMR is a proven system. Just don't get why its so difficult for people to grasp that in team PvP games you can bet that The P impacts gameplay. Game can be good but if P is filled with garbage the gameplay is... garbage. Its really simple. So... if you care about the game being better. Behavior systems for Ranked Need a rehaul more then anything else. I hope managed to convince you on this.

LindurosJul 10, 20191 votes

"It should be ok and conceivable to ban people from ranked because they played poorly."

So are you suggesting that Riot just start banning the worst 10% of players from ranked? Or do you think just 1% would be more reasonable? Or would it be more reasonable still to demote players that play poorly to lower ranks?

Let's look at this from a more macro "big picture" perspective. If you play 100 games, on average you will have comparatively bad players on your team and on the enemy team roughly the same amount of times. Your rank will be determined by how frequently your team comes out on top over a large sample of games, not just a handful of games. If you are able to carry "bad players" consistently then you will probably rise up the ranks while those lesser skilled players will drop. But if "bad players" consistently tilt you and make you play lower than your ability and flame in chat, then you will stay stuck at the rank you deserve. If Riot just gets rid of all the "bad players" from ranked, then there will eventually only be one division called "Challenger" and the rest of us will be playing other games with you.

Competitive games by nature are not really ever going to be consistently "easy" to win. Most of your original post just demonstrates that you don't understand the "competitive" nature of player versus player games. Even if you have a high win-rate, like 60% that still means you are going to lose 40 of every 100 games you play. You just have to learn to expect to lose sometimes while still maintaining a positive mental and move on to the next game. This is really a question of your own emotional management and ability to keep your cool while on a losing streak, knowing that you will come out of it and get some easy wins again on the other side.

If we all stopped looking at the chat window and instead looked at the mini map, the quality of everyone's games would be improved. So flaming in difficult games is only going to make things worse because you're taking eyes off the mini map to see your toxic chat.

DarthArchonJul 10, 20191 votes

This forum as issue with people talking pass each other

Competitive games by nature are not really ever going to be consistently "easy" to win. Most of your original post just demonstrates that you don't understand the "competitive" nature of player versus player games. Even if you have a high win-rate, like 60% that still means you are going to lose 40 of every 100 games you play. You just have to learn to expect to lose sometimes while still maintaining a positive mental and move on to the next game. This is really a question of your own emotional management and ability to keep your cool while on a losing streak, knowing that you will come out of it and get some easy wins again on the other side.

I don't see how it relate to what i'm saying? and obviously nobody thinks like that, making you look condescending a little...

The fact is if you can arbitrarily report someone for intentional feeding, why wouldn't be able to report someone for playing is role badly on purpose. Like splitpushing no tp with a slow shove champ who can't 1v1 and sticking to it.

My main argument is that the matchmaking system could offer more choice of playstyle and option to help player play with other who play the same way as them, there's alot of room to improve but clearly riot have other priority, mainly nuturing those who pay for skins (young boys and girls)

And keep in mind that this game is going for it's tenth year of existence soon and still almost nothing change about matchmaking. Role matchmaking took 4 years to implement, when other more recent moba did this months after release.