Is matchmaking turning you off from the game?

True Elo: The Elo rating a player would have if it were a perfectly accurate measure of their skill.
Current Elo: The Elo rating a player actually has. This can be above, below, or at their true Elo.
Elo Hell: A state where one's current Elo drops below one's true Elo, and the resulting decrease in allied skill causes winning enough games to climb to their true elo to be impossible.
Elo Purgatory: A state where one's current Elo drops below one's true Elo, and the resulting decrease in allied skill causes winning enough games to climb to their true elo to be incredibly frustrating, but possible.

In a 1v1 skill-based game, like chess, Elo is only determined by the skills of the individual players. In a 1v1 partly chance-based game, like some forms of Poker, Elo would have an uncertainty. In a team vs team game with random elements or enough complexity to display chaotic behavior, like MWO, Elo will have a large uncertainty.

I'll make some simplifying assumptions to start with, and slowly remove them as needed.
MWO is a 12 vs 12 game. Let's assume the matchmaker uses the arithmetic mean of the Elos of the players on each team, and matches the teams to have the same Elo. Let's also assume that each player can influence the game only as much as they change the average Elo of the team. For example: Team A has 11 players of Elo 1000 and 1 player of Elo 2000. Team B has 12 players of Elo 1000. Team A has an average Elo of 1083.3, and Team B has an average Elo of 1000. Using the expectation formula for Elo*, team A is expected to win 61.8% of the time.
Let's assume that the change in rating each player gets is equal to that gained or lost by the team. EG if the team would gain 12 Elo from winning a game each player's current Elo will increase by 12.

It's quite easy to see that Elo Hell can't exist. If Elo Hell exists, then a team with a player in Elo Hell will at some point face teams without a player in Elo Hell, and since the player in Hell means that his team has a higher true rating than current rating his team will win more than 50% of the time. Thus the player will slowly win more games than he loses, and his rating will rise to its true value.

Elo Purgatory is more interesting. The question here is "how many games does it take to reach one's true Elo." Let's take a player with a true Elo of 3000, and a current Elo of 1500. We'll use the new rating formula** with a K-Factor of 10 and try to see how many games it would take our player to reach his true Elo. We will assume all 23 other players on our player's team have exactly our player's current Elo.
Average True Elo: 1625
67% win chance by True Elo, 50% win chance by current Elo. Elo changes only go by current Elo.
Elo change for a win/loss: 5
So he should expect to gain (.67*5-.33*5)=1.7 Elo from the game.
For now I'll assume the gain continues to be 1.7 per game (it will actually slow as our player's rating approaches the true rating). In that case, it would take 883 games for our player to reach his true rating. In practice this number will be well over 1000 games. I may go back and show the actual curve at some point, but I don't have the time ATM.
Now let's try with a K-Factor of 24:
Elo change for a win/loss: 12
So he should expect to gain (.67*12-.33*12)=4.1 Elo from the game.
So with the linear model it would take 368 games to reach his true Elo, in practice probably around 500.

So depending on the tuning of the K-factor Elo purgatory may or may not be utterly horrible. Elo hell will never exist. Since Elo is hidden the only way to find the K-factor is to get it from the devs. It may also vary, some chess leagues use different K-factors depending on how experienced the player is. It's not likely under 10, so after about 1000-2000 games you should be out of Elo purgatory.

* The Expected win chance of A, as a decimal is 1/(1+10^((EloB-EloA)/400)). For B, it is 1/(1+10^((EloA-EloB)/400)). Where EloA is the current Elo of A, and EloB is the current Elo of B.
** Elo Gain/Loss of team A = K(ScoreA-ExpectedA) where ScoreA is 1 for a win, 0 for a loss, and 0.5 for a draw, and ExpectedA is the expected win chance of A. K is a constant. A low K means Elo changes slowly, a high K means it changes very quickly. If K is too low it takes too long to reach one's true Elo, if K is too high one's Elo will jump around a lot.

 

If ELO hell doesn't exist then where the hell am I? Game after game match after match its the same thing. Not which team is better but which team sucks less. What I don't understand is how I can be on a team full of muppets and still be in matches against teams of NGNG premades or Michael, Blagg or PGI developers (insert great players here). Its like every so often ELO takes a holiday and the matchmaker just gives up.

I am an exceedingly average player (see stats below) so when I come top damage or top kills match after match I am just simply the best of the worst which also means that they expect ME an average player to pick up the slack for below average and downright horrible players. An Elite player may be able to dig himself out of ELO purgatory but what hope do the average players have? I am good enough to be top damage and kills but no where near good enough to affect the outcome of matches the same way a good player is so I have no chance of fighting those much better players and no hope of improving my skills doing so. It appears the ONLY way would be to do nothing but drop with a lance. If that isn't ELO hell I don't know what is.


I do 100% PUG matches
Heavy:
Avg Dmg: 340.3
Win Percentage: 56.0%
K/D: 1.21

Assaults:
Avg Dmg: 308.6
Win Percentage: 55.0%
K/D: 1.57

 

Thats easy. I am in ELO Hell just the same as the rest of you. There are players out there who sit there and do nothing but 12 man drops in their elite units and win a lot more than they lose but the skill level of their team mates is on par with their own. I tend to do a lot of PUG dropping and there are matches where, if I didn't shave my head, I would be pulling out my hair at the sheer lack of skill I get surrounded with.

Atlas pilots wandering off with silly loadouts trying to chase down Ravens and Locusts is my new favorite thing.

 

Sounds like Hell to me.....

Seriously though, a view from a bad mechwarriors perspective, how many drops do I have to do where I dish out 500+ damage with at least 4 kills of my own for a team loss, before I end up dropping with people at my level. No team work, no strategy at that level... coming from a military background, there is more than a bit of frustration involved.

Last thing I did in the game was buy a jester.... I haven't actually played in weeks
Is matchmaking turning me off from the game?... more like the lack there of.

 

If your win rate is > 50%, your Elo is rising. You're not stuck, it's just a very annoying grind. Thus why I use "purgatory" for it instead of "hell." It also means that if you have less than about 2k games played you won't be at your true Elo.

I'm not saying it doesn't suck, just that there is hope. And maybe PGI will get a clue and use Glicko or such, but that's not terribly likely.

 

Wins / Losses: 2,167 / 1,672

I've had matches where I have wrecked face with 4+ kills and several hundred points of damage and the PUGs still manage to F it up somehow for a loss. Not only does it suck, it sucks hardcore. My problem is that I do a lot of PUG dropping when my guildies aren't on or when I want to speed grind or whatever and this is what I end up with; the bottom of the barrel.

 

And that's almost certainly due to the matchmaker being very loose about Elo, and not about your Elo being too low.
Or possibly there are multiple Elo ratings per player, one set for solo drops, one for dropping as a lance, and one when playing in 12-mans. I don't have enough information about the system to make really good conclusions. They might not even be using Elo, it might be a modified system like Sonas or such. Actually, it has to be modified, since Elo uses table lookups based on tournament category, which MWO doesn't have.

 

I don't understand some things...

1.
Let's assume some player P has rating R but really for his skill it should be freaking 2R. And matchmaker makes teams with all members in both teams with rating about R.
Then P should be like hero between rookies in both team. His team has big advantage by having him in it.

So, if you want to be in Elo hell you need to be overestimated. System will bring you in cool team which fight with cool team, and you just effective like half-disconnected because of your bad skill and therefore you get all those losings.

What's wrong in my reasoning?

2.
If bad Elo estimation just inserts you in Elo hell or doesn't insert, why then me and other players have those 50:50 be-steamrolled-or-steamroll-them extremely unfair matches? Or is Elo estimation so horribly bad that near ALL of players is both team are underestimated with different (and unpredictable) degree of underestimation?

3.
I think all those tonnage/weapon/mech class/etc. matchmaking is so bad now, that matches balance is unpredictably unbalanced in both terms of skill and tonnage/weapon/mech class/etc.. And this chaos is so bad that bad Elo is negligible and changes nothing in those matches of pure lack and randomness.
Isn't it?

 

I think theres a lot to what you say...I think. I agree that it appears that match maker is struggling to balance teams out and I can think of only one or two possible explanations

1.) Our server population is too low for effective match making to work (which scares me)

2.) We ALL suck and matchmaker just says screw it I don't care anymore

[move]SQUIRREL![/move]

I was in a match facing Garth and 3 guys from NGNG in a lance...our team got roflstomped and Garth says after the match don't blame the match maker. When a 3rd of the enemy team are elite players using comms and the other team gets muppets (me included) I honestly don't know who else to blame but the match maker. I hate matches where I feel like cannon fodder for the other team and that has been happening a lot lately. I have had some VERY close matches for sure but by and large the majority end in 12-4 or 4-12 routes and MANY 12-0 0-12 routes that just shouldn't be possible if the teams are even remotely balanced by weight and skill

 

1) Match maker probably works as follows:
Find a group with an average Elo within 10 of the player searching for a match. If no such groups exist, find one within 20. etc, etc.
That then spreads the Elo out, and if it's not clamped well you'll end up with groups where players have wildly varying Elos.

2) The steamroll-or-be-steamrolled thing has nothing to do with skill or Elo or whatnot. That's a feature of the game, and pretty inherent. Each team has a certain amount of armor and a certain amount of DPS they can put out. When a team kills an enemy player the other team not only loses that armor, they also lose the DPS. This makes it harder for the enemy team to get a kill in return, and the resulting fire from the friendly team becomes more focused. That means the next target on the enemy team dies faster, and the one after that faster still. Unless the enemy team gets a kill back very quickly, or gets a kill well out of proportion to their losses (say, losing 3 Spider 5Vs and killing a Stalker) they're not likely to reverse the trend. Essentially the "even, trade 1 mech at a time" games are an unstable equilibrium, and the "one team utterly crushes the other" matches are stable equilibria. You'll see the latter a LOT more than the former, no matter how well balanced the teams.

3) Pretty sure tonnage/mech/class etc aren't even considered ATM.

 

Amusing how a few decisions made by the OpFor resulted in their threat density being so low we were able to hunt them down with such force that just 2 in 12 managed to break into triple digit damage. I don't really have a point, just sharing this occurrence.

Likely the case at non-peak hours, I start not finding matches after midnight-1AM pacific. Certainly reaching the long end of the search clock.

The odds get longer for each mech lost, I agree. Another part of the steamroll nature of MWO is unorganized or ineffective deployment of forces, especially in PUGs. Atlases humping hills on Frozen City, tunnel rushers getting stopped at the exit on Forest Colony, entire lances flanking wide and not pincering in or getting a cap split/distract. While not entirely ineffective, these are low upfront yield roles/routes. If the enemy team is effectively deployed, they will be in a position to make a decisive blow, even before the first kill is scored and equilibrium is at hand. Equilibrium of mechs who're actually in the fight.

Other things potentially not accounted for:
*Mech price: DHS, XL engine, min/maxed brimming with firepower ballistic boats versus stock Blackjacks with Flamers.
*Mech mastery: Doubled Heat Capacity and Cool Run amount to what, 40% more alphas, and 15% more sustain? Speed tweak is obviously huge, as are the handling boosts for some mechs. Extra module slot can be a cool shot to influence the outcome of a close duel, etc.
*ECM count: sometimes, it'd be nice to have an ECM to fall back to and close up on LRM heavy teams with. Or have some amount of the element of surprise.
*Suicide lights: is this guy seriously grinding his Raven 4X by charging headlong into enemies, tagging as many as he can for kill assists?
*Average Damage/Match Score relative to team: Can this guy even do anything?
*Rounds played: Not sure why I ever see a trial mech hanging with the real mechwarriors. Friends of real mechwarriors, I guess.

I'd be up for some kind of leaguing/strictly ranked solo only mode. By leaguing I mean you are told you're in tier 1/2/3/4, and your win/loss earns you standing in the rankings. Maybe you can drop against one league tougher and one league weaker to challenge for promotion or defend your rank, but none of this dropping with people who look like they just installed crap. Or just some transparency on how MM actually works, instead of having us guess like chumps as fate rolls the dice match after match.

 

I understand that matches without respawn works such way that each kill influences win probability. It's just obvious in perspective of numerical advantage and common sense. That's why many of games with matches without respawn use additional mechanics like capping, missions, bombs, flags, etc. and avoiding pure death matches (see MWO, WoT, WarThunder, StarConflict).
However matchmaker sometimes (I don't say I have this every time, but about 1 time of 5, which is pretty annoying too) creates strange matches... He is like:

• “I'll make team with premade and other team without premades, there is no difference anyway.”
• “Why not put all Atlases I have in waiting in one team? That will be fun, lol.”
• “Let me make team of lights only, ha.”
• “Let's make one team of almost only LRMs, and add few spotters for them. Well, and other team will be slow, close-medium range, with 3 AMS for whole team and no ECM and no LRM.”
• “Nothing is better then some morning futility with strong tea in start of my day. Also in middle and other times of day and night as well.”
• “Multitasking is for plebeian, I'm too classy for it. This is why I fill one team and then try to fill another but cannot find with whom fill it, that's why I fill last 1-3 positions with the first comers or do not fill at all. But I don't care. Also, I could swap players between team, but that's for cattle too.”

P.S.: Interesting to note that teams of Atlases are nearly unstoppable despite of slowness. And teams of lights are even more hellish (like in memes about bees and other swarms). But teams of mediums and teams of heavies is not so powerful. It makes me think about medium/heavy class usefulness...
P.S.S.: I'm programmer. The fact that someone would write matchmaking like this has left me in despair.

 

I think you found the answer.
Matchmaking is made by a team of 10-20 subcontractors in Bangalore.
Manually.
When you press the Launch button, you & your build appear on their screen, and they drag & drop you into matches.

 

Something I'm seeing happen more often is a match starting with less than 12 people on one side. Being one down happens all the time, starting 2 down happens sometimes, and I've even started 3 down.

Just throw a warm body wrapped in a mech in there for Pete's Sake. If nothing else they'll absorb some damage.

 

Maybe not enough players available at that moment?

 

No, I work with some of these guys, they're ok. Most likely trying to make the best out of some idiot design decision. That's the problem: If you want to change something, you'll have a risk evaluation for changing the code. A (proper) change is more risky than an (idiot) patch. Hence we're stuck with idiot patches until this game is so fucked up and cornered by own idiocy that they have no other way out than to make a few huge changes.
I'm not very optimistic for MWO - just playing off my premium time (from the Phoenix package) and watching the drama unfold.

 

Marec has point. Because, yes, your (IanSane and Dr. Danger) assumptions about not enough player looks logical at first glance. BUT, why then things getting worse over time if players number is rising (at least launch obviously risen those number because it is... launch, cannot be other way)? (Please read very first post of this topic again, you'll see that it started from stating this problem).

P.S.: I was beta tester, and matchmaking was better for some strange reason. Well, we always can apply something like Hanlon's razor to this situation but I hope it's not that bad (because idiots is nearly worst thing that can happen to any thing... (history of my IRL county is example of this, we even have proverb related to it).

 

i believe that the matchmaking algorithm is totally broken and should be fixed with a better algorithm. I can't believe that we have too few players around the world. With a good algorithm you still can have good matches even with a little player base - you only meet the same guys more often. But do you have the feeling that you meet your opponents again and again? Not really, so player base is not a problem. The Algorithm is just very poor. But the true reason of a bad algorithm is, that the entire "secret" Elo-System is broken as well. Worst thing is that the whole system is totally intransparent. If it would be transparent, the community might give some suggestions. Whatever.

For the moment i'm not annoyed anymore, i just play the game as a fun.training until MWO acquires adulthood with well balanced game mechanics + versatile game options and real competitive Community Warfare really begins. For the Love for Battletech/Mechwarrior... for the sake of Loyalty, Honour and Fun!

 

Actually, there are some (10-20?) people I notice in the same matches with me pretty often.
I don't know if that's because the player base between 1800hZ and 2300hZ is significantly smaller.

[quote author=Blagg Zear link=topic=3727.msg28547#msg28547 date=1384275867]
[...] For the moment i'm not annoyed anymore, i just play the game as a fun.[...] For the Love for Battletech/Mechwarrior... for the sake of Loyalty, Honour and Fun!
[/quote]
Word.