91 replies to this topic

[Znail]

Odins Fist, on 31 January 2013 - 03:04 PM, said:

.
Hmmm.... So being it's in the future, how is it a problem already..??

It's a simple truth that it's easier to get the devs to change their mind about an idea they have yet to implement, then to try and get them to change something already in the game. This is even more true for something like an ELO system that can't just be changed without issues.

[Tikkamasala]

Aidan McRae, on 31 January 2013 - 03:00 PM, said:

but the most important thing a MechWarrior accomplishes in a match is a Damage Inflicted v. Damage Sustained ratio, or even simplify it to strict Damage inflicted.

Why do you think 1/(Damage Sustained) is a good measure for ones prowess in mwo? Soaking a lot of damage all over your mech should be preferable to just getting cored. Catching the bullet for a heavily damaged teammate seems to be a good idea as well. In your ratio both situations lead to a smaller value.
Damage inflicted could be an okay extra information for your elo, although just spreading the love and not killing mechs or destroying critical parts doesn't accomplish anything.

[Tikkamasala]

Yoseful Mallad, on 31 January 2013 - 03:18 PM, said:

The private match set up can really help and or hurt a team too. Lets say my 8 man team has a really high ELO rating and we decide to invite a much lower ranked team to a private match... not because we want to stomp a lower ranked team but maybe because the guys in that other lower ranked team are good friends of our and we want to have a friendly battle... regardless of ELO ranking. Well... if they beat us, their ELO ranking should jump up quite a bit and ours should drop quite a bit. If we beat them... ours may go up a few points but theirs should not go down all that much either.

If we can start custom matches our ELO won't change. At least that's how i would code it. Otherwise people could transfer ELO ratings too easily between teams.

Edited by Tikkamasala, 31 January 2013 - 03:26 PM.

[anonymous175]

[p4r4g0n]

Paul Inouye clearly states in the post that "One variable that is set by winning or losing is the WinFlag" which seems to imply that there will be more than one variable used in the calculation of score. Hence, it will not be based on just win / loss.

I believe that the main gist of your post is that based on the numbers provided in the example it may be very easy for a player to drop from 1300 to 0 particularly new players and take a while to climb back up to 1300. However, I'm not really sure I see the problem per se.

The likely trend I would expect to see (at least initially) is that new and (most) casual players who solo PUG are going to be clumped below or close to the 1300 level. The more serious / skilled players who solo PUG will be above 1300 as well as those who mainly premade (whether casual or serious).

In the long run, new players and below average players (due to technical / inherent disadvantages) will be in the lower band, most average players will be in the middle band and the better players in the upper band.

Any player wanting to play against the higher rated players will either have to up his / her skills if mainly solo PUGging or join a premade.

To sum up, solo PUGs will by default get the (almost) solo PUG queue they want mixed with some pretty bad premades (skill wise) while the premades will get the competitive matches they want.

As far as griefing is concerned, there's not much that can be done to stop anyone from smurfing other than what already exists currently. Better just some asshats griefing than having hugely mismatched battles i.e. stomps which appear to happen very frequently currently based on anecdotal evidence.

Assumption:

Generally, Premades win between 50+% (lowest I've seen on forum is actually 65%) up to 80% - 90%. Conversely, solo PUGs win about 30% - 40% of the time.

Figures derived from forum posts and is broadly consistent with my own experience solo PUGging and casual grouping.

P.S. The main issue I expect to occur due to ELO is the intentional losses through botting, suiciding, afk for smurf accounts created specifically to average down team ELO scores.

Edited by p4r4g0n, 31 January 2013 - 04:21 PM.

[Roadkill]

Znail, on 31 January 2013 - 02:14 PM, said:

Now to the main problem with this system. 25 points for each match means that 10 more losses then wins gets you -250 points and that is a massive change in ELO. Only around 52 losses more then wins are needed to go from noob rating (1300) to a zero one.

The main problem seems to be your comprehension of how Elo and the matchmaker work.

Every team you join will have its own Elo rating computed. If you are really bad, your team will continually have a lower and lower Elo rating, thus making the team you face easier for you to beat. To actually lose 50+ games in a row would mean that you actually deserve to have a 0 rating - you just plain suck.

Similarly, if you keep winning and your Elo rating keeps going up, your teams are going to face better and better competition. Winning 10 games in a row, even at the start, will be a pretty amazing accomplishment. You will need to be a god among mere MechWarriors to accomplish that because one of two things will be happening at the same time - either the enemy team is going to have a player on it who is winning constantly just like you, or your team is going to have a (different) gimp on it every time who is constantly losing in order to balance out the team's Elo rating.

So while what you say sounds totally plausible... it isn't. Elo ratings are self balancing and are recalculated after every match.

[Znail]

p4r4g0n, on 31 January 2013 - 03:52 PM, said:

Paul Inouye clearly states in the post that "One variable that is set by winning or losing is the WinFlag" which seems to imply that there will be more than one variable used in the calculation of score. Hence, it will not be based on just win / loss.

I believe that the main gist of your post is that based on the numbers provided in the example it may be very easy for a player to drop from 1300 to 0 particularly new players and take a while to climb back up to 1300. However, I'm not really sure I see the problem per se.

The likely trend I would expect to see (at least initially) is that new and (most) casual players who solo PUG are going to be clumped below or close to the 1300 level. The more serious / skilled players who solo PUG will be above 1300 as well as those who mainly premade (whether casual or serious).

In the long run, new players and below average players (due to technical / inherent disadvantages) will be in the lower band, most average players will be in the middle band and the better players in the upper band.

Any player wanting to play against the higher rated players will either have to up his / her skills if mainly solo PUGging or join a premade.

To sum up, solo PUGs will by default get the (almost) solo PUG queue they want mixed with some pretty bad premades (skill wise) while the premades will get the competitive matches they want.

As far as griefing is concerned, there's not much that can be done to stop anyone from smurfing other than what already exists currently. Better just some asshats griefing than having hugely mismatched battles i.e. stomps which appear to happen very frequently currently based on anecdotal evidence.

Assumption:

Generally, Premades win between 50+% (lowest I've seen on forum is actually 65%) up to 80% - 90%. Conversely, solo PUGs win about 30% - 40% of the time.

Figures derived from forum posts and is broadly consistent with my own experience solo PUGging and casual grouping.

P.S. The main issue I expect to occur due to ELO is the intentional losses through botting, suiciding, afk for smurf accounts created specifically to average down team ELO scores.

As for problems so is one that random luck has too much an impact on your rating if the gain/loss is too large. If your ELO can vary from day to day with several 100s just due to the daily luck, then it wont be a very accurate measurement of your skill, wich will make the entire system rather pointless.

Another problem is that the natural trend for new players will be to lose ELO. They will be new to the game and playing trial mechs, so it will only be natural for them to plumet in rating. But eventually so will they learn the game and get their custom mechs in order. But they are now likely to be surrounded by similar players that should be heading up in the rankings. So one side effect of lettings players go down in ELO is that 1000 ratings matches might be harder then 1300 ones.

Roadkill, on 31 January 2013 - 04:03 PM, said:

The main problem seems to be your comprehension of how Elo and the matchmaker work.

Every team you join will have its own Elo rating computed. If you are really bad, your team will continually have a lower and lower Elo rating, thus making the team you face easier for you to beat. To actually lose 50+ games in a row would mean that you actually deserve to have a 0 rating - you just plain suck.

Similarly, if you keep winning and your Elo rating keeps going up, your teams are going to face better and better competition. Winning 10 games in a row, even at the start, will be a pretty amazing accomplishment. You will need to be a god among mere MechWarriors to accomplish that because one of two things will be happening at the same time - either the enemy team is going to have a player on it who is winning constantly just like you, or your team is going to have a (different) gimp on it every time who is constantly losing in order to balance out the team's Elo rating.

So while what you say sounds totally plausible... it isn't. Elo ratings are self balancing and are recalculated after every match.

While the quality of your opponents will go up with higher ELO so will the quality of your team mates do the same. The main issue will always be if you have luck on your side. It's not even only about whom you have on your team, but it's obviously different if a Mechwarrior is bringing his ECM Atlas or some random mech he just bought. So the tenth match may not be any harder then the first win in the streak.

The main problem is that with a team based game like MWO so can't we expect every single match to be an accurate measurement of your personal skill and thus we need to use large number of matches to get any kind of accuracy. It would actually be preferable if they started to count ELO with all the matches we have played so far even. That would help some with the sample size and make it less needed to have a large gain/loss just to start to spread out the field.

Edited by Znail, 31 January 2013 - 05:12 PM.

[p4r4g0n]

Znail, on 31 January 2013 - 05:10 PM, said:

As for problems so is one that random luck has too much an impact on your rating if the gain/loss is too large. If your ELO can vary from day to day with several 100s just due to the daily luck, then it wont be a very accurate measurement of your skill, wich will make the entire system rather pointless.

Another problem is that the natural trend for new players will be to lose ELO. They will be new to the game and playing trial mechs, so it will only be natural for them to plumet in rating. But eventually so will they learn the game and get their custom mechs in order. But they are now likely to be surrounded by similar players that should be heading up in the rankings. So one side effect of lettings players go down in ELO is that 1000 ratings matches might be harder then 1300 ones.

When I solo PUG, which is most of the time, I already accept that I am at the mercy of what seems to be a fairly effed up MM that keeps throwing up opposition teams that are stacked with apparently more highly skilled players. While random chance still applies with ELO, I will be matched with players on both sides that are relatively comparable to myself.


I'll take that any day over what we have now.

As far as 1000 rated matches being harder than 1300 ones, that can't happen unless the MM is borked and no one from the 1000 ratings group is moving up. I think you're missing the point that it will never be static, there will be people moving up and down all the time and I think it will be really odd to see people whose ELO don't change at all over any significant period.

People may plateau out at some level but they will still move within a range and it may be a very broad range depending on whether they're playing new mechs, seasonal active playerbase variation, hours played, etc

ELO is not about player skill measurement per se, it is about helping to match you against comparable players. It may not be perfect but if it cuts down on the roflstomps, it will go a long way towards boosting player retention.

Edited by p4r4g0n, 31 January 2013 - 05:49 PM.

[Znail]

I think you have a bit too high hopes for the ELO values to be perfectly exact and telling about who will win.

But back to the important subject of this thread. Isues with the proposed ELO system. Another problem with alowing players to go down to low ELO rankings comes with the option pruposefully adding one player with really low points to your team to cheat the system. Consider 3 players of 1600 ranking that has one friend playing with 400 ranking forming a Lance, should they really be considered an average of 1300 in quality?

[TB Freelancer]

ParasiteX, on 31 January 2013 - 02:39 PM, said:

I don't think they will base Elo rating from just wins and losses.

They will be releasing a new combat score system before Elo. Which i assume will play a role in affecting you're Elo ranking.

So even if you lose, your elo may still go up, if you performed well in battle. Or at the very least, you wont lose as much of an Elo rating.

We wont know for sure until they release it. Until then. It's all speculation.

Yeah I'm not even sure if the combat score is part of the Elo system, separate from it, or if it actually is your Elo rating. PGI hasn't exactly been clear on that, and I see them making the same mistakes I used to make as a rookie when I wrote my first application on a professional level. They use terms or language that makes perfect sense to them, but us laymen just don't get what the hell they're talking about. Its an easy mistake to make, and I'm sure they're becoming much more conscious of it with every flame that their announcements sprout.

As it stands. I've got some very serious reservations about how limited Elo will be from what official posts I've seen. I strongly suspect it it will be a baby step in the right direction requiring months/years of collecting data and altering the system.

I had a long rant about how the shorthand of playing with the same guys is extremely powerful but I'll shorten it to this.....

As a Team Banzai member who's played countless competitive matches, I instinctively know who's taking the lead, who's out back, who's on my left, who's on my right right and in most case I know who is in each position, because thousands of matches produced a repetition that resulted for calls on coms being answered with "Already saw it, I'm on him and that ____ is one or two shots from being torn off!".

I and all the guys I play with (with only a few newcomers are veterans of thousands upon thousands of matches who simply instinctively know what the next guy is doing. We've been playing as a group for over a decade. Elo simply can't account for the power nor the shorthand we have as an organized group with the benefit of knowing how the next guy is going to play. We all have our rolls and fill them accordingly......

......I had a long rant written but this (for now....this is one of the few threads I'll revisit after posting.....because at this moment II'm half in the bag (drunk and about to #%#@ the wife until she can't remember her name))

[Kristov Kerensky]

First off, I share the thought that an Elo based system for ranking in a team game is a bad idea, due to my own personal experiences with it in other team based games and how it doesn't work. Elo is for determining ranking in Chess, 1v1 competition with a very strictly defined set of parameters. It's hard to game Elo for Chess due to how it's monitored, sudden shifts in Elo are rather noticable and attention is immediately payed to that situation to see what's going on, because people don't game their Elo in Chess, newbstomping isn't a thing in Chess after all. Games that are team based and use an Elo based ranking system, gaming the ranking is done all the time and it's hard to stop, if not impossible in most of the games using it currently.

As for the system PGI is working on and the problems it MAY have..we don't know yet, PGI doesn't know yet, they aren't even sure exactly how they are going to determine rankings yet. Paul mentions that they'll start with a basic win/loss and then modify that based on tonnage and go from there. We'll have to wait and see what they end up with, but I personally don't hold much hope it'll be any better then the POS's already in use in other game currently.

[One Medic Army]

Anyone else wondering if they'll be using the massive backlog of performance/match data they already have to give people ELO scores at the onset?

If it were me putting together the ELO system, I'd be running it right now, and comparing the win/loss performance estimated from the ELO system to what is actually occurring, using that to fine tune the system before it gets integrated into the matchmaker itself.

As to people gaming their performance, it'd be pretty easy to see if each match a person had an individual score (such as the new end of round screens will supposedly do) and then just have the computer look at players who have a high standard deviation in said scores. It may indicate running bots to lose on purpose to lower ELO, followed by bouts of pubstomping.

Edited by One Medic Army, 31 January 2013 - 08:41 PM.

[WarZ]

Another thought as to why you dont like this upcoming system is it will take away part of your ability to pug stomp and roll up stupidly good statistics. You will now be playing more "in your league" vs the other team. It wont be a group of 4 to 8 (im looking at you sync droppers), rolling a weak team of mostly noobs and/or the uncoordinated any more.

So instead of feeling like a superstar you are going to be fighting hard for your wins and kills vs just having them drop at your feet due to enemy ineptitude.

This mildly occurred to me and i realized sure my stats will go down a little, BUT it will be far far better than driving away new players to game. I'd rather play a competitive match. The truly epic stuff you remember is when you played good people and bad odds and still came out on top. Too many people get stat ' itis and forget that.

And yes I have ridiculously good stats for this game (over 10 to 1 KD and massive wins) even after almost 2,000 rounds, and that being a front line fighter that pugs 30'ish % of the time. And even though I generally smash most up, I only truly remember the awesome fights and taking on the good players and the good teams. Those make you feel awesome on the win.

So stop panicking that you are going to lose your "percieved awesome". Which is mostly bolstered by a weak system that lets you beat on poor teams.

[p4r4g0n]

Znail, on 31 January 2013 - 06:20 PM, said:

I think you have a bit too high hopes for the ELO values to be perfectly exact and telling about who will win.

Our discussion is purely theoretical and as I've said, it need not be perfect it just needs to eliminate the extreme mismatches of teams for a start. Whether it works better than that depends on implementation and that's entirely in PGI's hands.

Znail, on 31 January 2013 - 06:20 PM, said:

But back to the important subject of this thread. Isues with the proposed ELO system. Another problem with alowing players to go down to low ELO rankings comes with the option pruposefully adding one player with really low points to your team to cheat the system. Consider 3 players of 1600 ranking that has one friend playing with 400 ranking forming a Lance, should they really be considered an average of 1300 in quality?

Previously commented here in context of CW if / when. Also, we really have no idea if PGI will just use a simple average to determine a group's ELO, include modifiers / weighting to each group member's ELO based on other factors or even a modifier just because it is a group.

Edited by p4r4g0n, 31 January 2013 - 09:34 PM.

[Kristov Kerensky]

WarZ, on 31 January 2013 - 09:07 PM, said:

Another thought as to why you dont like this upcoming system is it will take away part of your ability to pug stomp and roll up stupidly good statistics. You will now be playing more "in your league" vs the other team. It wont be a group of 4 to 8 (im looking at you sync droppers), rolling a weak team of mostly noobs and/or the uncoordinated any more.

So instead of feeling like a superstar you are going to be fighting hard for your wins and kills vs just having them drop at your feet due to enemy ineptitude.

This mildly occurred to me and i realized sure my stats will go down a little, BUT it will be far far better than driving away new players to game. I'd rather play a competitive match. The truly epic stuff you remember is when you played good people and bad odds and still came out on top. Too many people get stat ' itis and forget that.

And yes I have ridiculously good stats for this game (over 10 to 1 KD and massive wins) even after almost 2,000 rounds, and that being a front line fighter that pugs 30'ish % of the time. And even though I generally smash most up, I only truly remember the awesome fights and taking on the good players and the good teams. Those make you feel awesome on the win.

So stop panicking that you are going to lose your "percieved awesome". Which is mostly bolstered by a weak system that lets you beat on poor teams.

So...WHO is waving their epeen around again? You really think an Elo system will stop pugstomping? Have you played the games that use such a ranking system? I'm gonna guess no, cause that is a known issue with Elo based systems, it's SO easy to game it and pugstomp all you want..so..yeah...

[WarZ]

I cant help my e-p is so pretty. Still doesnt change the reason why some, possible most are scared of elo.

[Taizan]

ELO is not supposed to be the solution against "pugstomping" - I assume by pugstomping you mean a team of 4 or 2x4 synced players dropping into pug matches - it's purpose is to have more evened out matches for random players and especially in 8vs8. I got totally frustrated when I PUG as soon as I see 2-3 trial mechs or later in spectator mode people looking up into the sky or walking against a rock because they do not understand the torso is seperate from bipedal movement. Personally I prefer to be matched up with players who are more experienced and around +/- 100 - 200 lets say my own "win chance level".

[Protection]

The problem I see with ELO is that MWO only has a few thousand players. It's a great start, but it's not exactly DOTA or LoL. So we might end up seeing really repetitive matchups, especially among the teams that play the game more intensely, who are too experienced and practiced and coordinated to reasonably play against Random Pug Groups, but there aren't many other similar teams online at the moment.

[Wraith05]

I don't think there will be any issue with the ELO. It is meant to match you up with people that have had similar game experiences. If you lose a lot, you'll fight others that lost just as much. If you win a lot, you'll fight others that win just as much. So the ELO will be working as intended.

As for peoples perception of it being meant for their epeen, i'm sorry but that's not what it's intended to do. A 600 isn't meant to show that player as bad or new. it is meant to match up to other 600 level players. We just view it as they suck.

Edit: Also it isn't meant to garuntee you wins, but make your matches more close instead of one sided.

Edited by Wraith05, 01 February 2013 - 09:19 PM.

[Znail]

Protection, on 01 February 2013 - 08:47 PM, said:

The problem I see with ELO is that MWO only has a few thousand players. It's a great start, but it's not exactly DOTA or LoL. So we might end up seeing really repetitive matchups, especially among the teams that play the game more intensely, who are too experienced and practiced and coordinated to reasonably play against Random Pug Groups, but there aren't many other similar teams online at the moment.

This was a large problem in HoN, that has a lot more players then MWO as well. The thing is that they few at the top of ranking couldn't get random matches at all as the odds that a two full random teams of that level tried to match up at the same time was too low. The only way to get a "random" high level match in the end was to do a 'sync drop' of two full teams. This of course ment that most of the time so would high level players be playing smurf accounts, wich made one wonder what the idea behind ELO is?

Another killer issue is one that would most likely have made me stop playing the first day, if they had an ELO back then. I understand that lots of people are looking forward to not having to play in the same team with as many bad players in the future. But consider those new people who just join the game. They have to look forward to first playing the game with Trial mechs, watching their ranting plumet and then slowly try and crawl their way back up again to eventually, maybe get a match where not everyone on their team sucks. How is that for 'fun' introduction to the game?