With my highschool knowledge of statistics, I tried to figure out my past 36 ranked games

Amor Omnia ·Apr 30, 2019·12 votes·7,711 views

Out of 38 games, 13 games were 1 sided
Out of 38 games, 20 games had smurfy cs performance (cs per minute above 7)
There was only 1 game where my team had more enemy towers taken down and still lost
There were 2 remakes

I played 578 games so far... Is it safe to say that the outcome of my lower silver elo games are not affected by smurfs? That's what im getting...

I guess i'll go back to play more league lel I wish there weren't 2 remake games. Time is wasting...

13 Comments

MarshbouyMay 1, 20194 votes

I wouldn't call CS/min above 7 "smurfy"

I've easily gotten 10 cs/min in my games on ADC, but we all know i'm dogshit at the game lmao.

Amor Omnia May 1, 20191 votes

I couldn't think of any other measure for skill discrepancy in mechanics...

Meteora707May 1, 20191 votes

just check if the players have over 70% wr

GripaAviaraMay 1, 20191 votes

That's a booster not a smurf

Meteora707May 2, 20191 votes

everyone who has been decent mmr will have 70%+ wr in silver

MarshbouyMay 2, 20191 votes

unless they're intentionally losing, playing mindlessly, or playing super off meta shit and messing around.

Meteora707May 2, 20191 votes

even so, in my experience playing weird shit is actually really good in silver cause they have no idea how to deal with it.

BârdMay 1, 20191 votes

Early game, sure.

What you look out for is whether or not they've managed to maintain that CS rate for the rest of the game.

You don't tend to see people with 400 CS at 40 minutes in Silver, for example.

MarshbouyMay 2, 20191 votes

Again, I've managed that before.

Some people in silver will spend a whole game mindlessly CSing (e.g. me). That doesn't mean they're good though.

edit: normally, I find it easy to catch up in CS later in the game because, while my team is playing ARAM, I can catch side waves and clear camps.

DJBlackVoiceMay 1, 20192 votes

Dont think a simple linear regression will help you here. Maybe try a classification tree or a neural net, think it would work better. Also you need more games (bigger population) to try to have some good analisys.

Sorry for bad english, hope you understand what i mean.

HateDaddyMay 1, 20192 votes

{quoted}

Dont think a simple linear regression will help you here. Maybe try a classification tree or a neural net, think it would work better. Also you need more games (bigger population) to try to have some good analisys.

Sorry for bad english, hope you understand what i mean.

Your English was fine!

That said, the issue really isn’t the type of analysis he conducted, it’s what he was measuring that was problematic - and problematic isn’t a great word to define this because there’s nothing really wrong with what he ran, it’s just poorly defined.

If you’re trying to determine if smurfs are in your game and you’re basing it on the relationship between CS/tower differential correlations, you’re really only explaining what victory looks like. Even if CS per minute is unusually high, that doesn’t suggest its a smurf. Players can CS moderately well in silver, especially if they get ahead. And what you have is called a positive correlation, or a positive linear relationship. And that should be expected - if you have larger CS/minute (even if it’s higher than normal), then you’re going to see that relationship lead to greater tower differences too - in other words, you shouldn’t see a negative correlation here where cs difference on average and towers taken diverge. And because regressions are based on correlations, you also cannot make cause/effect statements.

If you want to decide if smurfs are deciding the outcome of your games and you want to use CS as an indicator, you should compare CS performances in wins vs CS performances in losses. If you’re seeing a greater difference of CS on average in losses than what you’re seeing in wins, then perhaps you could interpret that as smurfs are crushing hardcore while you’re winning by an expected margin. There’s a lot of issues with this because this only explains smurfs impacting your losses (you can easily have smurfs help you win); it also assumes every smurf is significantly better than that elo, which also isn’t true. In truth, this is a hard concept to measure and the one thing your regression shows is that the more games you play, it’s more likely you’ll be placed in an elo you belong.

Nevertheless, good job. Neat presentation.

GripaAviaraMay 1, 20191 votes

There are tons of smurfs this season. While I don't know if they are just trapped in silver or not, it's wtf

5050BSMay 1, 20191 votes

So because Diamond players are getting like +9/-25 LP they dont play their main accounts because its a complete waste of time.

This means that they are Smurfing and many are in Gold+ so that ruins the Plat and Gold Ranks because they really are Diamond ranks now. So THOSE players are Smurfing and are in Silver so Silver Players are really playing vs Gold+ players Runing THOSE Ranks.

So Silver players are fighting hard to not fall to Bronze but because you need wins to keep or grow your MMR and you cant win vs Smurfs.

S9 is LITERALLY broken.