277 replies to this topic

[Vassago Rain]

Fyrerock, on 18 February 2013 - 09:23 AM, said:

I wonder how many of the average pre-mades that are now blaming there losses on the pugs, are still going to do that when those pugs are doing more damage then them.

All of them.

[SpiralRazor]

Vassago Rain, on 16 February 2013 - 09:33 PM, said:

Illustrated.

Scrub tier ghetto will be where most players are.....oh, wait..hmm...


Sorry, but ive played LoL for 4 years...i know this to be a truism. People with 500 wins that still dont know basic laning principles etc etc.

[Mr 144]

And conversely, the days of the wonderful 1200 damage screenshots with everyone else sub 200 will probablly vanish as well...unfortunately I liked my screenies...

Mr 144

[Vassago Rain]

Mr 144, on 18 February 2013 - 09:32 AM, said:

And conversely, the days of the wonderful 1200 damage screenshots with everyone else sub 200 will probablly vanish as well...unfortunately I liked my screenies...

Mr 144

I don't know. Those are usually against gatherings of golds.

[Mr 144]

Vassago Rain, on 18 February 2013 - 09:35 AM, said:

I don't know. Those are usually against gatherings of golds.

Doesn't matter, I probablly won't be placed with ALL of them at once anymore...geesh, this is turning into a team game, wth...

[MaddMaxx]

Did the Dev post these somewhere?

Quote

The ratings are on a scale of 0 to 2800 and you started with a rating of 1300

[WVAnonymous]

EJT, on 17 February 2013 - 12:27 AM, said:

Great post, OP. This is what I predict too, and I hope it all works out this way. Fingers crossed.

Also FYI, it's not "ELO" it's just "Elo". Elo is a guy's name, not an acronym. People get that wrong all the time.

Absolutely thank you for posting this. I never would have looked it up and the concept is powerful and simple at the same time. I wonder how PGI is mathematically structuring the implementation for this in a continuous multiplayer environment, since the original relative rating concept is intrinsically tied to 1 on 1 games. Time to surf the forum, I suppose.

[Caviel]

Some of the "other factors" that were mentioned are mech weight class and queue wait time. I would presume that means the longer you are searching for a match, the less particular MWO is going to be about finding ELO compatible people.

[Wintersdark]

WVAnonymous, on 18 February 2013 - 10:16 AM, said:

Absolutely thank you for posting this. I never would have looked it up and the concept is powerful and simple at the same time. I wonder how PGI is mathematically structuring the implementation for this in a continuous multiplayer environment, since the original relative rating concept is intrinsically tied to 1 on 1 games. Time to surf the forum, I suppose.

They're implementing it largely in the same way, calculating team Elo rating as the average of it's members, and applying the results as such.

Each player is rated based on the performance of his team. While this could be grossly inaccurate in a single given battle, consider that each player is 12.5% of his team. Each player has a substantial impact. As such, while your initial rating can vary wildly, it'll work out eventually.

Because Elo rankings are being preseeded right now, you should be placed with other players with a comparable win/loss ratio. If it turns out you're better than your ranking suggests (or you improve), your impact on the battle will lead to a somewhat higher frequency of wins for your team, compared to someone else in your spot who's just not quite as good.

Thus, as you're playing with different people all the time, each of them gains a very small Elo ranking bonus per win - as do you - but you get more wins. This pushes you into higher ranked teams for later games.

So, it does work out. You just have to look at the big picture, over hundreds of games, and with the knowledge that you're much more often than not going to be placed with roughly comparable players.




Another way to look at it: Lets assume, for some reason, you're really unlucky and get repeatedly placed with players who can't fight their way out of a paper bag, and somehow end up playing at the 1000 level (while you in fact are a 1400 player). Now, this is extremely unlikely, but theoretically possible.

The reality is that you're on a team with 7 1000 ranked players and you, ranked 1000 but actually 1400. You're fighting 8 other 1000 ranked players. Which team is more likely to win? It's not like now where you can be a great player on a terrible team, facing an overall average or great team. The battle you're in here is two roughly evenly matched teams, with one team having a crazed killing machine on it capable of soloing the entire opposing team.

You'll win LOTS of those matches, and very quickly climb out of the wrong bracket, while the other players on your teams are only winning the odd match, so their ranking doesn't change dramatically as they lose immediately when facing teams that suck less than them.

[ObsidianSpectre]

Mr 144, on 18 February 2013 - 09:32 AM, said:

And conversely, the days of the wonderful 1200 damage screenshots with everyone else sub 200 will probablly vanish as well...unfortunately I liked my screenies... Mr 144

Maybe it's just me, but I hate those games, win or lose. It's usually what gets me to give up pugging for the night.

[Comguard]

Whatever stats they are getting, they are messed up.

ECM. Premades. Disconnects and the resulting lopsided matches.

[Mackman]

Comguard, on 18 February 2013 - 01:58 PM, said:

Whatever stats they are getting, they are messed up.

ECM. Premades. Disconnects and the resulting lopsided matches.

But all of these happen to all people equally (don't know what they're doing with pre-made Elo, though). For every match that your team has a DCer, you'll play a match where the other team has a DCer. That means that the numbers will match up.

And using ECM as an excuse for your poor performance (or anyone's poor performance) is just sad at this point.

[80Bit]

As I have thought about it more, there are two things that have been brought up in this discussion that do concern me a little.

1) Most players will earn the same number of average C-Bills per match, instead of better players earning more.

A few people pointed at that after Elo, most players will be earning around the same C-Bills per match instead of more skilled players earning more. This is because right now, two things give good and great players higher C-Bill earnings. More wins, and more kills/damage per match. After Elo, great players will face other great players, and thus won't have those higher win rates or the kills/damage per match they used to, and thus will make fewer C-Bills. While this is not a big deal to me personally, the argument can be made that higher skill should equate to higher C-Bill/XP gain.

2) Elo manipulation can be profitable.

While some people have commented on how they are going to "trick the system" to drive their Elo down past where it should be. While it's easy to say "Who cares", I am concerned that some players will do this to boost their C-Bill gain. I would envision someone Alt-F4ing, or after that is actually fixed, simply playing fast mechs and rushing the enemy to die fast. This is going to increase their C-Bills per hour by boosting their games per hour rate, but of course the majority of those games will be losses since they left their team down a mech. Once in low Elo land they can start actually playing, and owning the heck out of of enemies, which is of course unfair for those low Elo enemy teams. Repeat, profit. PGI has stated they will have the ability to monitor for this behavior, so I hope they actually do and keep on top of it.


A potential solution to both this issues would be an Elo rating modifier to your end of match earnings. Perhaps top 25% players get a bonus 5% C-bills, top 5% get a 10% bonus, for example. This would make it unprofitable for good players to dip down into low Elo on farming runs. It would also help answer the calls of those that want to see their Elo rating. With this "bonus brackets" players would have skill goals to strive towards.

Edited by 80Bit, 18 February 2013 - 02:34 PM.

[p4r4g0n]

Aren't players already earning more relative to others in a match based on their ranking in the match? So if you have a 1:1 win/loss but consistently top the list, you'd earn more than those who don't even with a similar win/loss ratio.

[arghmace]

p4r4g0n, on 18 February 2013 - 04:17 PM, said:

Aren't players already earning more relative to others in a match based on their ranking in the match? So if you have a 1:1 win/loss but consistently top the list, you'd earn more than those who don't even with a similar win/loss ratio.

If someone consistently tops the list, then he's better than the rest of the players on that tier. He will move to tougher battles and his earnings will drop. Once again, make a difference between a single battle, or even 100 battles (still short term) and the truly long term gainings.

80Bit above wrote exactly what I've been rambling on about here. Only he wrote it better

[WVAnonymous]

Wintersdark, on 18 February 2013 - 01:27 PM, said:

They're implementing it largely in the same way, calculating team Elo rating as the average of it's members, and applying the results as such.

Each player is rated based on the performance of his team. While this could be grossly inaccurate in a single given battle, consider that each player is 12.5% of his team. Each player has a substantial impact. As such, while your initial rating can vary wildly, it'll work out eventually.

Because Elo rankings are being preseeded right now, you should be placed with other players with a comparable win/loss ratio. If it turns out you're better than your ranking suggests (or you improve), your impact on the battle will lead to a somewhat higher frequency of wins for your team, compared to someone else in your spot who's just not quite as good.

Thus, as you're playing with different people all the time, each of them gains a very small Elo ranking bonus per win - as do you - but you get more wins. This pushes you into higher ranked teams for later games.

So, it does work out. You just have to look at the big picture, over hundreds of games, and with the knowledge that you're much more often than not going to be placed with roughly comparable players.




Another way to look at it: Lets assume, for some reason, you're really unlucky and get repeatedly placed with players who can't fight their way out of a paper bag, and somehow end up playing at the 1000 level (while you in fact are a 1400 player). Now, this is extremely unlikely, but theoretically possible.

The reality is that you're on a team with 7 1000 ranked players and you, ranked 1000 but actually 1400. You're fighting 8 other 1000 ranked players. Which team is more likely to win? It's not like now where you can be a great player on a terrible team, facing an overall average or great team. The battle you're in here is two roughly evenly matched teams, with one team having a crazed killing machine on it capable of soloing the entire opposing team.

You'll win LOTS of those matches, and very quickly climb out of the wrong bracket, while the other players on your teams are only winning the odd match, so their ranking doesn't change dramatically as they lose immediately when facing teams that suck less than them.

So, the algorithm could be as brain-dead as "your W/L ratio for the last 15 matches is above 0.500 time to raise your rating". Maybe the math isn't as interesting as I'd hoped.

Still better than nothing, and looking forward to the patch.

[Mackman]

WVAnonymous, on 18 February 2013 - 04:42 PM, said:

So, the algorithm could be as brain-dead as "your W/L ratio for the last 15 matches is above 0.500 time to raise your rating". Maybe the math isn't as interesting as I'd hoped.

Still better than nothing, and looking forward to the patch.

I think you're misunderstanding Elo. If your team's Elo (which is likely going to be the average of all the individual Elo's) is higher than the other team's Elo, and you lose, then your personal Elo goes down: Likewise, if your team's Elo is lower then that of the other team, and you win, than your personal Elo goes up.

I'm not sure if the same will hold true for the opposite scenarios: If you win against a team you're "supposed" to win against (if your Elo is already higher than the team you beat), does your Elo still go up and does their Elo still go down? I don't know: In League of Legends, it goes up less if you were supposed to win (and down less if you were supposed to lose), and vise versa.

There's no "check w/l after X amount of matches to see if it's 50%." The 50% is the end result, not the mechanism by which it achieves the result.

[p4r4g0n]

arghmace, on 18 February 2013 - 04:29 PM, said:

If someone consistently tops the list, then he's better than the rest of the players on that tier. He will move to tougher battles and his earnings will drop. Once again, make a difference between a single battle, or even 100 battles (still short term) and the truly long term gainings.

80Bit above wrote exactly what I've been rambling on about here. Only he wrote it better

And this is an issue because the player can no longer manipulate the odds by various means so that he wins 90% of the time all the time thereby ensuring he always earns more than any else?

Sorry but I don't see the problem here.

On average, the C-Bills earned even without a 1:1 win/loss, as in my case, is adequate based on the current reward system even without hero mechs or premium time. If I want to earn more I should maybe try to improve my game or shell out for a hero mech or premium?

In any case, the presumed outcomes outlined by 80Bit are based on a fundamental assumption about how teams are put together by the matchmaker. As I've pointed out in an earlier post, while this assumption may be valid, it really should be noted that the current method of creating a team is not consistent with this assumption nor has there been any Dev comment that this will be changed.

If the current method of putting a team together continues to be used, I anticipate a significantly less straight forward clustering of players throughout the Elo rating range although it will still result in more reasonably balanced games.

Since the patch will be out in a few hours, I'm not going to post a lengthy explanation on this. Let's re-visit the issue after the patch notes / Command Chair post (if any) are out.

[Grits N Gravy]

Broceratops, on 17 February 2013 - 06:46 AM, said:

Elo Hell is a myth. The concept that there is a level of low elo where results are so random that you cannot have any effect on them just means you are at your elo level. The term exists only because players at this level don't have the deductive reasoning to understand they're bad and they're being protected from higher ranks.

False, Elo hell or the point where elo scores stagnate or the effort to raise your elo take tremendous effort and very long consecutive win streaks do exist as a function of the mathematical formulas used to create elo scores. The primary culprits are K factors and systems that clip point gains when a more skilled opponent beats a less skilled one. As is the case in our Elo formula.

It got so bad in LOL that they changed the function of their leader boards. Previously your rank on the leader board was solely based on your Elo score. They have radically changed this recently allowing players to advance up the leader boards via tournament style play. http://na.leagueofle...d.php?t=3004520

Broceratops, on 18 February 2013 - 06:44 AM, said:

I agree with Wispsy. I don't need a number going up necessarily but I need something. You cannot call a game competitive unless there is a ranking of some kind. I don't care if its leagues like in SC2, or a number like in LoL, or just random **** I can put in my cockpit win from tournies if they ever add those.

Leader boards are different from matchmaking. They do need to develop a leader board system at some point, but it should be taking in more factors than win lose to paint a more complete picture of "skill".

I honestly don't think the MM system is going to be marginally impacting on anyone's play. First off, the laws of probability tells us average solo dropper should have about 48% +- 5 percent win rate. All things being equal, unless the average player, plays only in an advantage situation there is no way mathematically to raise the average win rate as a function of the laws of probability. As soon as he does this, either by increasing skill or mech's he plays, he is no longer by definition average.

Mathematically, the function of a good match maker is to take advantaged and disadvantage players, divide them from the other levels of play and have play only themselves. Which would normalize the win rates and advantaged and disadvantage play styles. That's what Elo does; it's a mathematical way to represent how advantage one's play is against another.

It won't work any better than random drops for a number of reasons. The most obvious being, that the pool of players to match from, will be so small, that widely varying Elo scores will have to be matched against each other in order to ensure prompt matchmaking. Thus defeating the entire purpose of Elo matched making.

The games total amount of players logged on, is not the number of players that games can be made from. Only a tiny percentage of the player base is available to match at any given moment. The percentage of player available to match from can be worked by comparing the amount of time you que for a match compared to the amount of time you spend in a match, in the lab or idle. In most cases que time is less than 1 percent. Therefore, at any given time 1-10% of the player population is available to match from. Therefore, you can't build tight Elo matches and insure reasonable que times. As there are never enough advantaged and disadvantaged players in a que at a respective time, to build a match, of exclusively those players.

Match parity and que time are divergent factors. The tighter the Elo match, the longer the que time has to be, which is a function of the total size available to que. If you really want consistent 50/50 matches you have to be willing to spend the same amount of time you spend in battle waiting for opponents. The majority of players won't wait that long, most won't wait over a minute.

All that's going to happen with MM 3.0 is que times will increase, the quality of the matches will stay the same or go down and the amount of complaints about matchmaking and que times will increase exponentially. Complaints will increase as function of human physiology and the reactions to waiting and fairness.

http://www.nytimes.c...?pagewanted=all
http://davidmaister....-waiting-lines/

IMHO all the time that goes into an ELO matchmaking system is a waste, because the laws of probability get us to about the same point and Elo can't get us to a better point but could make things worse. The risk to reward doesn't justify the expenditure of time given the population size. The time would have been better spent doing a Battle value system, to ensure better mech parity in matches, creating a dynamic leader board and tournament system.

Edited by Grits N Gravy, 18 February 2013 - 07:28 PM.

[80Bit]

Grits N Gravy, on 18 February 2013 - 07:07 PM, said:

IMHO all the time that goes into an ELO matchmaking system is a waste, because the laws of probability get us to about the same point and Elo can't get us to a better point but could make things worse. The risk to reward doesn't justify the expenditure of time given the population size. The time would have been better spent doing a Battle value system, to ensure better mech parity in matches, creating a dynamic leader board and tournament system.

Since a lot of what you say is based off of our "low population", can you let us know what that low population number is, and where you are getting it?


While you are right that mathematically in a true random distribution, "average" players are going to have a near 50% win rate, this is currently not the case. Average and above average players (as determined by their typical in match performance on their team) are maintaining below average win rates. The distribution is not true random and is not producing quality matches, because of the grouping system. If it were random win rates would look when I played early World of Tanks, which was straight random match making (as far as skill) and my win rates across dozens of tanks were all 49%-51%. Here my solo win rates are 41%, even though when I sample my score boards my in match performance is more like 70th percentile.

As you point out the percentage of the population available in the <1 minute matchmaking window is quite low. During this window, it appears that either very few pre-made groups available, or that the match making is ignoring pre-made groups and not trying to pair them against each other. As a result, when a 4 man pre-made queues, they are now taking up 4 of the 16 slots for a match, and the rest of the slots are most often filled with the much more abundant solo players. A solo player has a 4 in 12 (33.3%) chance to be placed on the side with the pre-made group, and an 8 in 12 (66%) chance to be placed on the side opposing the pre-made group. Pre-made groups win the vast majority of matches they play, and thus, solo players of almost all skill levels are losing more than 50% of their matches. The current matchmaking fails to deliver fair matches to all players.


So to summarize the current state of MWO match making is not matching the laws of probability you are talking about. It is my belief that the Elo structure will indeed improve match quality by helping correct the pre-made placement problem in a round about way.