The Smurf Problem - MMR System is broken

[Wazowski_]

This is not a rant its a serious problem.

(Btw thanks for the responses guys, I've edited the main post to be more clearer and reflect some of the idea's people have shared.)

Here's the problem with Smurfing.

The Dota 2 match maker tries to make a fair contest of similar skill levels by creating 2 teams that have reasonably balanced MMR scores.

Smurfing cheats the match maker into thinking the player is worse then they actually are. So when the matchmaker makes 2 teams that it believes is fair its actually not, but the loosing team without the smurf is still treated as if they lost a fair balanced team fight (-25 MMR).


Taking advantage of a bug or oversight in a game for personal gain to the detriment of other players is cheating.


This is a serious problem valve.

It needs to be fixed.


People say that Valve doesn't care about the players, they only are about money.

Alright Valve, lets talk about money.

When a team looses to a smurf, they experience a differing range of emotions. Anger at the smurf, dissolution with the game, Anger at Valve etc. All of which don't make people want to keep playing Dota 2 or spend money on Dota Plus or spend money on merchandise from the store. You're letting these cheating smurfs drive people away from spending money on your game. Even if the developers don't give a single damn about players, they are obviously as thick as hell if they are ok with cheaters that are taking away their business. I mean they ban people that get their account boosted. Why not do something about the smurfs who hurt the game, the community and their back pocket a lot more.

Proposed Solution:

Dota 2 is smart enough already to determine a players skill level over the last 20 games (the hexagon stat chart on a players profile) as well as the skill level every game for a hero verses that players average. Valve clearly has the data to determine a players skill level outside of MMR. Why not implement smurf detection that detects when a players average skill level is well above the normal average for his MMR, then force the offending account to do the recalibration games again starting at an MMR bracket that matches their skill level?

For example. In the 1k pool, the average skill level on the players hexagonal chart might be about 4.5-5.5. Where as, when a 5k player, makes a smurf account their skill level on that chart is always 9-10. Why not have a system in place that detects this huge difference in averages and force the offending account to do the recalibration games again but starting at a MMR bracket that matches his skill level (in this case, his own 5k bracket)


Use the same system to prevent boosters as well? Use the system to detect boosters, or even just when the booster gives the account back to the player? Lets say a 2k player gets his account boosted to 5k. While his MMR is going to say 5k, his personal skill stats are going to go straight back to what they were before he got it boosted. Valve could use this to detect the fact that he belongs in 2k not 5k and force the account to re calibrate.


Maybe Valve needs to revalute how MMR is given? Maybe your personal stats should affect how much MMR you gain from a winning fight? For example you play well, get good personal stats for the game so you gain more MMR then the guy who was on your team but played really poorly? Or you loose a game but your stats reflect that you played really well so you don't loose as much MMR as the guy who fed 0-8 in the first 10 min?


I have had so many friends that are new to Dota 2 quit the game because of very experienced players, cheating the system because they need an ego boost from obliterating the new guys

Here's a great explaination of what happens when Valve lets players create smurf accounts:

"Smurfing is the equivilant of bragging you beat darksouls on the easiest mode. Sure you beat a game but you did it on handholding baby mode, theres no challenge. but many of these smurfs get a thrill from this.

But throwing all that into a PvP environment? it cripples games when they can engage "easy mode" vs other players, while its easy mode for them it forces nightmare mode on everyone on the other team. Its not fun or fair" ~ Credit Chili Dog Dave


Dota 2 already has the data. It shows us our performance over the last 20 games as well as every game based on our average with that hero. The data is there but its currently only used for personal comparison. Why not use it to prevent players attempting to cheat the MMR system by smurfing or boosting? I could write the code for this myself. Why has no one done this yet?

Even if this isn't the solution there has to be a solution to this problem. Its destroying the game for 1000's of people.


Valve your game is broken. Fix it.

[Leon]

Wazowski_ a écrit :

Dante a écrit :

Lol. dude its not “rare” just so you know last year alone valve ban hundred thousand account for boosting (announce in dota2 twitter). If your system implemented, i cant imagine how many people will use it to boost their account. Next time before you demand something do some research first and then have a viable solution. I’ve been in this forum for years and this kind of solution been repeated over and over again in other threads. So its not new to me.

My apologies i didn't realise you were referring to boosting. I agree boosting should be banned.

Wouldn't a system that CONTINUALLY compared a players skill level to the average in their MMR bracket prevent boosting as well? So lets say a 2k player gets his account boosted to 5k, but then when he starts playing the account again, the game recognises he's actually noob as hell and then forces him re calibrate (do the 10 solo and 10 party calibration games at 2k instead of 5k) at the MMR bracket that matches his skill level (2k).

Hahaha there is no system can recognize people is noob as hell or having bad day like my post above. Thats why human have feelings but computer doesnt.

[Wazowski_]

Dante a écrit :

Wazowski_ a écrit :

Don't you think its unfair if a rich noob buys his way to 5k with boosters then becomes a huge burden to his 5k team mates because he's noob as hell? Shoudn't he be put in his own bracket to learn at his own pace with players his own level?

The answer is not fair but the system cant determined wether people is that buy 5k account and become a burden or he just have a bad game but he is a 5k player. Again like i said its not that simple. Many equation need to be considered. If the player having lag or not, the player is currently not in a right mind. Not that simple bro. Its not that easy.

I agree, thats where the current MMR system works well. You have a bad day/game, you loose 25 MMR, you can have a better day/game tomorrow and win it back.

What i'm suggesting though is a way to prevent extreme cases of boosting and smurfing (more then +/- 1-2K)

[Leon]

Wazowski_ a écrit :

Dante a écrit :

The answer is not fair but the system cant determined wether people is that buy 5k account and become a burden or he just have a bad game but he is a 5k player. Again like i said its not that simple. Many equation need to be considered. If the player having lag or not, the player is currently not in a right mind. Not that simple bro. Its not that easy.

I agree, thats where the current MMR system works well. You have a bad day/game, you loose 25 MMR, you can have a better day/game tomorrow and win it back.

What i'm suggesting though is a way to prevent extreme cases of boosting and smurfing (more then +/- 1-2K)

About smurfing is not bannable offense doesnt matter how you put it. But boosting is bannable cause many ID has been banned last year on ban wave by valve.

[Wazowski_]

Dante a écrit :

Wazowski_ a écrit :

My apologies i didn't realise you were referring to boosting. I agree boosting should be banned.

Wouldn't a system that CONTINUALLY compared a players skill level to the average in their MMR bracket prevent boosting as well? So lets say a 2k player gets his account boosted to 5k, but then when he starts playing the account again, the game recognises he's actually noob as hell and then forces him re calibrate (do the 10 solo and 10 party calibration games at 2k instead of 5k) at the MMR bracket that matches his skill level (2k).

Hahaha there is no system can recognize people is noob as hell or having bad day like my post above. Thats why human have feelings but computer doesnt.

Bro honestly hahaha. Listen. In my MMR rank, even just looking at the last 20 games, the average players hexagonal stat chart is averaged at 4.5-5.5 out 10. But when you get a smurf, he's like 9.5-10 out of 10. Every single time. What i'm saying is that there should be a system where if a players average skill score is so far above the other players in his rank, he should be forced to do the recalibration games with players his own skill level.

Like wise if a player boosts his account, but then their average skill level drops to like 2.5-3.5 (over enough games that constitutes an actual average, not a bad few days) , they should be forced to recalibrate at their correct level

[Wazowski_]

JouYizz a écrit :

Wazowski_ a écrit :

I totally agree, but if thats the case, why did valve even create a ranking system in the first place? Would you put Miracle meepo mid against your 2k buddy you is learning how to play SF still? Its not a place to learn, you just get the floor wiped with your but. People are fed of smurfs in the lower brackets

the ranking system isnt actually that accurate as it can be exploited.. you experienced it firsthand.. nonetheless its there to put a semblance of skill bracketing... personally ive faced with a lot of smurfers and its annoying AF even to to point of me raging but hey its one of the reasons why i got out of the 2k bracket

Yeah i totally get that. The only problem is, I am a 2k player lol. That is my skill level. I've only been playing dota for 10months now, put 800 games into it and I'm starting to get my head around it, but i am a 2k player i can't get out yet hahahaha

[Leon]

Wazowski_ a écrit :

Dante a écrit :

Hahaha there is no system can recognize people is noob as hell or having bad day like my post above. Thats why human have feelings but computer doesnt.

Bro honestly hahaha. Listen. In my MMR rank, even just looking at the last 20 games, the average players hexagonal stat chart is averaged at 4.5-5.5 out 10. But when you get a smurf, he's like 9.5-10 out of 10. Every single time. What i'm saying is that there should be a system where if a players average skill score is so far above the other players in his rank, he should be forced to do the recalibration games with players his own skill level.

Like wise if a player boosts his account, but then their average skill level drops to like 2.5-3.5 (over enough games that constitutes an actual average, not a bad few days) , they should be forced to recalibrate at their correct level

Doesnt matter man, trust me there is hundred even thousand thread like this that i’ve been read over and over again. Its just plain simple your suggestion is not and probably never will implemented. And its not just your experience to meet smurf. Every single dota player experience it. Even me, but i still can manage to climb, my first mmr under 1k and i play this game from early 2017. Now im in 4k, so if noob like me can climb i’m pretty sure everybody can.

[∞ＰＡＲＡＤＯＸ∞]

We shouldn't base on hexagonal skill chart actually because there are versatile players that different roles in many games. He could have 4 points of everything but plays like a pro. If a player focused carry/mid in 20 games, it is expected to have high points on farming, pushing, and fighting. If he focused on supporting in 20 games, then, supporting is high. The point is, we shouldn't base on the hexagonal skills chart. I have a solution in smurfing but it is complicated and all small brains would just call it excessive effort or too strict or won't even try to understand at all.

[Chilli Dog Dave]

smurfs ARE an issue even if they tell you "I'm LeArNiNg nEw HeRoEs!"
my solutions are to just increasing how long it takes to get into ranked and having a different MMR for different roles (also make role queue base game not some underused crap from DOTA+ ) would go quite the way to slowing new smurfs down, if the work to smurf is far to much compaired to how long they'll be stomping games, they wouldn't make as many.
then making role queue base game would limit smurfs even more as games will take longer if they dont exploit the system, if they do they can be reported for not doing the role they signed up for.
then we need to adjust how stats themselves impact your rank, if your end game stats are far above your brackets average, they should gain x10 to x20 times the mmr from that win, why? because it'll shove them back where they belong faster, nullifying the smurfs ability to stomp for long periods of time.
if they truely are learning a new hero? they wont be stomping every game as hard so they'll know where their "new hero" placement should be.

edit: made this reply on reddit where the devs actually read.
https://www.reddit.com/r/DotA2/comments/cgp60d/just_some_random_suggestion_regarding_smurfs_and/
i recommend supporting it if you can so the pro smurf crowd dont bury it.

[Wazowski_]

Synchronicity ♦ a écrit :

We shouldn't base on hexagonal skill chart actually because there are versatile players that different roles in many games. He could have 4 points of everything but plays like a pro. If a player focused carry/mid in 20 games, it is expected to have high points on farming, pushing, and fighting. If he focused on supporting in 20 games, then, supporting is high. The point is, we shouldn't base on the hexagonal skills chart. I have a solution in smurfing but it is complicated and all small brains would just call it excessive effort or too strict or won't even try to understand at all.

Bro i understand what you're saying. I guess what i'm trying to say is Dota 2 already has a system in place for determining a players skill level at farming, supporting, fighting, etc... When ever i play a smurf, thats really smurfing (like 2k or more below his level) His averages on that chart are always double everyone else in the bracket.

I personally can handle smurfs that are 1k lower then they should be. Forces you to play harder, but 2-3k lower then they should be is just outrageous. It makes the game unplayable

[Wazowski_]

Synchronicity ♦ a écrit :

We shouldn't base on hexagonal skill chart actually because there are versatile players that different roles in many games. He could have 4 points of everything but plays like a pro. If a player focused carry/mid in 20 games, it is expected to have high points on farming, pushing, and fighting. If he focused on supporting in 20 games, then, supporting is high. The point is, we shouldn't base on the hexagonal skills chart. I have a solution in smurfing but it is complicated and all small brains would just call it excessive effort or too strict or won't even try to understand at all.

Bro whats your ideas to combat the smurfing/boosting? Why not actually make this a genuine thread with actionable ideas?

[Hornbach Loungeset]

The Day Is My Enemy a écrit :

this is a great longread, but the modern neural nets are capable of detecting cheaters so far, not of providing the fictional technology that you are imagining.

♥♥♥♥♥♥♥ spotted lmao

[Hornbach Loungeset]

Disconnected a écrit :

The Day Is My Enemy a écrit :

this is a great longread, but the modern neural nets are capable of detecting cheaters so far, not of providing the fictional technology that you are imagining.

tard spotted lmao

[Chilli Dog Dave]

Chilli Dog Dave a écrit :

smurfs ARE an issue even if they tell you "I'm LeArNiNg nEw HeRoEs!"
my solutions are to just increasing how long it takes to get into ranked and having a different MMR for different roles (also make role queue base game not some underused crap from DOTA+ ) would go quite the way to slowing new smurfs down, if the work to smurf is far to much compaired to how long they'll be stomping games, they wouldn't make as many.
then making role queue base game would limit smurfs even more as games will take longer if they dont exploit the system, if they do they can be reported for not doing the role they signed up for.
then we need to adjust how stats themselves impact your rank, if your end game stats are far above your brackets average, they should gain x10 to x20 times the mmr from that win, why? because it'll shove them back where they belong faster, nullifying the smurfs ability to stomp for long periods of time.
if they truely are learning a new hero? they wont be stomping every game as hard so they'll know where their "new hero" placement should be.

edit: made this reply on reddit where the devs actually read.
https://www.reddit.com/r/DotA2/comments/cgp60d/just_some_random_suggestion_regarding_smurfs_and/
i recommend supporting it if you can so the pro smurf crowd dont bury it.

pro smurfing players are already saying this idea are bad and trying to downvote bury the suggestion, even when i'v stated how toxically bad the current system is for anyone but the smurf. could use the assistance folks

[Leon]

Chilli Dog Dave a écrit :

Chilli Dog Dave a écrit :

smurfs ARE an issue even if they tell you "I'm LeArNiNg nEw HeRoEs!"
my solutions are to just increasing how long it takes to get into ranked and having a different MMR for different roles (also make role queue base game not some underused crap from DOTA+ ) would go quite the way to slowing new smurfs down, if the work to smurf is far to much compaired to how long they'll be stomping games, they wouldn't make as many.
then making role queue base game would limit smurfs even more as games will take longer if they dont exploit the system, if they do they can be reported for not doing the role they signed up for.
then we need to adjust how stats themselves impact your rank, if your end game stats are far above your brackets average, they should gain x10 to x20 times the mmr from that win, why? because it'll shove them back where they belong faster, nullifying the smurfs ability to stomp for long periods of time.
if they truely are learning a new hero? they wont be stomping every game as hard so they'll know where their "new hero" placement should be.

edit: made this reply on reddit where the devs actually read.
https://www.reddit.com/r/DotA2/comments/cgp60d/just_some_random_suggestion_regarding_smurfs_and/
i recommend supporting it if you can so the pro smurf crowd dont bury it.

pro smurfing players are already saying this idea are bad and trying to downvote bury the suggestion, even when i'v stated how toxically bad the current system is for anyone but the smurf. could use the assistance folks

Don’t bump your own comment. Just edit your own comment if you need something added.

[Wazowski_]

Chilli Dog Dave a écrit :

smurfs ARE an issue even if they tell you "I'm LeArNiNg nEw HeRoEs!"
my solutions are to just increasing how long it takes to get into ranked and having a different MMR for different roles (also make role queue base game not some underused crap from DOTA+ ) would go quite the way to slowing new smurfs down, if the work to smurf is far to much compaired to how long they'll be stomping games, they wouldn't make as many.
then making role queue base game would limit smurfs even more as games will take longer if they dont exploit the system, if they do they can be reported for not doing the role they signed up for.
then we need to adjust how stats themselves impact your rank, if your end game stats are far above your brackets average, they should gain x10 to x20 times the mmr from that win, why? because it'll shove them back where they belong faster, nullifying the smurfs ability to stomp for long periods of time.
if they truely are learning a new hero? they wont be stomping every game as hard so they'll know where their "new hero" placement should be.

Hey bro thanks for contributing.

A couple of thoughts to add to your ideas:

Chilli Dog Dave a écrit :

increasing how long it takes to get into ranked

This would definitely stem the tide of smurfs fairly well, however from Valves perspective, you have to look at what genuine new players are facing too. For Valve, they have to make the number of unranked calibration games high enough to deter smurfs, but also low enough not to deter new players from staying and playing Dota 2. The current 100 unranked calibration games is probably the sweet spot to be honest. Enough to deter some smurfs but not to much to deter new players from sticking with the game.

Chilli Dog Dave a écrit :

different MMR for different roles

I think this is not a bad idea. This might get pretty complicated for the noobs though. I think the ranked roles should definitely be a more common thing for serious players. I gues Valve needs to decide how they want to make money of Dota 2. Makes sense that the more serious players would be willing to fork out some money to play real games instead of trash ones.

Chilli Dog Dave a écrit :

adjust how stats themselves impact your rank, if your end game stats are far above your brackets average, they should gain x10 to x20 times the mmr from that win, why? because it'll shove them back where they belong faster, nullifying the smurfs ability to stomp for long periods of time.

I think this is a pretty good idea myself. Maybe not 10x the gain, but the concept is a good one. It would be hard to implement though i think like how do you compare the carries contribution to the team over the supports?. It would work well in a ranked roles situation i think because then you could compare your performance against other players in your role at your MMR level not just the specific game. Both things could come into play in determining how much MMR you get.

For example, you're the position 5 support, you win the game, but your average support skills verses the average for your MMR level were quite low so you only gain a little MMR. But lets say your the position 5 support, you win but your, but your support skill level for that match was higher then the average for your MMR level so you gain the 25 from winning plus extra based on how much better your support skill was.