Elo Is For Chess, Not Mwo
#161
Posted 19 October 2014 - 06:54 PM
the world is cruel.. if you dont learn how to play you need a beat up.. until you learn or ask how to play.
fair blablbalba we all had to learn it the hard way.. we did play with noobs and yes we did stomp noobs.
why`? bc the grp matchmaking sucked and you just went right up to the heavy boys with the heavy mechs.
this game just don´t have enough players to support an huge elo matchmaking...
and yes i went the hard way from a noob to a 600+ avg xp guy with 1.96 k/d as non teamer... tho my w/r is >50%
I just came back to check the 500$ mechs lol
*The non beta player wont even get the numbers i did adress.. but w/e realy sad that mw never made it.*
#162
Posted 19 October 2014 - 07:19 PM
Why do we not need a BV system? Because BV and Elo are correlated.
When you go into a match full of random people where both sides have equivalent Elo, you should have a 50% chance to win and a 50% chance to lose. If you bring zero or near zero BV to the match (a mech without guns or armor) you are reducing your team's chance to win. Sure, they could still win the 11 vs 12 match and give you Elo rating that you "don't deserve," but on average they would have a higher tendency to lose and thus make your Elo while bringing zero value mechs decrease over time.
The BV you bring to a match provides you the capability to win. Your wins translate to your Elo. The better equipment you bring to a match, the more likely you are to win, the more your Elo will rise. If you normally bring a Direwolf to your matches and decide to bring an Awesome instead, you will likely perform more poorly than you usually would and you might lose. But the fact that you're the type of person who considers bringing worse mechs sometimes will be reflected in your Elo.
If you only brought Direwolves all the time your Elo would have been higher, but because you're the kind of person who switches between Awesomes and Direwolves your Elo will settle someplace a bit lower. But your Elo will represent that fact and you will be matched accordingly.
It's possible that your whole team happened to do the same thing by bringing less performant mechs at the same time and the enemy team happened to all bring optimum mechs. In this case the BV balance may be way off even though the Elo appeared to be balanced. This however would be an edge case and most of the time people will bring their "average BV" that corresponds to what they used to earn their Elo rating.
If anything, you could use each chassis's win/loss rating as an Elo modifier for players, e.g. if a mech has a win/loss rating above 50% you could give the player piloting it a higher Elo when considering matchmaking. In the same way if they are below 50% you could give an Elo penalty.
You cannot assign BV to equipment because certain weapons on certain mechs make more sense, certain weapon combinations make more sense and certain weapons which appear good numerically are actually very bad in practice. As an example of each:
Small lasers on a fast moving light mech could work out. Small lasers on a direwolf are quite stupid.
A mix of medium and large lasers could work out. A mix of small lasers and large lasers is quite stupid.
The AC2 isn't that numerically inferior than an AC5, but in practice it has far less value.
Some people can also perform very well with certain weapons and other people can't. You can't necessarily say that PPCs have high BV for example because some people might not be able to hit moving targets at all with them and some people might. For those that can't who instead choose to take something they can aim, like lasers, they would indirectly benefit from the fact that opponents using PPCs are at a disadvantage due to higer perceived BV on PPCs. This would penalize people who have the skill to hit with PPCs and benefit people who cannot.
An excellent example of this is that competitive players used to all use PPCs but now they all use ERLLs because they perceived it was too hard to hit with PPCs now due to the reduced projectile speed. But what if some players could still hit fine with PPCs even at that speed? Such factors are based on mechanical skill, which is something that BV can't accomodate for. Elo on the other hand accounts for mechanical skill because mechanical skill plays into a player's win/loss ratio.
You have to rate mechs on their potential to win rather than on what they actually bring to the battle. Making a good loadout is part of player skill in this game. If you are not using your equipment (which everyone has the same access to) to its full potential it is your own fault. You should not receive a BV bonus against an opponent because you chose to take a less optimal configuration for your mech.
Similarly, since you are not weight restricted at the moment, if the fact is that heavier mechs perform better and you choose not to take the heaviest mech in the category of your choice then you are again making less optimal decisions and hurting your own Elo. However, since it is not necessarily true that higher weight equates to a better chance of winning, you could again turn to the per chassis global win/loss ratio that I referred to earlier, which would be an excellent way of determining which mechs are more performant than others and which is not as arbitrary as a system such as BV as it factors in some things like better hitboxes and higher weapon mounts which are things that BV doesn't assess.
Edited by oneproduct, 19 October 2014 - 07:57 PM.
#163
Posted 19 October 2014 - 07:39 PM
#164
Posted 19 October 2014 - 08:27 PM
Fire and Salt, on 09 October 2014 - 08:27 AM, said:
It just needs to GTFO when tourney time comes around because it unfairly penalizes good players.
What's the Matter? Can't have your cake and eat it too? Either you like Elo or you don't. This just means there will be different Elo bracket winners. That's better for the entire player base than the top 10% taking all the rewards. The top guys usually don't need the prizes anyway.
Koniving, on 09 October 2014 - 05:29 PM, said:
The first video you linked is a terrible example of counter play. First of all the sniper tanks (except for a few really terrible mid tier vehicles) really don't have limited ammunition. Left to their own devices most vehicles have enough ammo to destroy 50-60% of the team without help. They also pretty much did away with the invulnerable tank idea presented in the video. They even implemented a spotting mechanic that ruined passive spotting by scout tanks hiding in bushes and effectively remaining invisible. People knew you were there because they were told they were spotted by another mechanic.
On the other hand the link about elo and the sum being greater than the parts.... that is a great post.
Haji1096, on 10 October 2014 - 04:19 AM, said:
Unfortunately it is too left brained and focuses on the parts and ignores the big picture. This allows for one person to carry the team and have everyone else "do well" without doing anything. Elo is flawed because the game assumes you can carry no matter what mech you pilot. There are so many holes that have been posted over and over and over again. Random assortment would be better than what we have; at the very least you never had to wait more than 8-10 seconds for a match? I saw fewer stomps with random assortment than I do with a "balanced" Elo system.
There is a very easy approach to consider with BV instead. Look at the build the player is using and decide if it does more alpha or more sustained damage than the stock build. Or whatever else you want to measure. If these ratings indicate the build is better, increase his BV rating slightly. You don't need to itemize everything.... just the most powerful weapons. For example if you use Gauss add a 10% penalty since it is literally the best weapon in the game. If you use two add a 20% penalty (note increasing his BV would be a penalty).
Imagine a scenario where the best players used the crappiest lame build they can imagine and still win. Throw them against average players in supremely optimized builds and you would have an interesting fight.
Elo is completely unnecessary. Just look at the mech BV and the player's average match score and build a system from these two factors. This would mean that you could be really good and win the tournament with some middle of the road trash build OR you could win as the best of the best WITH the best build. You would see completely different environments when you play with one set of tools or the other.
#165
Posted 19 October 2014 - 08:42 PM
Glythe, on 19 October 2014 - 08:27 PM, said:
Please see my post above. Logically speaking, BV must be related to your success right? Otherwise we wouldn't measure it. Since BV is related to your success and your success determines your Elo rating the two are correlated.
How would you measure a player's skill without Elo for the "best players with low BV vs bad players with high BV?" Isn't that what we use Elo for?
Average match score is not necessarily indicative of performance. Unintuitively, lower match scores can be better because if, for example, you headshotted the entire enemy team you would have extremely low damage and no one would have assists.
In terms of your carrying/getting carried argument it too is false because it will average out over time. If you as a player are good you will tend to carry more games, even if in some games other people carry you. If you as a player have a high tendency to play mechs that cannot carry then your Elo will go down. The more you have a tendency to play better mechs that can carry, the higher your Elo rating will go. For any particular match this may not be true, but over the long term it will be, hence you saying that Elo ignores the big picture is plainly false as it is exactly the opposite. It ignores the immediate, short term picture which is what people are complaining about and think that they can magically fix. You can say that a coin will come up heads 50% of the time over the long term, but you cannot use that information to determine what the result of the next coin flip is going to be.
Your Elo will settle towards your average performance in the long term, that's what the system is designed to do. There may be bumps along the way, but like everything else it will normalize over time. That's just how statistics work. On the other hand, if you give something a BV and it turns out you gave it a bad battle value that will never self correct itself.
Edited by oneproduct, 19 October 2014 - 08:46 PM.
#166
Posted 19 October 2014 - 08:48 PM
MischiefSC, on 19 October 2014 - 06:31 PM, said:
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.
With a little bit of editing you've presented the answer to your own dilemma. There is a great deal of margin of error when you start matching pros with the average player base. This was the number one complaint of the Elo system in this game when first introduced. A lot of good people were getting complete BS allies to compensate for them being good. It is not fair for that team to be matched against an "average" team. Once again the sum is greater than the whole of the parts.
Random match making was better because the match making process was near instantaneous. If you play more games that favors the long term average that you will have more fair games overall. More games also meant more XP and CB for everyone. Let us not forget that a lot of free to play players intentionally sabotaged the armor on their mechs and ran right to the enemy to die so they could get into as many matches as possible (now a ban offense)
The old system let you play with up to 4 friends in one lance. OR you could host an entire full team to fight an entire full team. The new system says screw your friends you play alone or you get sexually violated in the group queue (to new players anyway).
Ever wonder why there aren't a lot of people playing this game? The Elo MM system is not fun; no matter how good you get you're always running in place at the same speed. The old system was fun. Most people can have a lot of fun doing really stupid/boring things if they are doing those activities with friends.
BTW I do enjoy playing as a group when you have 11 other people online. Otherwise it is a big waste of time.
#167
Posted 19 October 2014 - 08:59 PM
This is not exactly what Elo necessarily does however, as it is not the only way for a matchmaking system to work. It could try to present you with a match where the Elo ratings are not as close as they could be, but if your team has a far lower Elo rating than the other team, you will lose fewer Elo points as you were not expected to win. However you do need to be presented with these situations from time to time to determine if you could in fact beat players who currently have a higher Elo rating than you because it could be that your current Elo rating is not as high as it should be, and one of the ways to test that is to play against people rated higher than you.
Even so, our particular version of matchmaking does not go out of its way to try to create mismatched Elo matches, but when they do happen you will not be credited/penalized as much if the outcome goes the way that the Elo ratings indicate.
On the subject of the old max 4 or full 12 group queuing it might have been better as it would likely allow for closer matching, but we asked for them to let us queue with however many people we want. Yes, it probably causes problems as it makes it harder to mix and match parts, but you can't have both. The more you try to force people together and give the system less room to make changes the harder it is to make a fair match. But it was a system that we asked for and one that a lot of people use and enjoy, particularly the units.
If a group of people queue together and you have to keep them on the same team, how are you supposed to balance the parts? There simply isn't enough flexibility to find a 5 man group that complements a 7 man group in such a way that that team's Elo matches up well with the other team which may just so happen to be three lances queued up separately.
You also can't assign a value to the fact that a large group talking in a voice chat program can coordinate better. So even if you had 6 groups of 2 people vs a single group of 12 people it's hard to say how much of an advantage those 12 people will have. BV would not fix any of that. Both teams could have the exact same BV but it won't matter. The 6 groups of 2 people could even have much higher BV and still lose because the 12 man group gains much more from being coordinated. This is where the true fault in the system lies, but again you really can't do anything about it. The players who are very serious at the game play with voice chat. Even if you inflated their Elo rating to represent that how many other players will you find to match against them if they were already among the best?
What if a 12 man group queues with the highest conceivable BV possible (all direwolves, timberwolves, stormcrows, etc)? The only "fair" match would be an identical team, but if no such team exists you have to pick something else. Is it unfair to them? Maybe. But since they are the underdogs they will not lose as much Elo rating because they were predicted to lose, and this is because your Elo rating correlates with the BV you tend to bring to a match.
Edited to present arguments in a better order.
Edited by oneproduct, 19 October 2014 - 09:15 PM.
#168
Posted 19 October 2014 - 09:29 PM
oneproduct, on 19 October 2014 - 08:42 PM, said:
That's a very dangerous road to go down, but I'll give you the benefit of the doubt that the problem is with your phrasing.
Are you saying BV is not related to success and that is why we don't measure it? Are you trying to say good players win because they are good and would be just as good in a spider with a single machine gun as they would in a Timberwolf with the best build in the game?
What I was trying to say is that no matter how good the best player in the game may be... if you put him in the worst mech possible with the worst (semi viable) build possible he isn't going to excell. He will die to some average player with a build that works (the build might be great or it might be just slightly less terrible than yours).
Ever see a move or cartoon where a skilled swordsman is disarmed and utterly embarassed by someone using a stick/cane/umbrella? What you use to win matters. Elo doesn't care and this is foolish.
oneproduct, on 19 October 2014 - 08:42 PM, said:
oneproduct, on 19 October 2014 - 08:42 PM, said:
But here is where match score is indicative of performance: if you always or almost always have the best match score then you're probably better than most of the people on your team. This of course assumes that you are actually doing things to contribute to the match rather than farming match score (like laser raking for assists and carving excessive parts off AFK enemies). It doesn't matter if your average is 20 or if it is 200; it matters compared to all the other people that were in the game with you for all the games that you were involved with.
This again ties in with performance. To most people it would be more interesting to see how well a champion does when he is suddenly on a losing team. Can they only do well when everything is on their side? Or will they have near the same level of performance with zero assistance? Elo doesn't care and that ignorance reflects how it is a poor judge of skill.
oneproduct, on 19 October 2014 - 08:42 PM, said:
This might be true if you only played games alone... but just as there were when the Elo system was introduced there will always be people running around with an inflated Elo because they have friends that can carry them to victory more often then they can win alone.
#169
Posted 19 October 2014 - 09:42 PM
Quote
Are you saying BV is not related to success and that is why we don't measure it? Are you trying to say good players win because they are good and would be just as good in a spider with a single machine gun as they would in a Timberwolf with the best build in the game?
What I was trying to say is that no matter how good the best player in the game may be... if you put him in the worst mech possible with the worst (semi viable) build possible he isn't going to excell. He will die to some average player with a build that works (the build might be great or it might be just slightly less terrible than yours).
The problem is not with my phrasing, but how you read it. It means the exact opposite of what you got from it. BV is correlated (related) to your Elo. Thus you shouldn't need to measure both. If BV as a system exists with the purpose to determine that more BV should mean more success, then BV and Elo are quite the same. In the same way that Elo doesn't check your performance but merely that you won or loss, BV would do the same.
Quote
Again, these things are all correlated. If you have high accurracy you probably win more games. If you have more kills you probably win more games. All stats all lead to one thing: winning. Your win/loss ratio is the ultimate result of all the statistical things you do throughout a match.
Quote
The reverse is also true however. If you're almost always winning, you probably almost always have a good match score so again there is a correlation. The thing is that the fact you won is more general whereas match score can be gamed doing things like you said; scratching enemy mechs' paint for assist points.
Quote
This might be true if you only played games alone... but just as there were when the Elo system was introduced there will always be people running around with an inflated Elo because they have friends that can carry them to victory more often then they can win alone.
The Elo system already accounts for this. You gain or lose Elo points based on your expectation of winning. If the theoretical person that you're talking about who is bad but gets carried by his friends all the time wins matches that are lopsided when in their presence, he will be awarded a low amount of Elo points. If he plays on his own and is placed in a "fair" match with his "unfairly gained" Elo rating if he loses he will lose a large amount of Elo points because he was expected to win but is suddenly revealed to not be as good as he's been pretending to be.
Edited by oneproduct, 19 October 2014 - 09:56 PM.
#170
Posted 19 October 2014 - 10:08 PM
oneproduct, on 19 October 2014 - 08:59 PM, said:
This is not exactly what Elo necessarily does however, as it is not the only way for a matchmaking system to work. It could try to present you with a match where the Elo ratings are not as close as they could be, but if your team has a far lower Elo rating than the other team, you will lose fewer Elo points as you were not expected to win. However you do need to be presented with these situations from time to time to determine if you could in fact beat players who currently have a higher Elo rating than you because it could be that your current Elo rating is not as high as it should be, and one of the ways to test that is to play against people rated higher than you.
Even so, our particular version of matchmaking does not go out of its way to try to create mismatched Elo matches, but when they do happen you will not be credited/penalized as much if the outcome goes the way that the Elo ratings indicate.
To address the last part first... for organized play you should have a ladder. Let the best team win. Throw Elo out the window and assume every team is going to be top notch and well organized. You had that and it failed. I think in part it failed because it tended to bring out the worst in people. What you have now is a system where full teams on coms are regularly paired up against partial teams without the same advantage. And people wonder why no one wants to play in the group queue.
Over a long enough timeline random assortment is just that. It matters a lot less that you got stomped when you can play more games and get more rewards. Statistically you should be taking just as many stomps as you give. On the other hand with the Elo system if the guy who is supposed to carry doesn't then you've probably lost and will get stomped. Maybe he's having an off day maybe he is letting his little brother play or maybe he is drunk. I see a lot more stomps now with Elo than I did with completely random assortment (before even the first weight adjustments). But I suspect the extra 4 players per team are far more influential in causing a shift in the meta that leads to more stomps.
First let me say you're making a serious flaw in assuming that random match making automatically ends in a stomp. It worked just fine for a long time without problems. I think what you mean to say is with a completely random system more bad-new players get stomped. And my initial response to this is why is this a problem? If you're good you should be rewarded by winning. What ever happened to the arcade mentality? Winner stays loser pays (loser has to insert another coin). But the point is moot as bad players get stomped anyway. I support give the new players protected match making (this is also why we need single player training mode). Secluding bad players from good players means they won't see what they should be doing instead of what they are doing. They won't learn from their mistakes and instead will just at best become the "champion" of the baddie bracket. Besides when the system changed over people were WAY more upset about long queue times than getting rolled at every turn. Random got it right quite often.
I like that you try and point out that when the system fails with bad match making you don't really get punished for it. Yea you do. You've put players into a match that the game thinks they have no chance to win. Wow thanks. I think I'd just rather have random assortment with instant match making and play a lot more games every night.
In the long run Elo may keep your win rate at 50/50 but it doesn't prevent stomps. The practice of handicapping a good player with low score allies against an average team actually encourages it. With no Elo when you won you felt like you were just a little better than the other guys. If you win too much with Elo you often feel you get screwed over on team selection (especially when your match score/damage/kills is nearly identical on the loss as it is on a win). It isn't fun when Elo punishes you for winning.
#171
Posted 19 October 2014 - 10:26 PM
Randomness is what you use when you decide to not make a system. Because it is completely random, you have a high chance of having mismatched teams, thus you have a high chance of matches being stomps, even if in the long run the tendency is 50/50.
Any system you build to replace randomness is going to reduce that problem, or at least should unless the system is so bad that it actually exacerbates the problem. I highly doubt that matchmaking while attempting to use Elo could do so as Elo as a system has a lot of evidence to support that it works. Of course in some games it will work better than others. But if you take a player's win rate (which is correlated to accuracy, kills and all those other stats you want to measure) and you use that to try to make the teams a bit more even, it surely must be better than using nothing at all and doing things completely randomly.
Quote
Yes, you certainly suffer the loss which isn't fun, but you missed the point. In both random matchmaking and Elo based matchmaking you can end up in situations where you're pretty doomed to fail from the start. The difference is that Elo will try to avoid this, since it has a numerical basis to try to do so, and if it does happen anyways, you will not be as penalized as you would otherwise. The stomp would happen either way, but in the Elo version at least you don't suffer from it as much in terms of a hit to your statistics.
As seen in screenshots of the MWO matchmaking system it also tries to minimize standard deviation between the players on one team, meaning that it won't put extremely low ranked Elo players on a team of otherwise highly ranked Elo players to try to balance it. The worst player in the game won't be matched with the best player in the game to fight two average players. Whereas if you had random matchmaking this could happen quite often by fluke. Also, consider the alternative. Let's say we intentionally tried not to match high Elo players with low Elo players to try to even things out. Surely a match of the best player + an average player vs an average player + the worst player is going to be less evenly matched than the first case, right?
The real source of people's unhappiness with the system is really that large groups of coordinated players using voice chat will have a much higher tendency to win games, and no matchmaking system can account for that. I would easily agree with you that I don't think that groups should be allowed to be that big in order to encourage fairer matches, but people who play in units want to play with their friends and it is better to be able to play unfair matches with your friends than to not be able to play at all.
But then that highlights a good problem. Larger groups of friends get preferential treatment compared to smaller groups of friends who are unfortunately going to have to fight the larger groups.
By the way, I just want to say that I've been enjoying this conversation and I really appreciate that we can keep it going without just shouting incorehently at each other. Thanks for being civil with me.
Edited by oneproduct, 19 October 2014 - 10:33 PM.
#172
Posted 19 October 2014 - 10:36 PM
oneproduct, on 19 October 2014 - 09:42 PM, said:
We need to take a step back and define our terms then because BV for me means something completely different. It is a predicted measurement of how valuable any one particular mech would be in a match. This could range from a spider with 1 machine gun and 1 armor (not 1 ton of armor) to our fully loaded favorite Direwolf build. BV should exist without any determination of player skill.
BV would be largely a replacement for Elo and would serve the same purpose but be a better representation of predicted match outcomes. There is really no need to bother calculating Elo the way the game does right now. If you marked players as good/bad/ugly (ha ha just kidding.... lets try /average) based on kills damage assists and wins rather loosely you could pair that with a BV system. This would allow great players to play "bad" mechs and not be penalized for it. Basically the idea is bad player+good mech = good player +bad mech. This would also prevent the game from thinking an Awesome = Direwolf.
oneproduct, on 19 October 2014 - 09:42 PM, said:
oneproduct, on 19 October 2014 - 09:42 PM, said:
Never gonna happen because he only plays with his friend (like a lot of people do). Also that's not how Elo is supposed to work is it? Team A is roughly supposed to Equal Team B. Team A just happens to always have his friend on it. Team B in theory should be assigned some random guy who is equal to both friends (Let's roll back to the old lance queue not the elite team queue). Why then would the game be assigning Team A as the winner?
#173
Posted 19 October 2014 - 10:54 PM
And I normally solo drop.
However, over the weekend I dropped in a 2 person group for 2 hours and EVERY match was a ROFLstomp with me on the losing team.
The thing that made me think this Elo matchmaker is "broke fo sho" was when I dropped into a match that matched me in my two-man-team premade with two other random premades against.... {drum roll please}.... [Lord] 12 man group!
Aren't they a professional MWO team?
Once the teams met up the match was over in 30 SECONDS (maybe less).
ROFL....and wait for it......waaaait for iiiiit........... *stomp*!
It was NOT fun. Makes me wana stick to solo drop pugging
#174
Posted 19 October 2014 - 10:54 PM
Quote
That's exactly how I see it too. Let's step through it then:
BV = how valuable you are in a match (true according to your statement)
More valuable in a match = more likely to win (must be true otherwise it would mean that BV doesn't work)
Winning = Elo (true because Elo is based on winning)
Therefore BV and Elo are correlated.
Quote
In this you are correct. However, if you are the type of person who enjoys playing "bad" mechs it will be reflected in your Elo. If you bring bad mechs often that means that the average BV you bring to a match is lower which means that your average chance to win is lower which means that your Elo will be lower. Thus if you joyride in mechs often it will try to match you against people who have an Elo appropriate for that.
However, like I pointed out before, you cannot say this is true for any one particular match, just like how you can't know if the next coin you're going to flip is going to be heads or tails even though you know it'll be 50/50 on average. Sometimes you'll take that awesome and find an enemy direwolf. Again, my proposal was to use the chasses's win/loss ratios as an Elo modifier, which would allow you to joyride in bad mechs more easily for these individual cases.
Quote
There certainly is a difference. Let's consider a simple example though:
Let's say that we have a 1 on 1 game where each player has 100 health and your score is how much health you had left at the end of the match. If you were a perfect player your average score would be 100 since you would never take damage. If you are the worst player your average score would be 0 because you would never have any health left if you lost every game.
You could win every match with exactly 1 health left. This would be an example of very good win/loss ratio but poor performance. However, it would be extremely unlikely for you to always end up so close to death without dying. If you win all of your games, it is far more likely that you have a very good score.
If you have a very bad score but are winning it means that you are being carried. But how often can you rely on being carried? Let's say our "carried" player only ever plays with his good friends and they only ever play with him. They would all have identical Elo. But what happens when they get matched up with a team of equivalent Elo instead of stomping over easier players? Well if every player on the enemy team is actually good unlike the carried player where everyone on the team is good except for him then the carried player's team is going to start losing.
So yes, it is possible for someone to have poor performance and good win rate, but the likelihood of it would be low. Or, he's being carried by the rest of his team but when that team runs into another team where every man pulls their own weight then the carried player's team is going to lose.
#175
Posted 19 October 2014 - 11:01 PM
Diablobo, on 08 October 2014 - 08:40 PM, said:
Have you figured out the flaw with using Elo for MWO yet? It's not too hard to see....if you don't see it, then you have no business even trying to comprehend the intricacies of Elo in the first place.
Elo assumes both players have the same basic starting point. Apart from white getting the first move, chess is an extremely balanced game. Both players have the same pieces, and both players' pieces can do the exact same things.
Is that what happens in MWO? Does someone who takes a zero unlock Commando with 3 small lasers, a standard engine, and single heat sinks stack up against a fully mastered, fully optimized Spider? No. They do not. Neither do the other mechs that come into the matchmaker that are not as powerful as the fully optimized meta builds. Using Elo with a game that has such widely divergent starting positions is just a recipe for disaster. It turns out that PGI has cooked up a wonderful dish of one-sided mismatch stew.
Until PGI ditches Elo and uses some sort of Battle Value system, there can be no matchmaker that is anything other than random, luck of the draw crap. The sooner they acknowledge this fact, the sooner we can all get on to enjoying our epic well matched battles we were promised. Instead, we have to be saddled by noobs on our team who think it is fun to pilot an Atlas with a couple of med lasers and some LRMs with single heat sinks and no unlocks, while the other team has fully unlocked and optimized meta builds and pilots who know how to use them.
There will never be a decent matchmaker as long as we use Elo. It is supposed to be used for teams that have identical starting positions. As we all know, that is not the case with MWO.
This has been argued with PGI since it was first introduced. 18 months later, they still have a broken MM, and no intention of dropping the ELO system. Last patch has made the MM particularly bad.
#176
Posted 19 October 2014 - 11:07 PM
Yes, players who regularly play in zero unlock, stock configuration mechs are going to lose to people in fully unlocked, optimized builds. When they lose their Elo will go down and then they will have a higher tendency to fight against similarly imperfect opponents.
For most people the time frame where you don't have unlocks and you are in a stock configuration is low, so for a few matches you might suffer but in the long term it will be fine.
BV is not suitable for MWO. A small laser on a jenner has moderate value but a small laser on a direwolf has almost no value. A jagermech has good value, but will usually lose in a 1 on 1 fight against a firestarter because the firestarter pilot can use his mechanical player skill to outmaneuver the jagermech, which is not something you have in TT (mechanical skill).
If you want to demonize something it should be the large number of players you can bring in a group queue game. This is what makes it so hard to balance things because you have less wiggle room and large groups on voice chat have a significant advantage that can't be measured.
Also I feel the need to point out that I don't think that most people realize that MWO is a slippery slope game, or perhaps even what a "slippery slope" is. Slipperly slope means that once you start losing you tend to continue losing even faster. From the point that the enemy team kills one of your allies, your team lost 1/12 of its potential firepower. This means that your team is more likely to continue to lose. By the time a game gets to 0-4, one team has lost such a dramatically high percentage of their firepower that the game easily slides towards a stomp. There is a reason why it is so rare to have close matches.
Comebacks are and should be rare. When you're down 0-4 it's not like the 8 remaining players on the losing team suddenly get more powerful and should unexpectedly kill the 12 enemy players.
Edited by oneproduct, 19 October 2014 - 11:21 PM.
#177
Posted 19 October 2014 - 11:43 PM
oneproduct, on 19 October 2014 - 10:26 PM, said:
oneproduct, on 19 October 2014 - 10:26 PM, said:
How are we defining working? If working is defined as getting more people to play MWO then it is by definition a failure of catastrophic proportion.
I was in a 500+ member affiliation when Elo was implemented. The majority of people talking on the forum and the people I spoke to on TS agreed that the Elo system actually encouraged more stomping. From what I've seen Elo is a mechanism to sabotage your win rating as a good player and to hold the hands of bad players.
oneproduct, on 19 October 2014 - 10:26 PM, said:
Did you just say your Elo statistic wouldn't be penalized from getting stomped in a match you were predicted to lose? That's terrible considering you can't see your Elo and you actually have an advantage when your Elo is low. The very least the game could do is underrate you for the next game to make up for the last game (for putting you in an unfair match in the first place). Where is the option to opt out of unfair Elo matches. Why would the game intentionally set up an unfair Elo match. Where is my complementary bonus XP/CB for putting me in a match where you knew I was going to lose. How does that in any way make me feel better. At least with a random system I know you weren't trying to screw me over. And at the end of the day I *do* get compensation because I played more games.
oneproduct, on 19 October 2014 - 10:26 PM, said:
oneproduct, on 19 October 2014 - 10:26 PM, said:
I have no doubt that the Elo system reduces player win percentage to try and equal out to 50/50. I do have doubts that the system has significantly reduced the number of lopsided wins. I'd rather we cared more about ECM and clan mech distribution. But I'd also like to say why is it that games feel the need to balance the good/bad players. Isn't it a bit unfair to have bad players only play bad players? Shouldn't they at some point rise or fall based on their own skill? Back when things were random you saw amazing/terrible players who ruined the game with rather low frequency. This game can actually pretty boring when every game feels like a tournament of champions. Shouldn't that level of play be saved for competitions and community warfare? Why is it that companies feel the need to have players win rates capped around 50%?
oneproduct, on 19 October 2014 - 10:26 PM, said:
oneproduct, on 19 October 2014 - 11:07 PM, said:
BV is not suitable for MWO.
It is... but I dont think it is in the way that you are thinking. If you want to think about it that way you can use BV as a modifier for your Elo.
Want to lower your Elo? Don't use consumables (wish that was a thing).
Want to lower your Elo? Use a tier 4 mech (wish that was a thing)
Want to raise your Elo? Use super optimized builds like JJ+ 2 energy+ballistic builds (wish this was a thing)
Want to raise your Elo? Use clan mechs or gauss or both for an even higher value (wish this was a thing)
Yes the conversation has been quite good but I think we've run away with this thread.
Edited by Glythe, 19 October 2014 - 11:51 PM.
#178
Posted 20 October 2014 - 12:24 AM
Quote
Yet, it will even out. You will have as many games that end in a stomp as games that don't end in a stomp. The idea is to use a system to try to reduce the number of stomps because having an even number of stomps is not a good thing when you can aim to have fewer stomps.
Quote
It's not like this is the first game to use Elo. It's not like they invented their own system from scratch, though they, and many games that use an Elo system, may decide to use what the system is telling them differently. Starcraft 2, Dota 2, LoL and CS:GO all use Elo and these are on the short list of the most competitive games out there.
It's entirely possible that they still have some problems with it, but it is very likely that even with problems that it is better than completely random matchmaking.
Quote
I was in a 500+ member affiliation when Elo was implemented. The majority of people talking on the forum and the people I spoke to on TS agreed that the Elo system actually encouraged more stomping. From what I've seen Elo is a mechanism to sabotage your win rating as a good player and to hold the hands of bad players.
Yes, a fair matchmaking system is designed to sabotage your win rating. If you were an above average player and random matchmaking was used, then on average you would fight against players who are worse then you and you would win far more than you lose.
This might be fun for that above average player but it is not fun for the loser and it may be boring for players who are more interested in a challenge than about winning. It will try to match you against equally skilled players so that you will only win about half the time.
You have to understand that this is the basis of matchmaking. Even if you are a good player, you're intended to lose every other match you play on average. People don't like this and would rather win all their games but that is a perception problem, not a problem with the matchmaker.
Also, if people left because of a perceived problem with the system that doesn't exist (I'm not saying that it necessarily doesn't exist however) then again it is not the fault of the system. Keep in mind that for all you know if they kept the random matchmaking with the newer 5+ player group queues that it could have been much worse.
People don't like change and tend to remember "the good old days" as better than they were. The good old days didn't have to deal with 5+ player group queues that we have to deal now and larger teams are essentially shifting the matchmaking towards organized stomping just by the virtue of them being a team. Again, this is something that any matchmaking system would have problems with.
Quote
Because sometimes there simply aren't players of equal skill for the matchmaker to make a match with. Let's say there were only two 12 man teams of unequal skill that played MWO and nobody else at all. You're asking for an option that would simply never put you in a match instead of at least getting to play, even if it would be a stomp. As a hint, the longer you are searching for a match the worse the match is likely to be because the fact that it is waiting so long means that it can't find a suitable match for you so it will slowly start to consider less suitable matches.
Quote
This is stil better than having completely random matchmaking! How likely do you think it is for a highly skilled player to do something incredibly stupid? It is far more likely for a random player to do something stupid. And yes, good players can have bad days but this is completely outside the matchmaker's control. If you used BV someone using a high BV mech could be having an off day as well. It is better to gamble with an educated guess such as Elo than to gamble completely randomly.
Quote
Because if you don't have that 50% target then the result is the good people would have high win rates, like 75% lets say, and the other majority of players would have extremely low win rates. They would be stomped so bad they wouldn't even want to play. Then once they quit, who's left? Only the good players who will then be forced to play against each other and then you'll go back towards the 50% goal anyways. You can only win by making other people suffer. You can't ask people to suffer more than you do, so we aim for 50% so that the suffering is shared equally.
Quote
Yes, it is their fault for not having voice chat, no questions there. And yes, I completely agree with you that the ability to play in large groups ruins the game for small groups of people. People say that if they couldn't play in their large groups that they would quit, but I'm pretty sure people could make do with running groups of 4. But that's for another discussion.
Quote
Want to lower your Elo? Don't use consumables (wish that was a thing).
Want to lower your Elo? Use a tier 4 mech (wish that was a thing)
Want to raise your Elo? Use super optimized builds like JJ+ 2 energy+ballistic builds (wish this was a thing)
Want to raise your Elo? Use clan mechs or gauss or both for an even higher value (wish this was a thing)
This is a fine concept except that it is impossible to assign BV in a game like this that has so much mechanical skill involved. BV only exists in the MW TT game and in the TT game there is no mechanical skill. I pointed out before why BV doesn't work and you haven't refuted those points yet. In summary, there are at least four problems, shown by these examples:
1. A small laser on a jenner has value. A small laser on a direwolf has practically no value.
2. A mix of medium and large lasers has good value. A mix of small and large lasers makes a lot less sense.
3. In TT PPCs hit or miss based on a dice roll. In MWO they hit or miss based on mechanical skill. For people with bad aim PPCs would have a bad return on investment. For good players they'd be penalized for their ability to aim well with it. BV doesn't account for mechanical skill.
4. A jagermech with 6 MLs has more BV than a firestarter with 6 MLs but the firestarter would likely win that 1 on 1 fight due to mechanical skill used to outmaneuver the jagermech.
So again, for the last time, if you want to modify a player's Elo a much better option would be to use the "potential goodness" of a mech. It is very easy to determine this potential goodness by looking at that mech's win/loss ratio as determined by the average stats of everyone who has ever piloted it. If the JM6-S wins 60% of the time, then give the pilot riding it a slight Elo boost.
You cannot assign BV when the usefulness of the weapon is determined by player skill. In TT everyone rolls the same dice to see if they hit. It will always do a fixed amount of damage rather than lasers in MWO having their damage dependent on how long you can keep your beam on target and missile weapons having their damage dependent on how many of them you can actually get to land on your target.
Edited by oneproduct, 20 October 2014 - 12:33 AM.
#179
Posted 20 October 2014 - 12:25 AM
Glythe, on 19 October 2014 - 08:48 PM, said:
With a little bit of editing you've presented the answer to your own dilemma. There is a great deal of margin of error when you start matching pros with the average player base. This was the number one complaint of the Elo system in this game when first introduced. A lot of good people were getting complete BS allies to compensate for them being good. It is not fair for that team to be matched against an "average" team. Once again the sum is greater than the whole of the parts.
Random match making was better because the match making process was near instantaneous. If you play more games that favors the long term average that you will have more fair games overall. More games also meant more XP and CB for everyone. Let us not forget that a lot of free to play players intentionally sabotaged the armor on their mechs and ran right to the enemy to die so they could get into as many matches as possible (now a ban offense)
The old system let you play with up to 4 friends in one lance. OR you could host an entire full team to fight an entire full team. The new system says screw your friends you play alone or you get sexually violated in the group queue (to new players anyway).
Ever wonder why there aren't a lot of people playing this game? The Elo MM system is not fun; no matter how good you get you're always running in place at the same speed. The old system was fun. Most people can have a lot of fun doing really stupid/boring things if they are doing those activities with friends.
BTW I do enjoy playing as a group when you have 11 other people online. Otherwise it is a big waste of time.
Random matchmaker was not better - it was horrible. I say this as someone who won almost 90% of matches for the last couple of months before Elo came in.
There are more people playing now than prior to the new mm. Look at tournament populations. Also a ton came back with Clans and the group queue.
Do you understand that your complaint comes across as the game is only fun for you when you get to play against less skilled people than yourself or you can drop with a numerical advantage (12man) and skew your odds of winning, not by your skill but by playing against less competent opponents or because you're in a 12man in the group queue occasionally getting to play against an aggregate of 2man teams and the like; effectively replicating the worst examples of pre-Elo game experience?
Random matches were almost always stomps. Close matches were rare and far between. I could drop with 2 friends and the 3 of us would win 18 out of 20 matches, the two we lost because we ran into another 3man team. Same sort of thing dropping solo; it wasn't hard to get a ton of kills and damage because half the other team was window licking mouthbreathers. Half of my team was terrible too; the key was to kill the other teams terribads before anyone competent on the other team killed all your own terribads. There wasn't really a challenge to the game if you were even a smidge over average. There were top tier players who'd get, literally, a 100 game winning streak. We're talking solo and 2man play there too, this was back in 8v8 with a max of 4.
That's not just bad it's horrible. It was bad and it was bad for retention.
As I've said, the main problem some people have with Elo is it works - you play with and against people around your own skill range, though there's a couple of people in every match who are going to be significantly above or below your Elo.
That's because the vast majority of players are about average and people way down either end of the curve are going to spend most games filling in on teams full of mostly average players. That's the sort of place where the MM needs to put the top and bottom tier folks together. If there was enough people on either end of the curve to populate their own matches that'd be great -
there isn't though. A wider player base will help but that will effectively just push the top and bottom further down the curve but the same problem still exists. Most players are average or close to it. High/low Elo scores are going to need to fill in with mostly average players. The difference is with Elo it's not random scrubs getting chewed by people 2% up the curve; it's reasonably balanced for everyone EXCEPT the top and bottom 5% or so, who are going to end up in oddly stacked matches by dint of being so far up or down the curve.
Need to match tonnage, weight classes and approximate Elo plus game mode. That cuts available folks a lot of ways.
#180
Posted 20 October 2014 - 12:56 AM
MischiefSC, on 20 October 2014 - 12:25 AM, said:
I always thought that my W/L is so bad because I did play so few games in the past but the ELO could be indeed the difference. Indeed I'm matched most the time vs players whose experience is the same as mine + my defiance vs meta builds.
AHA - there is the meta build and the stock build- so we need BV?
No we don't - the choice not to use a min maxed Mech can have effect on my ELO - it may drop - maybe 20-50 points because i don't think that the build of a single player has so much impact on a game.
We don't even need weight classes nor map or game mode selection - but better victory conditions for the battles.
Simple reason:
hypoteticaly player Timmy drops all the time with his AC 40 Jaeger on Forest Colony with Skirmish game mode. His ELO will rise - the BV of his Mech is well known. He becomes a valuable asset on this map - and BV + ELO proof that.
But one day Timmy thinks its time to drop on another map with a TimberWolf.
he doesn't know the map, he has a High ELO for heavy Mechs, he doesn't know his Mech - he will play like a newb. And again although everything seems to be fine - people start to complain that the MM sucks because they have to play with Noobs.
ELO works
1 user(s) are reading this topic
0 members, 1 guests, 0 anonymous users