Jump to content

Elo Is For Chess, Not Mwo


198 replies to this topic

#141 NocturnalBeast

    Member

  • PipPipPipPipPipPipPipPipPip
  • Shredder
  • Shredder
  • 3,685 posts
  • LocationDusting off my Mechs.

Posted 10 October 2014 - 10:04 AM

View PostThorn Hallis, on 10 October 2014 - 02:16 AM, said:

Are we asuming that a player with a high ELO score would enter a battle with a subpar chassis? If this player does, would he really expect to be successful?

Any battlevalue system that does not include player skill would be ripe for abuse.


Roadbeer's suggestion for adding BV to ELO seems like the best solution.

#142 RalphVargr

    Member

  • PipPipPipPipPipPip
  • 292 posts
  • LocationTureded, Lanth Subsector, Spinward Marches

Posted 10 October 2014 - 02:28 PM

View PostMadcatX, on 10 October 2014 - 12:39 AM, said:


I agree, but sadly from what I see here, the only alternative is apparently Battle Value, which I too disagree with because it lacks the human component. Plenty of PvP games out there using all sorts of different systems.

On the flip side, could be worse, could be using a static tier system like WoT where really good players can play with newer players at any given time or really good players just decided to stick at a particular tier.

And stock tanks vs a fully upgraded one with fully trained crew...

There are better, and worse, thing then elo out there.



I am in recovery from World Of Warplanes, where you simply do not *dare* play above Tier 5, because you will be eaten by the sharks. Not so co-incidentally, WoWP has an average of about 1,000 active players, tops, at any one time. The "hardcore" killers have driven all the prey (like me) out of the pool.

You need low-skill players like myself to keep the game ecosystem alive. When suckers like me wise up, and walk away from the poker table, the sharks turn on each other. Not a good business model.

#143 Victor Morson

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Elite Founder
  • 6,370 posts
  • LocationAnder's Moon

Posted 10 October 2014 - 02:33 PM

View PostEd Steele, on 10 October 2014 - 10:04 AM, said:

Roadbeer's suggestion for adding BV to ELO seems like the best solution.


I've been begging for "BV" type matchmaking forever. I do not want them to use TT values for this.

A simple:

Player Skill (ELO is still awful, but some measure of it) + Mech base value + Mech upgrades (T1/T2 stuff - i.e. endosteel) + Weapon loadout (a lot of high tier weapons, or boated weapons) = Player Value.

Then matchmaker could just look at the teams and fill them with Player Values that get both teams to roughly the same amount.

Seems super simple for matchmaker to handle this and it sure would be easier to get games going than to worry about an ELO stat that doesn't even matter, really.

#144 Victor Morson

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Elite Founder
  • 6,370 posts
  • LocationAnder's Moon

Posted 10 October 2014 - 02:36 PM

View PostMadcatX, on 10 October 2014 - 12:39 AM, said:

I agree, but sadly from what I see here, the only alternative is apparently Battle Value, which I too disagree with because it lacks the human component. Plenty of PvP games out there using all sorts of different systems.


I think some measure of pilot skill (NOT team based ELO because that is utterly worthless) should figure on top of the 'mechs BV, for a total "Player BV." There's no reason they have to be mutually exclusive systems.

Also, the player element is even visible in the 'mechs. A new guy with non-boated standard weapons running a standard frame with no mastery skills like Speed Tweak (which should also absolutely figure into this!) would be far, far, far less than a top tier Timberwolf pilot running meta guns even before player BV started to factor.

#145 Diablobo

    Member

  • PipPipPipPipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 1,014 posts
  • LocationOn your six

Posted 10 October 2014 - 05:17 PM

Yes, I agree with the suggestion to combine player skill rating with a BV-like system. Obviously the TT values are not going to carry over, and there shouldn't be any sort of team W/L affect on the rating at all. Instead of wins and losses, the performance per match (score) should be used. Basically the score would be a good predictor. If someone consistently gets a high score in their games, it can be said that they are more skilled than the average players who gets carried by good teammates. A good player saddled with bad teammates is just as unfair as a weak player who teams up with higher skill players. Neither player should be penalized/rewarded.

I think that wins and losses shouldn't really be considered, but it might be interesting to add scores to the win loss Elo rating. Factoring in scores would give a more accurate picture of a player's skill. Adding scores to wins and losses certainly would be an improvement.

Edited by Diablobo, 19 October 2014 - 05:20 AM.


#146 MischiefSC

    Member

  • PipPipPipPipPipPipPipPipPipPipPipPip
  • The Benefactor
  • The Benefactor
  • 16,697 posts

Posted 10 October 2014 - 06:02 PM

Win/loss is the only real interpretation of how likely someone is to win or lose. That's the whole point.

Again, the best players in the game, do they win more than they lose? Worst players, lose more than they win?

Any other system can be gamed. Right now I'm running a TW with UACs and a good TC. It does a lot of damage and destroys a lot of components. It's not the best killer but it gets me a lot of damage and a high score. It doesn't make me the most useful but it does grind cbills.

#147 Diablobo

    Member

  • PipPipPipPipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 1,014 posts
  • LocationOn your six

Posted 19 October 2014 - 05:28 AM

Having someone who achieves a high W/L ratio by grouping up with successful teammates does not automatically make that person better. Being in a really good team artificially inflates rating just like being in a bad team would deflate. I'm not sure of the exact statistics, but I'm pretty sure that if you look at the end of match splash screens, you will see that the winning team usually does more damage than the loser. There is another problem as well. If you are on a team that manages to capture in assault or win on points in conquest, it says little to nothing of your individual skill level. Also, it is possible to boost the rating of a disconnected player when the team he drops from still manages to win. It happens rarely, but the idea of a disco getting an increase irks me. Ratings should be based on individual performance, not how good your team is.

Edited by Diablobo, 19 October 2014 - 05:32 AM.


#148 Willard Phule

    Member

  • PipPipPipPipPipPipPipPipPip
  • The 1 Percent
  • The 1 Percent
  • 3,920 posts
  • LocationThe Omega Company compound on Outreach

Posted 19 October 2014 - 07:17 AM

The biggest problem with the whole Elo system is how progression is handled.

Right now, YOUR individual score is based upon the performance of a lot of random TEAMS....unless you're a group player, I guess.

Elo, by itself, would be just another system to develop a numerical value to an individual to indicate their skill if it were based on your performance, win or lose, period. Base it off of K/D, W/L, Average damage, whatever....but making us dependent on the performance of a team for our own progression, THEN artificially inflating new player's Elo score to the mid-range and forcing the general populace to have to carry them pretty much guarantees that you will NEVER go high enough on the scale to actually start playing with people of your skill level.

Fix that problem first, then worry about the whole BV thing. Once you can come up with a way to rate individuals, you'll have a solid number to use as a "pilot score." Without a "pilot score," It doesn't matter what values you assign to the 'mech....there has to be a value for the pilot or then entire thing is meaningless.

That and ECM is an issue. You cannot simply assign it a value and move on. Electronic systems, especially ECM, should be tracked and divided as equally as possible on top of the whole BV thing....but that's a mechanic for later. Figure the whole BV thing out first.

#149 ztac

    Member

  • PipPipPipPipPipPipPip
  • 624 posts

Posted 19 October 2014 - 08:46 AM

ELO is used in many games. But it should not be, In fact it is odd that devs use ELO ratings and totally fail to understand their own games and that you can not use an ELO or any other rating without it getting highly complicated (rating for every mech for every player taking the loadout into consideration too).

There is also the fundamental flaw that this is a team based game , so any rating would in part be based on your team!

But devs are too smart to acknowledge this.

#150 Willard Phule

    Member

  • PipPipPipPipPipPipPipPipPip
  • The 1 Percent
  • The 1 Percent
  • 3,920 posts
  • LocationThe Omega Company compound on Outreach

Posted 19 October 2014 - 02:53 PM

View PostNick Makiaveli, on 19 October 2014 - 09:04 AM, said:

Elo isn't perfect. But the OP is acting like ONE match can turn around you Elo. A high Elo player in anew Commando is likely a better teammate than a low one in a tricked out Spider. Even if not, it takes a lot of matches to actually make a difference.


Even more so if you don't play in the Group queue ever. Considering that there is no Elo separation there...the better you are, the more you have to carry...it's almost impossible to move your Elo at all.

View PostNick Makiaveli, on 19 October 2014 - 09:04 AM, said:

Lastly, if you don't like it, then wait for CW. Last I heard Elo won't be involved so it will be more realistic. See what I did there? Don't like it, don't play it.


The only problem with that statement is that CW is going to be released slowly. Lone Wolves and Merc Corps won't even be able to participate at first. Basically, they're releasing it for the organized Groups first....then they'll find a way to shoehorn us solo players into it. It's going to be a complete mess. But, what do you expect?

View PostNick Makiaveli, on 19 October 2014 - 09:04 AM, said:

Yes yes I know nothing will change if we don't complain but sure why not? Just toss a bunch of resources to totally revamp the MM right after they just tweaked it.


Not sure what you mean by "tweak." The MM was completely re-written around the time of the Clan release. That's when we got the whole "the better you are, the more you carry" MM that we have now.

The only "tweaking" was the addition of the "soft map" thing that they took out when the Group players started whining.

If you're primarily a Group player and only see the Solo queue during events, you're not seeing the day to day derpness of the solo queue....believe me.

View PostNick Makiaveli, on 19 October 2014 - 09:04 AM, said:

Not like we need ghost heat looked at. Or the orphan pilot skills. Etc.


True enough. Just more stuff on the list that is critical but won't get fixed in our lifetime.

#151 Nick Makiaveli

    Member

  • PipPipPipPipPipPipPipPipPip
  • Bridesmaid
  • Bridesmaid
  • 2,188 posts
  • LocationKnee deep in mechdrek

Posted 19 October 2014 - 03:05 PM

View PostWillard Phule, on 19 October 2014 - 02:53 PM, said:


Even more so if you don't play in the Group queue ever. Considering that there is no Elo separation there...the better you are, the more you have to carry...it's almost impossible to move your Elo at all.


You are saying that Elo is used in the solo queue? That's news to me and I think you are wrong.



View PostWillard Phule, on 19 October 2014 - 02:53 PM, said:

The only problem with that statement is that CW is going to be released slowly. Lone Wolves and Merc Corps won't even be able to participate at first. Basically, they're releasing it for the organized Groups first....then they'll find a way to shoehorn us solo players into it. It's going to be a complete mess. But, what do you expect?


Uh...didn't they just talk about LW and Mercs being able to change House allegiance between campaigns/seasons? And that if a Merc unit can't field a full team of 12, that solos would be added?


View PostWillard Phule, on 19 October 2014 - 02:53 PM, said:

Not sure what you mean by "tweak." The MM was completely re-written around the time of the Clan release. That's when we got the whole "the better you are, the more you carry" MM that we have now.

The only "tweaking" was the addition of the "soft map" thing that they took out when the Group players started whining.


Not sure we are communicating here. They are working on other things, and the MM works well enough for many players so hardly makes it a high priority.

View PostWillard Phule, on 19 October 2014 - 02:53 PM, said:

If you're primarily a Group player and only see the Solo queue during events, you're not seeing the day to day derpness of the solo queue....believe me.


Nope, I often play in the solo queue when I am just not in the mood to play with a group, or only have a short time, expecting family aggro etc. Or sometimes when I just want to test a new build, or farm Cbills by wiping out derps in my 30 LRM Kit Fox.

View PostWillard Phule, on 19 October 2014 - 02:53 PM, said:

True enough. Just more stuff on the list that is critical but won't get fixed in our lifetime.


Hyperbole much? Maybe it's time for you to take a vacation and play another game. Come back with a fresh perspective.

#152 Willard Phule

    Member

  • PipPipPipPipPipPipPipPipPip
  • The 1 Percent
  • The 1 Percent
  • 3,920 posts
  • LocationThe Omega Company compound on Outreach

Posted 19 October 2014 - 03:51 PM

View PostNick Makiaveli, on 19 October 2014 - 03:05 PM, said:


You are saying that Elo is used in the solo queue? That's news to me and I think you are wrong.


Supposed to be would be a better way of describing it...but, what with the system being as janky as it is, it's hard to tell.

In theory, New Players are supposed to have an Elo that's low enough to keep them out of general population until after their first 25 matches...that was put in patch notes...but it doesn't happen.

After the first 25 matches, no matter how well or crappy you did in the first 25, you are given an Elo score of 1700 or 1800....right smack in the middle.

Then....the only way to rise to the top or sink to the bottom is for your team to do differently than what the Matchmaker predicts for you in any given match.

Movement on the scale, positive or negative, is so SLOW that it is hardly noticeable.

Then add to that the range that the Matchmaker is allowed to search through to make teams of even "team average" Elo scores.....meaning, yes, they will balance high Elo players with multiple low Elo players.

Overcomplicated, underthought, broken and ignored......that's the Elo/Matchmaker system in a nutshell. But, those descriptive terms could be used for just about anything PGI has ever "fixed" in the game.




View PostNick Makiaveli, on 19 October 2014 - 03:05 PM, said:

Uh...didn't they just talk about LW and Mercs being able to change House allegiance between campaigns/seasons? And that if a Merc unit can't field a full team of 12, that solos would be added?


Something like that. But the problem is with changing allegiance. I don't know if you've got a Company or you've ever tried to change yours but it doesn't work like that.

When I set up Phule's Company, I was aligned with Clan Wolf. The other two guys in Phule's Company didn't want to be Clan, so I tried to switch to Merc Corps. I had to disband the unit, switch my allegiance, reform the unit with a different spelling/symbol and reinvite the previous members.

That's why I'm a bit leery of the whole "give you the chance to change allegiance between seasons" crap. One would think that if a Mercenary unit took a contract from a House, it would have temporary allegiance automatically. We'll have to see how it pans out. From my standpoint, since I'm the only one in my Company still playing, I went back to Wolf so I could play on the Clan side of CW....I intend to be one of those "filler" guys.



View PostNick Makiaveli, on 19 October 2014 - 03:05 PM, said:

Not sure we are communicating here. They are working on other things, and the MM works well enough for many players so hardly makes it a high priority.


Ok, not meaning to pull your chain, but are you referring to the Group Queue or the Solo Queue?

Perhaps the times we play are significantly different or your Elo is significantly higher than mine and you simply don't experience what I do but......ever since they got rid of that "soft vote" thing, the solo queue reverted to a state that was worse than before it.

During the "soft vote," it was cool. Very, very few "new players" in the mid-range games...lots more instinctive teamwork, better played games. Sure, you had to play Conquest every once in a while, even when you didn't have it checked off, but the other team had just as many guys that don't know anything about Conquest as our team did....so, we had a brawl somewhere while others were capping. It wasn't too bad. Well worth the tradeoff for teams that were much more equal skillwise.

I can imagine how the Group queue felt, though. You get all your guys together for a league match, the other team does the same (a feat in and of itself)....get everything all set to go and.....Conquest. Not part of the program. Everyone bail. Requeue, try again. And since the Organized Groups are the focal group for PGI, its understandable that they were catered to. Sucks for us solo guys, but it is what it is.

View PostNick Makiaveli, on 19 October 2014 - 03:05 PM, said:

Nope, I often play in the solo queue when I am just not in the mood to play with a group, or only have a short time, expecting family aggro etc. Or sometimes when I just want to test a new build, or farm Cbills by wiping out derps in my 30 LRM Kit Fox.


I'm guessing we play at significantly different times then, if you're not seeing what I'm seeing there. Most of the time, it's the team with the most ECM covered LRMs (and fewest trial mechs) that win.



View PostNick Makiaveli, on 19 October 2014 - 03:05 PM, said:

Hyperbole much? Maybe it's time for you to take a vacation and play another game. Come back with a fresh perspective.


Dude. Seriously. I know you're not new here. Think back to the beginning. I've only been here since Closed Beta, but I've heard some of the ghost stories.

Tell me about one thing.....ONE thing....that was a problem early on that was absolutely, completely fixed without having to apply bandage after bandage, overcomplicating the issue and breaking it further.

I'm not saying that to shake your tree or to pick a fight, I'm just trying to make a point. Even if you can name one or two things, the list of items that have been a problem since Alpha and Beta that STILL haven't been corrected is much, much longer.

The Elo/Matchmaker system was originally a bandage. It was a placeholder for something better they were planning to implement...they threw this in, kicked it, beat it, slapped it around and eventually re-wrote it. It's still the same bandage they came up with in the first place....just more complicated, broken and useless.

Ghost Heat was a temporary fix for something they didn't know how to address at the time. Once it got implemented, it got tweaked over and over until someone high up said "hey, this is a dandy mechanism for manipulating things easily. Let's keep it" and we got stuck with a really ridiculous heat system that's not even close to either BT or canon lore.

I could go on and on about temporary fixes that stayed with us, but I've made my point.

It is what it is, man. I'm still here, I still play the game. I get bored and/or ragequit far more often than I used to and just go play Pathfinder Online but, it still doesn't change my feelings for the lore and franchise. I don't have a lot of faith in PGI to actually develop the game the way they say they're going to....not after all the behind the scenes shenanigans with Trannyverse and crap.

#153 gregsolidus

    Member

  • PipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 1,352 posts

Posted 19 October 2014 - 03:59 PM

We really do need a long term adjustment to matchmaking that both factors the player and his equipment into the the player and his equipment into the equation but the game is such a cluster**** of things that need to either be added or tweaked we may not see it for years.

#154 MischiefSC

    Member

  • PipPipPipPipPipPipPipPipPipPipPipPip
  • The Benefactor
  • The Benefactor
  • 16,697 posts

Posted 19 October 2014 - 04:37 PM

New players start at 1000 or 1200. 1800 is an excellent score.

#155 Willard Phule

    Member

  • PipPipPipPipPipPipPipPipPip
  • The 1 Percent
  • The 1 Percent
  • 3,920 posts
  • LocationThe Omega Company compound on Outreach

Posted 19 October 2014 - 04:59 PM

View PostMischiefSC, on 19 October 2014 - 04:37 PM, said:

New players start at 1000 or 1200. 1800 is an excellent score.


....and is what is given away after the first 25 matches, no matter how well you did or didn't do during them.

Maintaining the 1800 you're given is easy, since movement on the scale is inconsequential and the spread allowed to find teams is so huge, it hardly matters.

#156 Belorion

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 5,469 posts
  • LocationEast Coast

Posted 19 October 2014 - 05:06 PM

This is an old tired argument, never mind that tons of video games now use this system to great effect.

#157 oneproduct

    Member

  • PipPipPipPipPipPip
  • 213 posts

Posted 19 October 2014 - 06:07 PM

View PostDukeDublin, on 09 October 2014 - 04:56 PM, said:

League of Legends took elo out, that should be evidence enough.

Just an example don't hate.


LoL did not remove ELO, they merely hid it from players and grouped them into leagues based on their hidden ELO. DotA and CS:GO use ELO as well as examples of other games.

#158 Mawai

    Member

  • PipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 3,495 posts

Posted 19 October 2014 - 06:18 PM

View PostDiablobo, on 10 October 2014 - 05:17 PM, said:

Yes, I agree with the suggestion to combine player skill rating with a BV-like system. Obviously the TT values are not going to carry over, and there shouldn't be any sort of team W/L affect on the rating at all. Instead of wins and losses, the performance per match (score) should be used. Basically the score would be a good predictor. If someone consistently gets a high score in their games, it can be said that they are more skilled than the average players who gets carried by good teammates. A good player saddled with bad teammates is just as unfair as a weak player who teams up with higher skill players. Neither player should be penalized/rewarded.

I think that wins and losses shouldn't really be considered, but it might be interesting to add scores to the win loss Elo rating. Factoring in scores would give a more accurate picture of a player's skill. Adding scores to wins and losses certainly would be an improvement.


I think before you can use match performance for a rating system you need to show how that match performance skill actually correlates to win/loss.

Here are some cases:
- LRM boats can do a lot of damage and get a lot of assists ... but may or may not contribute that much to winning the match
- There are some players out there who could care less about team play ... they optimize their play to survive and do maximum damage, they snipe, they do their own thing, they do a decent amount of damage and get good kills and assists ... but then love to whine about how their side didn't win, how everyone else was fail and it is obvious from the numbers that it was everyone else's fault ... when in a number of cases I have seen it was the fault of the individual for not supporting the team ... for letting a lance get cut off and die
- then there are the ECM snipers (Raven, Cicada, Spider) ... I have seen some of these post great numbers ... but are they helping the team actually win? What about the ECM pilot that stays with the team ... protects from ECM ... does a little front line sniping ... doesn't get great numbers but really makes a difference in win/loss?

It only takes about 150 damage to core an Atlas through the front. Sometimes there is an efficiency factor that you have to consider in not just how much damage is done but how it is applied.

#159 MischiefSC

    Member

  • PipPipPipPipPipPipPipPipPipPipPipPip
  • The Benefactor
  • The Benefactor
  • 16,697 posts

Posted 19 October 2014 - 06:31 PM

View PostWillard Phule, on 19 October 2014 - 04:59 PM, said:


....and is what is given away after the first 25 matches, no matter how well you did or didn't do during them.

Maintaining the 1800 you're given is easy, since movement on the scale is inconsequential and the spread allowed to find teams is so huge, it hardly matters.


Sorry, let me clarify.

New players have a 1000 or 1200 Elo during and after their 25 matches. There are not enough people playing MW:O with an 1800 Elo to consistently fill matches; that's why you've got high Elo players matched with low Elo. 1800 and up is exceptional; as a percentage of the population people with an 1800 Elo are probably around 10% or less.

Most players don't get far from 1200. That is not, however, a problem with Elo - it's a problem with people. Most people are not that good. Not at MW:O or anything else. Practice doesn't make perfect; perfect practice makes perfect. Someone with 2,000 drops in bad mechs with bad habits repeating the same bad behaviors is only good at being bad; they're not, in truth, dramatically superior to a newbie. They just feel they should be because we've got this post-modernist idea that knowing a little about something is the same as being an expert.

Elo works in MW:O and works fine. The math has been put forward repeatedly. If anyone has any actual mathematical evidence to the contrary and wants to put it forward then please do so - you'll be disproving the nature of statistical analysis and at least one, possibly three, fields of mathematical study. You'll be famous! Why not go ahead and bring that forward?

Here's a useful phrase:

The probability of observing a result given that some hypothesis is true is NOT EQUIVALENT to the probability that a hypothesis is true given that some result has been observed.

Elo works just fine in MW:O or any other game for matchmaking because the goal is to try and get matches as competitive (equal) as manageable with a small margin of error. In many games and competitive sports that's not the goal; the goal is to just determine overall who's best. That's a ladder ranking. We do not have, nor want, a ladder ranking in MW:O - that would involve letting the best players perpetually stomp the average players. It would mean that if you're even a bit above average you'd win most the time; if you're even a little below average you'd lose almost every match.

We had that before Elo came out and it was terrible and made the game unplayable for new players. Elo works - that's the problem for some people. If you're average then you need to focus on actually being a better player; not just playing more matches, to get better at playing. That's not about getting a slightly better mech. It's about figuring out what better players than you do and doing that consistently.

Which is more effort than most people are willing to put into a game, so they stay average and complain that they're not given an escalating scale of reward for their tenure.

#160 Mystere

    Member

  • PipPipPipPipPipPipPipPipPipPipPipPipPip
  • Bad Company
  • Bad Company
  • 22,783 posts
  • LocationClassified

Posted 19 October 2014 - 06:43 PM

Anybody who says "ELO" instead of "Elo" has absolutely zero credibility with me. Nada. Zilch. Zip.

They obviously do not know what they are talking about if they can't even spell it right.





8 user(s) are reading this topic

0 members, 8 guests, 0 anonymous users