Jump to content

Balancing Teams


30 replies to this topic

#21 PPMcBiggs

    Member

  • PipPip
  • The Defiant
  • The Defiant
  • 42 posts

Posted 06 June 2013 - 08:41 AM

View PostSgtKinCaiD, on 06 June 2013 - 08:09 AM, said:

When half of your team did individually less than 100 dmg, it's not a domino effect.

Here's a guy who gets it.

Let me see if I can boil the issue down.

A players elo changes when his team wins or loses. This may or may not be an indicator of that players performance. I would say it is most likely not, but I digress. Now, because elo changes based solely on your teams win or loss your elo has little to do with how you performed individually. So, elo is really the average performance of your teams over time. Can you see where I am going with this?

This is elo hell.

Elo has nothing to do with my skill. Or, at best, elo is only 12.5% (1/8) my skill. The rest of the elo number is the average performance of my team mates.

Now, to address those who say elo needs a larger pool to work correctly here is an idea: make elo only activate when the pool is large enough otherwise completely random.

#22 xDeityx

    Member

  • PipPipPipPipPipPipPip
  • 753 posts

Posted 06 June 2013 - 08:43 AM

View PostPPMcBiggs, on 06 June 2013 - 08:41 AM, said:

Now, to address those who say elo needs a larger pool to work correctly here is an idea: make elo only activate when the pool is large enough otherwise completely random.


I agree with you here. MWO doesn't have the population to support a skill-based matchmaking system right now. It was better when it was based on tonnage only.

#23 PPMcBiggs

    Member

  • PipPip
  • The Defiant
  • The Defiant
  • 42 posts

Posted 06 June 2013 - 08:48 AM

View PostxDeityx, on 06 June 2013 - 08:38 AM, said:


Here's an article on why Elo Hell is a myth.

Here's a video.

You have 7/8 chances to get a bad player on your team. You have 8/8 chances for a bad player to be on the enemy team. Statistically speaking you play more often with idiots on the enemy than you do with idiots on your own team assuming you are a good player. The common denominator across thousands of matches will be you.

People think that Elo doesn't work because of the Dunning-Kruger Effect and because of perception biases that make it seem like the idiots are always on your team because you don't notice the enemy idiots as much. You always spec your own team when you die, and idiots on the enemy team you just kill and move on rather than examining their badness in detail like you do with your own teammates.

edit: i diot is a filtered word? Really? Really? /facepalm

Perhaps this article is true but because PGI is nice enough to give me stats at the end of the round I can prove that my team is bad, the other team is not and neither am I.

How do I get out of elo hell? I can prove that I am in it.

That article is really bad too. Whoever wrote that article assumes that elo is perfect and that particular implementation of it is perfect.

I dont think elo is perfect, and I really doubt PGI has implemented it perfectly. I mean after all I still have to pound on backspace rather than just hold it down.

Since, as you pointed out, the player base for MWO is so small that elo can't work properly why doesn't PGI disable it, or only turn on when the player base is large enough?

#24 NoZaku4U

    Member

  • PipPipPip
  • 59 posts
  • LocationThe Citadel, Pacifica

Posted 06 June 2013 - 08:58 AM

In my opinion, ELO is completely broken in MWO for one simple reason. Mechs are not balanced.

Take chess for example. ELO works great because both players have the exact same tool set available to them. The game is balanced.

A player in a Hunchback who has the same ELO as a player in a Stalker is at a huge disadvantage because the tools available to them are not equal.

Some combined algorithm of ELO and Battle Value might help. But mech capabilities have to be taken into account.

#25 xDeityx

    Member

  • PipPipPipPipPipPipPip
  • 753 posts

Posted 06 June 2013 - 09:03 AM

View PostPPMcBiggs, on 06 June 2013 - 08:48 AM, said:

Perhaps this article is true but because PGI is nice enough to give me stats at the end of the round I can prove that my team is bad, the other team is not and neither am I.


If you are pointing to any single round as an example of Elo not working, you're missing the point of Elo. It's strength is in the statistical evidence for it working. Everyone has anecdotal evidence of it not working. Also, how would you prove that your team is bad and the enemy team is good? Damage? Everyone has games where they walk into an ambush or just get popped early. Also some assets are inherently capable of producing more damage than others because of the (admittedly bad) role warfare that's in the game.

View PostPPMcBiggs, on 06 June 2013 - 08:48 AM, said:

How do I get out of elo hell? I can prove that I am in it.


Play more games. Eventually your Elo will be a reflection of your skill as you have demonstrated it. You can't prove you're in it though.

View PostPPMcBiggs, on 06 June 2013 - 08:48 AM, said:

That article is really bad too. Whoever wrote that article assumes that elo is perfect and that particular implementation of it is perfect.

I dont think elo is perfect, and I really doubt PGI has implemented it perfectly. I mean after all I still have to pound on backspace rather than just hold it down.


I share your skepticism of PGI's abilities but in this case all they are doing is copying something that has been proven to work in other games. The reasons that Elo isn't working out for MWO as well as it could have nothing to do with the characteristics of an Elo rating system but rather the characteristics of MWO (low population).

View PostPPMcBiggs, on 06 June 2013 - 08:48 AM, said:

Since, as you pointed out, the player base for MWO is so small that elo can't work properly why doesn't PGI disable it, or only turn on when the player base is large enough?


Because PGI thinks it works, or at least thinks that it is better than nothing. They might even be right that it is better than nothing, but since they are not transparent with their data there is no concrete way to know. They also need to test the matchmaker and fine tune it for when there are enough players for it to make sense. It's all about collecting data and to a lesser degree creating a pacifying effect for the whiners who complain about stomps.

In general though, if a player isn't breaking even in wins:losses they are more than likely a bad player who is falling victim to the Dunning-Kruger Effect if they think they are in "Elo hell."

#26 Star Captain Obvious Kerensky

    Member

  • PipPipPipPipPipPipPip
  • Bad Company
  • Bad Company
  • 500 posts

Posted 06 June 2013 - 09:04 AM

The Elo system will balance itself out after thousands of matches. However, if no one wants to play through thousands of matches because of unbalanced teams (Tonnage, Skill, etc), then Elo hurts the game more than it helps.

This is where the Elo system fails, and fails miserably.

Edited by Eldragon, 06 June 2013 - 09:04 AM.


#27 SgtKinCaiD

    Member

  • PipPipPipPipPipPipPipPip
  • Overlord
  • Overlord
  • 1,095 posts
  • LocationBordeaux

Posted 06 June 2013 - 11:39 AM

One point was forgotten : i'm pretty sure that a vast majority of players doesn't know that they are in ranked match system as the game does not propose a skirmish mode.

#28 Albert Meyburgh

    Systems Engineer

  • 74 posts
  • Location404

Posted 07 June 2013 - 09:50 AM

Hello everyone! Thanks for the interesting discussion. I like hearing the perception, arguments and suggestions put forth by people who understand parts of the system and those who have no idea--both perceptions are interesting! I'm unsure what details I'm allowed to share, so I will lean towards being conservative and not say much. Some things people have pointed out/explained/suggested are valid... some are not... :) We are studying the system, looking at probabilities, correlations, predictions vs outcomes, deviations, all things statistical, and have faith that things have improved, and will continue to do so over the months! I feel your pain--just be patient! No one said this was going to be easy on you or your pride!

#29 BlackBeltJones

    Member

  • PipPipPipPipPipPip
  • 460 posts

Posted 07 June 2013 - 10:29 AM

View Postsrccoder, on 07 June 2013 - 09:50 AM, said:

Hello everyone! Thanks for the interesting discussion. I like hearing the perception, arguments and suggestions put forth by people who understand parts of the system and those who have no idea--both perceptions are interesting! I'm unsure what details I'm allowed to share, so I will lean towards being conservative and not say much. Some things people have pointed out/explained/suggested are valid... some are not... :) We are studying the system, looking at probabilities, correlations, predictions vs outcomes, deviations, all things statistical, and have faith that things have improved, and will continue to do so over the months! I feel your pain--just be patient! No one said this was going to be easy on you or your pride!

Why say you'll "lean toward being conservative" in your statements when you really said nothing? To offer a qualifying preface does not necessarily mean that what followed was even remotely informative and the qualification does not mean that what you said would even be considered a conservative version of the information you have access to.

Added: No offense to you personally - I always question people when they express effort with no actionable path or constructive value.

Edited by BlackBeltJones, 07 June 2013 - 10:36 AM.


#30 keith

    Member

  • PipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 1,272 posts

Posted 07 June 2013 - 10:36 AM

added a weight balancing system would have WENT alot farther then JJ shake. saying only allowing X assaults, X heavy, then filling with mediums and lights. thats MORE BT lore then JJ shake. last time we heard something about that was, its after launch... BUT the horrible JJ shake got thrown in very quickly

#31 Rocket Puppy

    Member

  • PipPipPip
  • Overlord
  • 97 posts

Posted 08 June 2013 - 08:03 AM

ELO Hell can be a real possibility depending upon how the game does math. A player already at the top will have an easier time starting a new character and getting their than someone at the bubble of crossing brackets.

If you are doing very well the game may continue to match you with tougher and tougher opponents, sending you soaring past the lower or medium skilled players as the game is actively trying to find the correct tier of players to match you with. A player gradually moving up can get stuck on a bubble as he win/loses matches right on the edge some times due to circumstances outside of his control.

Some of the best Halo 2 players were constantly stuck at middling tiers due to this. A great player can make a big difference on a team but can not make up for the skill of 7 other players. This was a big reason Halo 3 and later dropped a modified ELO system and moved to trueskill.

If you feel you are stuck on a bubble try to find players of similar skill and play with them, cut out the random chance of poor players and make sure you always play with people who can pull their own.

ELO is far from a perfect system for team-based games. It is an extremely good system for 1v1 games.

Edited by Rocket Puppy, 08 June 2013 - 08:07 AM.






1 user(s) are reading this topic

0 members, 1 guests, 0 anonymous users