92 replies to this topic

[Albert Meyburgh]

Chemie, on 10 June 2013 - 01:35 PM, said:

First thanks for at least chiming in here. Nice to see developer engagement. My question; What does telemetry tell you? Is my observation that higher ELO players are being matched against new players/trial mechs etc real? I would assume you could look at the standard deviation of the match ELOs to see that the variance has increased? Or just run a count of "bad/extreme ELO matching"? or in games with 4 high ELO< how often are there trial mechs? or recruit with <25 games? As I stated, I have not seen this imbalance in MM since ELO was first introduced so something has changed. Thanks. (PS How about chaning the queues to be premade and pug; that way we COULD drop against our own lancemates of equal ELO).

I think that the 'new player change', could explain someone seeing more trial mechs yes. Though a formal analysis would be best left to a future mm update when the data is properly digested on this end. My gut feel as an engineer is that if you mess with statistical systems you will lose but I have to wait for logic to back up emotion.

As for separating the queues, this is something I want to avoid since every time you split queues, you reduce the numbers in the queues, which cause the matchmaker to have even less to work with--though, this is completely up to design as to how to approach that and on one hand I do appreciate the sentiment.

[Nicholas Carlyle]

srccoder, on 10 June 2013 - 02:11 PM, said:

I think that the 'new player change', could explain someone seeing more trial mechs yes. Though a formal analysis would be best left to a future mm update when the data is properly digested on this end. My gut feel as an engineer is that if you mess with statistical systems you will lose but I have to wait for logic to back up emotion.

As for separating the queues, this is something I want to avoid since every time you split queues, you reduce the numbers in the queues, which cause the matchmaker to have even less to work with--though, this is completely up to design as to how to approach that and on one hand I do appreciate the sentiment.

Any chance you could explain this to Bryan? Because he said he's 2 months away from releasing 3PV which based on what he said, would split us into 3 queue's (3PV, 1PV and 1 & 3PV).

And of course 95% of us voted against the idea.

Or do you guys not talk to Bryan?

[Vodrin Thales]

Chemie, on 07 June 2013 - 03:30 AM, said:

or are the queues that empty?

I was playing last night in assault queue and had at least 2 trials mechs in each match and one game a guy said "my first time every playing".

Now I do not profess to be an expert but I assure you my ELO is sufficiently high that I should not be seeing trial mechs players (nor have I until last night). Are the queues THAT EMPTY?

(3000 wins/800 losses with 5:1 K/D; I only post this to avoid the "you just suck" responses).

As stated, I have never seen trial mechs and newbs in drops since ELO was implemented. Also, I know good players might take one for a spin now and again but the reality is not 2 per match.

MM balances ELO's on the two teams. If the queue is relatively empty the MM will just make sure there are the same number of noobs on both teams, and about the same number of vets. And I've played against you Chemie, you are definitely a strong player and a scary good shot at times.

[Renthrak]

srccoder, on 10 June 2013 - 09:22 AM, said:

The matchmaker tries to get homogeneously mirrored teams when it can, but it only has 2 minutes to do that before a timeout happens, so if someone's been waiting for awhile without perfect matches coming along, then the matchmaker will form the best heterogeneous match it can! Of course everyone would quixotically prefer perfectly mirrored matches, but that's the best that can be done in that 2 minutes of waiting.

How can this be improved? Well, either you need more people queuing up per minute (in order to increase the probability of perfect matches being formed), or you need a longer timeout in order to increase the consistency of the team compositions.

Do you think it would make any difference to the matchmaker if instead of using the player's ELO score directly to find matching players, the ELO score was used to place players in a 'bracket' for a given range of ELO scores, allowing a simplified search for players beginning within the same bracket?

Functionally, the idea would be to have the 'bracket' assignment occur when the player's ELO changes. That way, the matchmaker doesn't need to check the ELO rating for finding players, because the bracket number was already determined ahead of time.

An example would be:
Player A has an ELO of 1550, placing him in Bracket #5. Bracket #5 contains all players with an ELO between 1401 and 1750. When Player A searches for a game, the matchmaker simply looks for other players tagged with Bracket #5. If there are not enough Bracket #5 players to fill the game, it expands to include Bracket #4 and Bracket #6, and so on.

If I understand the way the matchmaker currently works, this would not fundamentally alter the way it functions already. The idea would be to make the ELO matching portion more efficient. Ideally, that could allow the weight matching portion to take more resources, hopefully finding a more ideal match within the same 2 minute timeout limitation.

I wouldn't expect very much of a difference on a small scale, but when hundreds or thousands of players are searching simultaneously, it could add up.

Edited by Renthrak, 10 June 2013 - 03:05 PM.

[Hauser]

Renthrak, that makes no sense. Perhaps it helps if you see Elo as a really fine grained bracket system.

Edited by Hauser, 10 June 2013 - 03:50 PM.

[Chemie]

srccoder, on 10 June 2013 - 02:11 PM, said:

I think that the 'new player change', could explain someone seeing more trial mechs yes. Though a formal analysis would be best left to a future mm update when the data is properly digested on this end. My gut feel as an engineer is that if you mess with statistical systems you will lose but I have to wait for logic to back up emotion.

As for separating the queues, this is something I want to avoid since every time you split queues, you reduce the numbers in the queues, which cause the matchmaker to have even less to work with--though, this is completely up to design as to how to approach that and on one hand I do appreciate the sentiment.

I am an engineer (a real one meaning professional engineer) and I love statisitics.

BTW, I do not want a third queue. Pug and pre-made. Lose the 8-man. It will improve ELO because at least all the pugs are pugs and all the premades are premades...mixing just messes with ELo calc and weight balancing.

And no, "new player change" does not explain why a high ELO player would suddenly have a bunch of games with 1100 ELO players driving trial mechs and typing in "this is my first game...be gentle" (and being serious and not trolling)

[Taemien]

SgtKinCaiD, on 10 June 2013 - 10:30 AM, said:

I must repeat myself then :
This is not a surge of new players, it's the opposite : a lot of players are not playing MWO right now due to the JJ shake (and maybe the nerf of poptarts), search the forum about "motion sickness" and you'll see.

This phenomenon is more noticeable now thanks to that, but several players, me included, have been suffering from this, since the introduction of ELO ratings :
http://mwomercs.com/...w-when-to-quit/
http://mwomercs.com/...uch-a-wide-net/
etc.

PGI doesn't want to acknowledge that the number of players in MWO is not sufficent for the ELO system to work properly right now. They were forced to change several times the MM in order to lessen the waiting time, which is a good indication on the problem. Now it appears to widen even more its search as it does not take the mech tonnage into account anymore :
http://mwomercs.com/...ctions-anymore/

The MM was better without the ELO. We wanted a basic CW (a small lobby to prepare and launch 8v8) and PGI threw the ELO ratings in order to keep bad PUGStompers at bay ! Take that big meanies ! But guess what ? Players are still getting stomped except that it's the MM who decide now who will get stomped or not.

I have never been able to prove that people were leaving. And believe me I have tried. I always argue against a point or position I intend to take. I believed people bluff when they say they will quit. So I tried to see if I could prove that people are quitting, but I couldn't.

I did this by offering a guy who was adamant about quitting a means to prove his point. I did this by offering to kill his account for him. Basically he would allow me to take his account, change the pass to something I couldn't remember (so I wouldn't get a free account, it would really be dead). And in return I would purchase $25 in MC to put my money where my mouth was in this whole ordeal. This way the guy wouldn't feel like he was losing something alone.

He never replied and went on to other threads and is probably still playing (I know he continued to play after the debate because he continued to make posts about things happening ingame). In otherwords, he never quit.

So... if you have a better way of showing proof that players are quitting, then I'd like to see it. Until then the game is growing, not shrinking. The vocal minority on these forums aren't enough to convince me.

By the way, the fact that MWO doesn't have enough to support perfect ELO isn't good enough. WoW had 12 million at its peak and it wasn't enough to give balanced arena matches. It wasn't enough for their LFG tools to fill up in under 30 minutes depending on your class.

Edited by Taemien, 10 June 2013 - 05:59 PM.

[ICEFANG13]

I also figured I would point out that, if you look at this page, info from PGI (if correctly counting, and I assume it is, as it increases when I look at it, and then re-tab the page) indicates a slow overall decrease. Large bumps are usually hero mechs, new mechs, and show a difference of gameplay-mechlab-gameplay.

http://img402.images...52481184314.jpg

This is the part 1-2-3 of the graph un-put together.
http://nces.ed.gov/n...p?temp=909123.1
http://nces.ed.gov/n...p?temp=986919.8
http://nces.ed.gov/n...sp?temp=1058013

[SgtKinCaiD]

@Taemien :
I did not say explicitly that people are leaving, i say that a lot of players are not playing currently due to the JJ shake that causes them troubles (see JJ feedback thread for example, 43 pages atm, enough said : http://mwomercs.com/...shake-feedback/ )

The MM normaly matchs people by ELO ratings and then tonnage, right ?
Then why is there a surge of new threads complaining about tonnage mismatch, new players getting paired with veterans ?
-> http://mwomercs.com/...er-still-sucks/
-> http://mwomercs.com/...775-matchmaker/
-> http://mwomercs.com/...attle-matching/
-> http://mwomercs.com/...he-match-maker/
-> http://mwomercs.com/...-mm-turned-off/
-> http://mwomercs.com/...d-steamrolling/
-> http://mwomercs.com/...ctions-anymore/
-> http://mwomercs.com/...uired-to-drive/
-> http://mwomercs.com/...117-i-hate-elo/
-> http://mwomercs.com/...echs-per-match/
-> http://mwomercs.com/...-refuse-to-pug/
-> http://mwomercs.com/...alancing-teams/
-> http://mwomercs.com/...-tuning-effort/

Why is there a growing number of threads talking about showing individual ELO, organising ELO by brackets, creating a stock mode only for new player :
-> http://mwomercs.com/...us-see-our-elo/
-> http://mwomercs.com/...nce-discussion/
-> http://mwomercs.com/...eight-matching/
-> http://mwomercs.com/...ake-elo-public/

Because the MM doesn't have enough players to work correctly. And this phenomenom was amplified last patch : a lot of player have stopped playing in one go (for various reasons : JJ shake causing motion sickness, poptart builds not being so easy to pilot anymore, etc.)

As I already said, this problem is not new, we have been several players to talk about it, but as it impacted only a selected few, nobody was paying attention. And now ?

PS : i'm not talking about people leaving the game, i'm talking about the current pool of players. People tends to stop playing for a while and then come back to check the new stuff.

[John MatriX82]

SgtKinCaiD, on 11 June 2013 - 02:07 AM, said:

@Taemien :
I did not say explicitly that people are leaving, i say that a lot of players are not playing currently due to the JJ shake that causes them troubles (see JJ feedback thread for example, 43 pages atm, enough said : http://mwomercs.com/...shake-feedback/ )

The MM normaly matchs people by ELO ratings and then tonnage, right ?
Then why is there a surge of new threads complaining about tonnage mismatch, new players getting paired with veterans ?
-> http://mwomercs.com/...er-still-sucks/
-> http://mwomercs.com/...775-matchmaker/
-> http://mwomercs.com/...attle-matching/
-> http://mwomercs.com/...he-match-maker/
-> http://mwomercs.com/...-mm-turned-off/
-> http://mwomercs.com/...d-steamrolling/
-> http://mwomercs.com/...ctions-anymore/
-> http://mwomercs.com/...uired-to-drive/
-> http://mwomercs.com/...117-i-hate-elo/
-> http://mwomercs.com/...echs-per-match/
-> http://mwomercs.com/...-refuse-to-pug/
-> http://mwomercs.com/...alancing-teams/
-> http://mwomercs.com/...-tuning-effort/

Why is there a growing number of threads talking about showing individual ELO, organising ELO by brackets, creating a stock mode only for new player :
-> http://mwomercs.com/...us-see-our-elo/
-> http://mwomercs.com/...nce-discussion/
-> http://mwomercs.com/...eight-matching/
-> http://mwomercs.com/...ake-elo-public/

Because the MM doesn't have enough players to work correctly. And this phenomenom was amplified last patch : a lot of player have stopped playing in one go (for various reasons : JJ shake causing motion sickness, poptart builds not being so easy to pilot anymore, etc.)

I think the same, probably numbers got shrinked, with SRMs not in shape, lrms back to a certain degree of efficicency (seeing 5xlrm15 stalkers isn't rare) and all the sniper gameplay has certainly driven out many players.

Yesterday night I was going around in an assault 4 man (sometimes 3A+1H) and we met several teams the other side to feature ligths while my team didn't at all (not even mediums) or with only 2-3 assaults and mixed mediums/heavies.

[Taemien]

I have no doubt there is a bunch of players not playing. But I doubt its for the reasons above. MWLL had screenshake and that didn't stop players from using JJs or jump sniping. MW4 had a way worse sniping game and it went on for years and years. And neither game had any sort of match making.

You have to understand something about the gaming community. It complains. ALOT. I was playing Everquest the other day and people were complaining that it was too easy and a myriad of other things. But those things hardly keep them from playing it. Look at WoW and its millions and millions of players. At its peak, it was absolutely riddled with complaints. If you went by the complaints alone, you would have thought the game had no real players.

I listen to constructive feedback, not complaints or whine. The moment anyone tries to insult the devs in any way possible, I know its just another whine thread from a player who wants attention, not a player who actually wants to see a fix. And players have been pretty sneaky about how they word things to 'appear' to be constructive when they are not.

A good example is the thread titled, "Game Making 101" the OP tried to sound legit.. but come on... the tittle alone shows that he was arrogant enough to suggest he knew more about game making than the devs. A veiled insult, and I reamed him for it until he abandoned the thread.

The match maker isn't perfect. Its not possible to have a perfect match maker. You need literally thousands of players in the queue at the same time for it to work. I've said this in other threads but I'll say it again. If you have 10,000 players online at the same time (which is double what CB had). Assume 90% of them are playing in a game at a given moment. That means 1000 players are available for a queue. However some of those thousand will be afk. Another bit of them will be browsing forums, and another bit will be fiddling around in the mechlab. So that may give you about 200-300 at the most queuing up at the same time. Giving about 10-20 match makes.

Sorry to say, but that isn't nearly enough to get perfect results. Well.. it will actually. But only for a handful of games. The rest will get mismatched because well, its got 2min to match you and its going to match you with 15 other people (if solo queuing).

Now lets say its an offpeak time and we see a fifth of the amount online at a time so 40-60 in the queue. Well you've got 2-4 match makes right there. Yeah its going to be imbalanced unless by some coincidence everyone there is the same score.

Now I'm not going to say people aren't playing. Because I'm not even playing myself. Well I do, but only about 3-4 matches a week on average. The reason for this, and its the same reason for everyone I play with has. There's just not much to do. The game is pretty basic. You ready, launch, shoot, and repeat.

MWLL had this problem too. And I played it about as often as I play MWO. But I did that from 2009 to 2012. So this doesn't make the game dead.

Now MW4 was a different story. That was being played nightly and many many games per night. This was because I was playing in a planetary league and was responsible for practice matches with my unit. I don't have a unit in MWO. I don't have a need or desire for one. There's no purpose behind it. I don't have any feeling of progression, progress, or incentive. Win or lose, I get Cbills. Cbills I kinda need for some configs I kinda want. But without any sort of league play, I have little incentive to unlock more variants and work them up.

Judging by how popular leagues are in the MechWarrior and MechCommander franchises, I'm pretty sure I'm not alone in this. Again this doesn't make MWO a failing game. Just not an often played one. I do believe that CW will change that. When players have territories to fight over then group play and competitive play will take off and you'll see more playing.

Though if done right, ELO will be largely a non-issue. Ideally players will just group up in a lobby with their fellow faction, unit, and friends, discuss what mechs they take, launch and kick some butt regardless of weight or ELO score. Thats how it went down in previous titles. And like myself, I think alot of players are looking forward to that.

So don't let the amount of complaint threads or the length of some of them worry you. MWLL had just as many, and just as long (and longer, just look at some of their legging threads). And their community was only about 200-300 strong. MWO is much bigger as at this moment of time (midday EST) there's 1300 on the forums. I think we're doing alright.

[Jestun]

Chemie, on 10 June 2013 - 03:39 PM, said:

BTW, I do not want a third queue. Pug and pre-made. Lose the 8-man. It will improve ELO because at least all the pugs are pugs and all the premades are premades...mixing just messes with ELo calc and weight balancing.

I assume by "PUG" you mean "solo"? In which case, no.

Forcing the matchmaker to have to match all the 2,3,4,5,6 & 8 man teams (7 man would, of course, be impossible as there would be no solo player to take the last spot) would have serious effects on either the time spent queueing or the balance of the matches as the ELO criteria are relaxed.

For example, a 5 man team can only play when a 3 man team also queues. A 2 man cannot help as that would once again leave 1 empty spot and no solo players to fill it.

[Chemie]

Jestun, on 21 June 2013 - 07:38 AM, said:

[/size]

I assume by "PUG" you mean "solo"? In which case, no.

Forcing the matchmaker to have to match all the 2,3,4,5,6 & 8 man teams (7 man would, of course, be impossible as there would be no solo player to take the last spot) would have serious effects on either the time spent queueing or the balance of the matches as the ELO criteria are relaxed.

For example, a 5 man team can only play when a 3 man team also queues. A 2 man cannot help as that would once again leave 1 empty spot and no solo players to fill it.

Just make people have to queue in increments of 2. So 2,4,6,8,10,12. That works much easier. I would even be fine with forcing a lance increment of 4,8,12.