Riot, are you lying to us?

Plasma Servant·Oct 9, 2016·10 votes·979 views

Okay, there are ranged champions, there are meele ...

if they were 50/50, having all ranged on one team in ARAM would not be crazy, but to have all ranged on one team and all melee on the other would be 1:500 ... how often do I see it? Way more than once in 500 games.

I own all but 2 champions (140 champions) so you would think it would be a poisson distribution ... if ARAM were really ARAM. So how in the course of 2 days have I seen viktor 5 times?

If it isn't random, advertise it as such.

35 Comments

LostFr0stOct 9, 20162 votes

Ayyy, you get all melee vs ranged games too? Thought I was goin' crazy.

Critmaster GarenOct 16, 20162 votes

i dont know why op is downvoted.

aram wont be truly random as long as riot doesnt enable all champions for everyone in that mode.

until then, we have to face sona in 50% of our games.

Irelia BotOct 16, 20161 votes

or nid + lux + ezreal

Plasma ServantOct 15, 20161 votes

okay riot, post the code for selection of the random champion ... not anything else, just how you select random ... I've done enough samples that I am positive that it is not random. So either you have a bug or you don't know what you are doing.

DrCyanideOct 16, 20161 votes

You realize it's harder to make it not random than it is to keep it random, right? Pretty much any programming language has a library that handles all the random events for them. Why would Riot spend more time and QA effort to make a worse random system?

Plasma ServantOct 16, 20161 votes

If they reseed the random number generator then it will not be random. It will come out with the same sequence of numbers over and over.

DrCyanideOct 16, 20161 votes

If they reseeded every match you'd get 1/10 champions every time you queued up until you bought a new champion (one for each player slot on a team, then two teams). Again, nothing that effects Melee vs Ranged.

Plasma ServantOct 16, 20161 votes

If they reseed the random number generator then it will not be random. It will come out with the same sequence of numbers over and over.

I suppose I just want to know what they are doing. Are they either stupid and don't know how to program it, or are they liars. They could easily post just the code that does the random selection for each champion so it is publically able to be verified. I doubt they will do that. And I seriously doubt they are using the built in (intel processors) noise generator for the random number generator. If they are using a pseudo random generator, they are using a really stupid one.

DrCyanideOct 16, 20162 votes

If they reseed the random number generator then it will not be random. It will come out with the same sequence of numbers over and over.

You already know the answer to that is "they're not doing it", because otherwise you'd only roll 10 champions. You've far exceeded that in the last week alone.

I suppose I just want to know what they are doing. [...] They could easily post just the code that does the random selection for each champion so it is publically able to be verified.

You know how many times people want to "see the code"? Every few months there's another post about someone wanting to see the code for matchmaking, or mystery chests, or champion mastery, or something. Showing the code is a Pandora's Box.

Are they either stupid and don't know how to program it, or are they liars.

I think you need to show some proof that it's broken, rather than than make baseless claims. You played a grand total of 1 game this week (out of 33 ARAMs) where it was all ranged vs all melee.

Honestly, it doesn't sound like you're observing Pseudo-random behavior, it sounds like you wish there was pseudo-random behavior.

Of my champions, only 22 are melee, 38 are ranged. I'm almost twice as likely to roll ranged, just because of my pool. Don't assume it's "50/50", because it won't be.

Plasma ServantOct 16, 20161 votes

okay, I'll do yet another sample, do the math, show the poisson distribution project and why I am 95% confident is it wrong. I find it hard to believe that with 140+ champions (I have all but one) that I would get viktor in 4 of 5 games).

Plasma ServantOct 17, 20161 votes

Okay, there were 132 champions ... my bad on counting. Hard to tell where I was in the store.

frequency distribution (how often each of the champions is randomly selected) should follow a bell curve (i.e., a normal distribution). Having any champion have extremely large values (for example, aatrox selected 20 times with the next most frequent selected twice) would suggest non-random. there should be a continually decreasing number of champions selected more (not at every level, but normally distributed). My sample size is still a little small ... I've not played 500 games since starting this ... but it does seem that viktor shows up a lot (only champion that I've had initially before rr 8 times, next highest was 3 times for graves).

Plasma ServantOct 16, 20161 votes

I have 141 champions. the mix should be that I'd see nearly all of them. there are ones that just don't show up. Ones that show up way too often (that 4 of 5 convinced me I should do the math).

DrCyanideOct 16, 20161 votes

There aren't 140+ champions, there's 134 with Ivern's release
Not everyone owns that many champions. You can only do the math on your own roles, not your teammates or enemies.

Plasma ServantOct 16, 20161 votes

Okay, I counted wrong? lol there are 132 ... I probably counted the free to play twice. But even so, I have (assigned to me) way more of viktor than teemo. I've almost never seen malzahar ... I saw mord just once (and I play aram a lot). I'm not talking about other people getting champions, I'm talking about what champion I get not being random. It is supposed to be.

It might be they knock out from a team those that have already been picked, but there are plenty of champions that are not on my team, yet never picked, or ones that I get way more often than what they ought to show up.

DrCyanideOct 16, 20162 votes

Define "ought to show up".

I ran a small Python program using the built in random function, and here's what I got (note that the boards messes up the spacing, but it's easy enough to fix).

import random hits = {} for _ in range(0,132): ... c = random.randint(1,132) ... if not c in hits: ... hits[c] = 1 ... else: ... hits[c] += 1 ... hits {1: 1, 2: 2, 3: 2, 4: 3, 6: 2, 8: 2, 11: 3, 12: 1, 130: 2, 14: 1, 15: 2, 16: 1, 17: 1, 18: 1, 19: 1, 21: 2, 22: 1, 25: 1, 28: 2, 29: 1, 30: 1, 31: 1, 32: 4, 33: 1, 35: 1, 37: 1, 38: 4, 39: 1, 41: 4, 42: 1, 46: 2, 47: 1, 48: 3, 49: 1, 53: 2, 55: 1, 56: 1, 57: 2, 58: 1, 59: 1, 62: 2, 63: 1, 64: 1, 65: 2, 66: 1, 67: 1, 68: 1, 69: 1, 70: 2, 73: 1, 74: 1, 75: 1, 76: 4, 77: 1, 13: 1, 80: 1, 82: 2, 83: 1, 84: 2, 86: 3, 88: 1, 91: 4, 92: 1, 93: 1, 94: 1, 95: 2, 131: 1, 101: 4, 104: 2, 107: 1, 108: 1, 109: 2, 111: 1, 113: 2, 115: 3, 117: 2, 119: 1, 120: 1, 121: 2, 122: 1, 123: 1, 124: 1} len(hits) 82

So using a tried-and-true random function, running 132 times (probability says everyone picked once):

50 were never picked
49 were picked once
22 were picked twice
5 were picked three times
6 were picked four times

By the sounds of it and by the look of your match history, it would seem safe to say that yes, they are just using a random system. A pseudo random generator would likely have put more emphasis on getting you champions you haven't played recently.

WoonStruckOct 16, 20161 votes

IIRC, you cannot reroll champions that have already been rolled, even if someone rerolled them.

Considering that, its likely that champions are 'reserved' when you arrive in champion select. That would mean you actually roll 3 champions at a time. If that's the case, when it comes to over-laps and since its dependent probability, it likely gives priority to the one with fewer champions to avoid situations where someone would end up without a champion at all.

All of this is basically speculation from my own observations, though. Can you either confirm or deny that you've never seen a rerolled champion rolled by someone else? That'd make or break this notion right away.

DrCyanideOct 16, 20161 votes

There have been people who have complained on the boards about getting champion A, rerolling chamion B, rerolling again and getting champion A. (Needless to say, they were upset at that, which is why they were posting.) I don't think I've heard of someone getting champion A, then rerolling champion A though.

The reroll code likely removes any champion currently on your teams selected champions (including your own) from your champion pool, and uses a random function to select a new champion from your remaining pool. From a programmers perspective, it makes sense that they'd reuse that code for the initial rolls too, running it 5 times per team.

... Actually, it's just the Random Champion code that's been around for years (since beta?)...

Plasma ServantOct 17, 20161 votes

that is about what I'd expect from random. The distribution should be normal. so before you start seeing a champion picked 8 times, you'd expect a lot more champions picked once. Getting the same champion (before rr) in 4 of 5 games smacks of not random ... having him 8 times in two weeks?

Plasma ServantOct 17, 20161 votes

Oh, btw, probability says that they won't be picked once. It says that the frequency of the champions selected would be normally distributed (bell curve). if you see one that is way more frequent, it tends to point to non-random. At some point, the probability of the actual process being random becomes less than the probability that it is something other than random.

One thing that I think ought to be part of this is that the summoner with the MOST champions should have his selected first, then the second most, and so on. So that those that have limited champions (i.e., they have an ARAM account) get last choice. Those that have all champions should have near perfect normal frequency distributions.

Plasma ServantOct 16, 20161 votes

And I've rr a ton of times when I see it all melee to obtain a ranged. So if there are just one ranged, it very well could be rr allowed a melee to become ranged. It should not be necessary.

WoonStruckOct 15, 20161 votes

ARAM only accounts.

People only buy champs good at ARAM.

Sad thing is, it doesn't even help them too much. I don't abuse such a thing and I'm still in top 4% of the MMR distribution at ~1,731.

Comparatively, you're ~882. There's something more than going against range that's losing it for you. Hell, it could even be that there are more ARAM-only accounts at lower ARAM MMRs, leading to a 'if you can't beat them, join them' kinda thing.

Plasma ServantOct 16, 20161 votes

I have all but one champ ... and that I'll have next week. I'm looking at the distribution of how often a champion is selected from my own group and it is not random. I don't know if they are using a pseudo random generator and resetting the seed, but they ought to be using the true random generator built into intel processors.

Plasma ServantOct 16, 20161 votes

and besides, testing for random is not anything related to MMR. It is related to doing a test of the null hypothesis of "the number of times a champion is selected follows a poisson distribution" and picking a confidence interval and see if it is rejected.

This question has nothing to do with the skill of the person playing ... it is the same for all summoners regardless of skill level.

WoonStruckOct 16, 20161 votes

The higher you go, the fewer people there are that rely on ARAM only accounts.

The people that use ARAM only accounts are the ones that feel like they need that advantage...which are lower MMR players.

Most players of decent skill wouldn't care to take the time to buy all of the A tier and higher champions for ARAM on another account just for the purpose of playing ARAM considering most of the decent skill players primarily play summoner's rift.

BreakuOct 16, 20161 votes

How do you find ARAM mmr?

WoonStruckOct 16, 20162 votes

https://na.whatismymmr.com/WoonStruck

BreakuOct 16, 20161 votes

The ARAM one is neat, but it's ranked MMR is way off of what op.gg has.

http://prntscr.com/cuuvet

WoonStruckOct 16, 20162 votes

Its measured in a different way that's likely more accurate.

In the past, ~mid-gold was never up around 1700-1800; It was around 1500-1600.

However, its a newer measurement, so it might be off either way.

That little blue message at the top is new, and values might be off atm. 1200 seems a bit low for you haha.

BreakuOct 16, 20161 votes

Oh hm. It's def interesting though. I guess I should try harder in my ARAMs-I didn't even know it had an MMR.

TyrekGoldenspearOct 16, 20161 votes

You also have to remember that each player has a different amount of ranged and melee champs, and therefore a much much more random experience in the sense that you cannot accurately predict the outcome of an ARAM pick unless you knew all the champs owned of all the players.

Then you could plug numbers into the formula, (using dependent chance since one players pick could remove anothers from being picked as well) and find out the chance of champs being picked on a team.

Plasma ServantOct 17, 20161 votes

okay, I've seen viktor (before rr if I had to) 8 times in the last couple of weeks. I've not seen any champion that often. So if it isn't random, make it balance.

RPK9KM4K4AOct 17, 20161 votes

What you are describing is the antithesis of random, looking for more or less equal distribution of each character being picked is NOT random.

https://youtu.be/mEN7DTdHbAU?t=1m24s

Plasma ServantOct 17, 20161 votes

No, but I do look for the frequency of the distribution to be normal.

The WhamboozlerOct 17, 20161 votes

ARAM isn't random. People make ARAM accounts that only have strong ARAM champos unlocked. It's common enough that you probably end up with at least one of them in every single match you play. Because people are assholes.

Plasma ServantOct 18, 20161 votes

riot could fix that too ... just have all champions available in aram, or at least assign the people with the most champions 1st.