Addressing the issue of ELO-hell.

Sweet! So when are you going to hook me up with some tasty ELO?

PayPal me 50€ and I boost your account into the heaven

PayPal me 50€ and I boost your account into the heaven

Cool, get me rank 1 Ostheer in 1v1 while streaming on twitch with camera and 50€ are yours.

Cool, get me rank 1 Ostheer in 1v1 while streaming on twitch with camera and 50€ are yours.

too trash for 1v1 Sadge

Thinking of it, I guess one of the main problems with ELO is that it doesnt care about who you are playing with in a first place and your performance in the game.

For instance, for instance if I'am rank 10000 paired with rank 10 vs another rank 10's, I will gain shit load of ELO in case my team wins, Elo doesnt care if I got hard carryed.

If I'm rank 10 paired with 10000 against same 10000, I will lose shit ton of ELO, dispite me trying to hard carry them.

At the same time, I will gain almost nothing if I managed to carry bad low rank teammates.

This is really a big risk low reward system. Because you legit can lose almost of your rank grinding prossess in a few really unlucky games.

Not saying that such indepth calculation is nessesery, but, at very least, in teamgames, teams could have had assigned rank based on the ranks of all players in mentione team. So if the game is super one-sided, at least loosing side wont lose crap ton of ELO and winning side wont get free ELO after stomping them.

If you look closely 50% of Hell consists of 66.66% of elo

https://web.pdx.edu/~newsomj/uvclass/ho_outlier%20and%20sample%20size.pdf

Standard error and standard deviation decrease as sample size increases. Rosbone’s insane rumblings about matchmaking fucking him up 10 times in a row and then continuing to do so with a significant effect no matter how many games you play have no basis in reality. If you play 200-300 games, matchmaking quality will regress towards the mean and you will get very similar numbers of good and bad cases of it.


Bayes theorem is irrelevant, I never said if you get 10 bad cases of matchmaking in a row you are more likely to get good matchmaking next time. Learn to read.

Lol, you just keep digging.......
There is nothing wrong with my ability to read.

You grabbed a random article on the effects of an outlier. You probably didn't notice the portion about the article analyzing the effects of a single outlier on a NORMAL distribution. The complaint about elo hell boils down to the fact that it isn't a normal distribution, and anyone with rankings less than about 6000 will get a right-skewed distribution for matchups.

If you wanted to make a counter-argument that was coherent instead of just slinging insults, you could've pointed out that people end up placed in the ladder depending on how they manage the right-skewed distribution, toxic teammates, etc. The distribution is worse for people that are higher on the ladder. If you're someone like Aerafield, Sanders, etc, practically every one of your teammates is worse than you, so you'd better learn to shut up and play.

If you look closely 50% of Hell consists of 66.66% of elo

Best use of statistics in this entire thread!

I don't think "ELO-hell" is a real thing, at least from my experience; I've never had a problem getting to my usual "top 200" rank in 4v4 (or any mode), on any of the factions I've tried. The only advice I can really give is to "not leave games". Leaving games almost always results in a loss, which of course lowers your rank. As a result, even if you win every match you think is "fair", you're still probably taking a lot of losses due the games you've left, cancelling out any wins. In my case, I only leave if one player on my team is clearly afk (no movement for the first 2 minutes), or if the game has clearly been lost (getting base-pushed, facing tanks when we've never held any fuel, etc.).

With that said, the match-making system desperately needs an overhaul in CoH3, for many of the reasons listed already. The biggest change, in my view, would be to limit the possible "range" of ranks within a match. As pointed out in the Legendary Team Game Matchups thread, you can currently get into some incredibly one sided matches, especially if you wait long enough in queue, or matches where the skill levels don't make much sense (top-50 players matches with ~5000 ranks).

This needs to be limited, as it creates an awful experience for everyone involved; the lower-rank players aren't going to be able to compete, and even worse, won't learn anything from such a one sided match, and the "pro" players are going to either find the match incredibly dull due to how one-sided it is, or incredibly frustrating as their teammates won't be able to play at their level.

Other changes would also be great, like leaver penalties, or adding a "skill bonus" to pre-made teams so they face "random team" players of a higher level (to compensate for the pre-made team's better communication), and of course a "ready check" when a match is found; but I think all of those are a much lower priority.

Best use of statistics in this entire thread!

Agreed

We have had several forum posts about this already.

WHAT I WOULD DO WITH THE CODE:
- Give premades a rank. Do not force them to smash 10 noob teams first.
- Do not make games where players are all ranked much higher than their enemy.
- Do not stick people on the same map they just lost on.
- Do not put people in the same spawns they just played on.
- Do not make matches with the weakest factions only.
- Do not stick the worst player in a match with other members in the next match.

I could go on and on....

Relic could easily do the first two and those are the biggest contributors. I care way more about you guys than Relic does, so I would do more steps.

Write this list to Santa Claus next winter maybe. On the other hand Soviet is by no mean the weakest 4vs4 faction and your vision of it is probably distorted because it is the most played allied faction at lower level being the easier.

If you want to get out of Elo Hell, start playing 1vs1, improve, then come back to 4vs4 and uplift your side. It doesn't work every match but instead of a 40/60 victories you may get 55/45 or 60/40 victories ratio.

We have had several forum posts about this already.

WHAT I WOULD DO WITH THE CODE:
- Give premades a rank. Do not force them to smash 10 noob teams first.
- Do not make games where players are all ranked much higher than their enemy.
- Do not stick people on the same map they just lost on.
- Do not put people in the same spawns they just played on.
- Do not make matches with the weakest factions only.
- Do not stick the worst player in a match with other members in the next match.

I could go on and on....

Relic could easily do the first two and those are the biggest contributors. I care way more about you guys than Relic does, so I would do more steps.

The first item makes sense, and I do agree that imposing some restrictions and handicaps on matchups between random and premade teams would probably be easiest and most impactful tweaks to implement. But as we go further down the list I fail to see how this would improve things noticeably without the drawbacks (i.e. drastically increasing search times) heavily outweighing the gains. What people don't realize is that putting multiple exclusion criteria like this into the MM algorithm makes finding games not linearly but exponentially harder the more players and restrictions are involved. If all of the above was implemented and you start searching right now, I'd wager you'd maybe find one game before the heat death of the universe - if you're lucky, that is.

heat death of the universe - if you're lucky, that is.

So your banking on Dark Energy continuing the expansion of the universe ... Top 1v1 players are backing Dark Matter and designing new strats around the big crunch. Which of course will happen long before Relic fixes the servers

I of course are pessimistic and figure the universe will under go a Phase Change just as Coh3 is released and we will all disintegrate. Then the last soul on Earth will be some marketing guy at Relic who will make a video about game design and Phase Change mechanics added to Coh3.

On the other hand Soviet is by no mean the weakest 4vs4 faction and your vision of it is probably distorted because it is the most played allied faction at lower level being the easier.

You saw all the graphics I posted earlier in this topic from PagePs site about soviets being the worst faction in 4v4 looking at only rank 200 games and above. I can assure it is much worse at lower ranks.

I also described why Soviets are UP in detail earlier in this topic.

https://www.coh2.org/topic/109981/addressing-the-issue-of-elo-hell/post/883742
https://www.coh2.org/topic/109981/addressing-the-issue-of-elo-hell/post/883688

Think it would be helpful if people would agree about what the are talking about.

So is everyone talking about 4vs4 random games?

You saw all the graphics I posted earlier in this topic from PagePs site about soviets being the worst faction in 4v4 looking at only rank 200 games and above. I can assure it is much worse at lower ranks.

I also described why Soviets are UP in detail earlier in this topic.

https://www.coh2.org/topic/109981/addressing-the-issue-of-elo-hell/post/883742
https://www.coh2.org/topic/109981/addressing-the-issue-of-elo-hell/post/883688

Those numbers don't show anything about faction balance, Honestly I used to think they do like you but in reallity they don't.

There are so many parameters behind that like what I said: maybe the pool of soviet players is less skilled than other allied factions, maybe a top100 Soviet player = top300 UKF player in term of skill. And maybe a top100 Soviet player = Top500 OKW players. Seems silly or stupid, well prove it.

There are 0 strong correlation between average V/D ratio and skill required per faction.

Another example is USF, the faction look like balanced if we look at those numbers, but that's probably not correct otherwise with all those sweat tears about Path&Scott OP all over the place the faction should be rocketing about 80% victory rate (at least below TOP200), which isn't the case which may show that the faction is artificially running about 50%+ victory rate thanks to Path&Scott and that without it would be far away from those 50%.

Lol, you just keep digging.......
There is nothing wrong with my ability to read.

You grabbed a random article on the effects of an outlier. You probably didn't notice the portion about the article analyzing the effects of a single outlier on a NORMAL distribution. The complaint about elo hell boils down to the fact that it isn't a normal distribution, and anyone with rankings less than about 6000 will get a right-skewed distribution for matchups.

If you wanted to make a counter-argument that was coherent instead of just slinging insults, you could've pointed out that people end up placed in the ladder depending on how they manage the right-skewed distribution, toxic teammates, etc. The distribution is worse for people that are higher on the ladder. If you're someone like Aerafield, Sanders, etc, practically every one of your teammates is worse than you, so you'd better learn to shut up and play.

The only guy who knows what he is talking about kudos thanks a lot.

You saw all the graphics I posted earlier in this topic from PagePs site about soviets being the worst faction in 4v4 looking at only rank 200 games and above. I can assure it is much worse at lower ranks.

I also described why Soviets are UP in detail earlier in this topic.

https://www.coh2.org/topic/109981/addressing-the-issue-of-elo-hell/post/883742
https://www.coh2.org/topic/109981/addressing-the-issue-of-elo-hell/post/883688

The first link is not top200 data. The faction matrix for top200 in 4v4 is basically empty since those matches don't happen. You must have drawn from all data available.

Axis definitely have an advantage in 4v4, and Soviets are one of the weaker factions. But there are quite some caveats though, Esxile pointed out some of them. I'd add the fact that SOV and OST are part of the base game. New players will probably pick this one up and check the game out or get it during one of the free CoH2 giveaways. This *could* drag down Soviet win rate as well. Is it proof? No. We have another hint that Wehrmacht is also often trailing behind OKW by ~1%, but also here we cannot tell it apart from actual imbalance.

Next, a lot of factors that you mentioned in other posts that get you into ELO hell (assuming it exists to such a degree that a reasonably skilled player can get dragged into it), will also randomly push you out. Is it fun or fair that CoH2 4v4 matchups have a high degree of randomness? Definitely not. Is it proof of real existance of ELO hell? Also definitely not.

No one is debating that sometimes the match maker just screws you over, even for multiple games in a row. At the same time, the MM will screw the opponent over and you'll rise again in ELO. IIf a faction is weak by design, ideally everyone gets downmatched to opponents with less skill but better faction. This is what the system will also converge to given enough games. Even if your point is that you will play with more "weak" Soviet players: Those have been downmatched as well. The last patch is months ago, rankings should have reorganized. It is shitty that the fluctuation around that optimal ELO score in CoH2 is quite high considering even an unbalanced game often lasts between 15-30 minutes plus search time. Just the random fluctuation can cause you to "lose" 2-3 evenings until you're likely to get your ELO back, either by skill or pure chance.

Relic definitely needs to fix the match making, but I doubt it has anything to do with "ELO hell".

Let's fix MM!

• Premade teams get no elo when playing against randos. Optionally, have checkbox to avoid playing with randoms when playing premade.
  
• Maximum difference of ELO between teams: 150.
  
• Way better map pool (especially 3v3 and 4v4) to minimize cheese strats.
  
• Make MP F2P in order to get more players in to optimize algo.
  

Also, responding to some cucks who seriously think that "it's all about fun":

Unfair MM -> Unfair game -> Elo Hell -> Unfun game.

Good luck having fun when your 2-3 teammates are barely able to build a single LV the entire game and end up spamming MGs and ATGs.

Lol, you just keep digging.......
There is nothing wrong with my ability to read.

You grabbed a random article on the effects of an outlier. You probably didn't notice the portion about the article analyzing the effects of a single outlier on a NORMAL distribution. The complaint about elo hell boils down to the fact that it isn't a normal distribution, and anyone with rankings less than about 6000 will get a right-skewed distribution for matchups.

If you wanted to make a counter-argument that was coherent instead of just slinging insults, you could've pointed out that people end up placed in the ladder depending on how they manage the right-skewed distribution, toxic teammates, etc. The distribution is worse for people that are higher on the ladder. If you're someone like Aerafield, Sanders, etc, practically every one of your teammates is worse than you, so you'd better learn to shut up and play.

The lower you are on the ladder, the more probable it is to be matched with a higher skilled opponent. At the same time, this loss will result in a lower ELO adjustment. I am not a computer scientist, but shouldn't the ELO system fix this scewed distribution by estimating the win chances before the game? I always though that this is how it works.