104 replies to this topic

[valkyrie]

I think figured out what the single biggest issue with ELO is. Aside from the obvious tonnage matching issue, it's the fact that for mid to high ranked players, it's now become "cheese or die."

Right now, I'm trying to play a Treb -7M. I'm not doing it for high scores or for street cred, I'm doing it for fun. Normally I don't play LRM boats, especially not one as starved for tonnage as a Treb, but I love my Mediums. I do ok with it, for an LRM boat. A kill here, 400 damage there. Not bad considering I usually roll my Founder's Hunchback with a totally different playstyle. The problem is that I'm usually the only one doing that well, since I've got a pretty good ELO thanks to dropping with Kong (which, despite being a glorified PUG group more than anything, I can at least trust not to use awful builds), occasionally running something horribly OP like a Splatcat or Raven -3L, and generally being a pretty decent pilot.

ELO (to my understanding) matches based on skill and balances teams accordingly. That means I could drop in a game with a moderately good ELO, but be grouped up with a bunch of newbies who have no experience or awful builds, and the other team gets "ok" players who have all just discovered D-DCs, Raven -3Ls, and Jumpphracts. Needless to say, we get slaughtered.

But, that's not the only issue. Now, let's say I'm dropping with Kong. Kong, being a group that discusses builds at length and tends to play fairly often, probably has a moderate ELO score in most cases. Let's say I drop with 3 other Kong members, like we've done every night for months now. I'm still in my Treb. Kong, with a few exceptions, tends to avoid "cheesebuilds." The rest of the community doesn't really share this viewpoint. We drop against an equally skilled team...who is all using cookie cutter cheese builds that everyone on these forums complain about. Again, we get steamrolled.

Now, let's assume PGI gets weight balancing back in the game. Hooray! I drop in my "meh" Treb, and three others drop in usable Catapult -C1s, Hunchback -4SPs, and maybe Vass or Connor drops in a D-DC like usual. The enemy, being equally skilled, knows the ins and outs of the game, and takes a Boomcat, SRM Centbomb, and a D-DC or 6 PPC Stalker. Guess how this one is probably going to end?

See, the problem is twofold - even if you could take the "skill weighting" out of the equation and ensure equally skilled opponents rather than "1 good player and 7 morons vs. 8 ok players", you still have the issue of ELO not balancing builds - at best, it'll balance weight (eventually). In essence, the longer I stay out of a CheeseMech, the more my team suffers, because other players of my skill WILL be using them. Taking anything that isn't a tried and true, borderline-broken build becomes a handicap in matches of equal skill. It's the Chromehounds effect all over again - customization goes out the window because uniqueness gets you killed in most cases.

I'm not naive enough to think this wasn't an issue in Phase 2 matchmaking, but it is worse now. In Phase 2, I could take a Medium 'Mech and know that there's going to be a Medium on the other team. Now? Could be a Medium...or it could be a Splatcat. Or worse. And, in a perfect world, they'll be every bit as good as I am, unrestrained by their choice of 'Mech.

PGI really needs to re-evaluate the use of ELO in MWO. The more I play now, the more I realize it's doing far more harm than good. Right now, it's just going to further homogenize 'Mech choice (especially in PUGs), and we all know stagnation is the quickest way to kill a game.

[Gaan Cathal]

valkyrie, on 26 February 2013 - 08:17 PM, said:

I'm not naive enough to think this wasn't an issue in Phase 2 matchmaking, but it is worse now. In Phase 2, I could take a Medium 'Mech and know that there's going to be a Medium on the other team. Now? Could be a Medium...or it could be a Splatcat. Or worse. And, in a perfect world, they'll be every bit as good as I am, unrestrained by their choice of 'Mech.

Actually, in theory Elo is based on success, not skill. I.e. a higher W/L ratio in a given class of mech gives you a higher Elo score in that weight class. So if you are consistent with your cheese/lackofcheese mech build choices, that should actually be factored in by default, since you would win more with the objectively superior mech (generally, there are skill ceiling issues with some FotM builds). So if Elo is actually working perfectly, then you're better than the Splatcat pilot, he's just being carried by his mech.

[Loxx]

Cheese builds are the easy solution for low to average players who have reached the top of their skill set. They can't survive in anything else, nor do they wish to spend the time getting better. So Instead they choose the easy route and go for cheese.

[valkyrie]

Gaan Cathal, on 26 February 2013 - 08:23 PM, said:

Actually, in theory Elo is based on success, not skill. I.e. a higher W/L ratio in a given class of mech gives you a higher Elo score in that weight class. So if you are consistent with your cheese/lackofcheese mech build choices, that should actually be factored in by default, since you would win more with the objectively superior mech (generally, there are skill ceiling issues with some FotM builds). So if Elo is actually working perfectly, then you're better than the Splatcat pilot, he's just being carried by his mech.

Maybe. The problem with that? There's only so much I can do from behind the controls of an LRMTreb. It's even worse when my higher ELO is being "balanced" by having a team of lower-skilled players. It's putting all of one team's eggs in one or two baskets.

[Mycrus]

Did not one of your guys make that 80+ kill splatcat video?

Anyway, i notice that if you PUG more ELO gives you fairly good games...

If you 4-man more balance is quite skewed... Heck, vassago has been posting this rather vigorously...

Edited by Mycrus, 26 February 2013 - 08:36 PM.

[valkyrie]

Mycrus, on 26 February 2013 - 08:34 PM, said:

Did not one of your guys make that 80+ kill splatcat video?

Anyway, i notice that if you PUG more ELO gives you fairly good games...

If you 4-man more balance is quiet skewed... Heck, vassago has been posting this rather vigorously...

Yeah, Vass was bragging about that for a long time before ELO went live. Still, am I the only one who sees that's an issue? It's like you get punished for playing casually with friends.

[Homeless Bill]

You people can ***** as much as you want, but it's not changing. The number of 8-0 rolls I encounter are perhaps 20% of pre-ELO, even with the tonnage mismatches.

ELO is a boon to newbies, and that's who needed the help. I'd like easy mode where I could play my Spider every now and then, but they need to cater to a larger market - not people like you and me.

[Mycrus]

valkyrie, on 26 February 2013 - 08:35 PM, said:

Yeah, Vass was bragging about that for a long time before ELO went live. Still, am I the only one who sees that's an issue? It's like you get punished for playing casually with friends.

Well, ive noticed that if you play 4s you can't goof around as much and are forced to play seriously...

You can still play balanced mechs, but drops are not forgiving of tactical mistakes...

I get you... playing with high elo players in 4man means the death of casual play...

[Zaptruder]

valkyrie, on 26 February 2013 - 08:27 PM, said:

Maybe. The problem with that? There's only so much I can do from behind the controls of an LRMTreb. It's even worse when my higher ELO is being "balanced" by having a team of lower-skilled players. It's putting all of one team's eggs in one or two baskets.

The only outcome you should be trending towards in the long term is a 50/50% win loss ratio.

If you're losing more than 50% now, it's because your stats have been padded by playing in a less competitive queue prior to Elo. If you're winning more than 50% now, it's because your stats have been depressed through factors that cause the queue to be more competitive than it is on average (i.e. you run 8 mans, or you solo drop against premades regularly).

Everything else is just noise; what mechs they're using, how many people they have in their groups, what playstyle they have, whether or not they use ECM... that doesn't matter.


Yes, on an individual game level, Elo could be more granular to provide a better/more equal experience.... on a macro game level, once player Elo scores have been adjusted through actual matches, then it'll provide the most equal experience possible.

[One Medic Army]

Gaan Cathal, on 26 February 2013 - 08:23 PM, said:

Actually, in theory Elo is based on success, not skill. I.e. a higher W/L ratio in a given class of mech gives you a higher Elo score in that weight class. So if you are consistent with your cheese/lackofcheese mech build choices, that should actually be factored in by default, since you would win more with the objectively superior mech (generally, there are skill ceiling issues with some FotM builds). So if Elo is actually working perfectly, then you're better than the Splatcat pilot, he's just being carried by his mech.

This is why I'm staying clear of cheese.
Indulge once and you'll start to jump your Elo to levels where you will be forced to continue to cheese just to compete.

[EgoSlayer]

valkyrie, on 26 February 2013 - 08:17 PM, said:

I think figured out what the single biggest issue with ELO is. Aside from the obvious tonnage matching issue, it's the fact that for mid to high ranked players, it's now become "cheese or die."


<snip>

ELO (to my understanding) matches based on skill and balances teams accordingly. That means I could drop in a game with a moderately good ELO, but be grouped up with a bunch of newbies who have no experience or awful builds, and the other team gets "ok" players who have all just discovered D-DCs, Raven -3Ls, and Jumpphracts. Needless to say, we get slaughtered.

But, that's not the only issue. Now, let's say I'm dropping with Kong. Kong, being a group that discusses builds at length and tends to play fairly often, probably has a moderate ELO score in most cases. Let's say I drop with 3 other Kong members, like we've done every night for months now. I'm still in my Treb. Kong, with a few exceptions, tends to avoid "cheesebuilds." The rest of the community doesn't really share this viewpoint. We drop against an equally skilled team...who is all using cookie cutter cheese builds that everyone on these forums complain about. Again, we get steamrolled.

Now, let's assume PGI gets weight balancing back in the game. Hooray! I drop in my "meh" Treb, and three others drop in usable Catapult -C1s, Hunchback -4SPs, and maybe Vass or Connor drops in a D-DC like usual. The enemy, being equally skilled, knows the ins and outs of the game, and takes a Boomcat, SRM Centbomb, and a D-DC or 6 PPC Stalker. Guess how this one is probably going to end?

See, the problem is twofold - even if you could take the "skill weighting" out of the equation and ensure equally skilled opponents rather than "1 good player and 7 morons vs. 8 ok players", you still have the issue of ELO not balancing builds - at best, it'll balance weight (eventually). In essence, the longer I stay out of a CheeseMech, the more my team suffers, because other players of my skill WILL be using them. Taking anything that isn't a tried and true, borderline-broken build becomes a handicap in matches of equal skill. It's the Chromehounds effect all over again - customization goes out the window because uniqueness gets you killed in most cases.

And the previous match making prevented these exact same scenarios how?

If 'Cheesebuilds' are so overpowered these player's ELO will rank them up to the point where you won't see them anymore and you'll be matched against other players who are not running these builds. It becomes self correcting.

[Star Captain Obvious Kerensky]

The only way to end the Cheese builds is if ELO takes mech variant into account. A Raven 3L is worth more to a team than a Raven 4X, Raven 2X, or any flavor of spider or Commando. Some variants are just better than others.

PGI has all the data they need to accomplish this as well: Run the ELO algorithm against each variant.

[Kdogg788]

Zero cheese and almost zero difference in win loss, k/d etc. Even my a1 runs some LRMs to be balanced.

-k

[OuttaAmmo NoWai]

So....you just complained about, what, 8 or 9 "cheese" builds? If there are so many builds that you don't approve of, are they really OP? Are they really un-counterable if you can choose ANY one of them and play them yourself?

[valkyrie]

OuttaAmmo NoWai, on 26 February 2013 - 08:57 PM, said:

So....you just complained about, what, 8 or 9 "cheese" builds? If there are so many builds that you don't approve of, are they really OP? Are they really un-counterable if you can choose ANY one of them and play them yourself?

Because they disgust me. We're playing a game that gives you almost complete control of your ride, and I'm supposed to play one of a handful of "proven" builds just to avoid dying a horrible, fiery death in every match? **** that.

[Mawai]

Well ... there are 8 mechs on the team. You don't need cheese builds to win ... you need teamwork

Almost any mech can do decent damage when played well (though I have particular trouble with the founder's hunchback ... the 4P and 4SP just play much better for me .. what build do you use for the 4G?)

Finally, if ELO is working correctly, everyone will end up with a W/L ratio of about 1:1. You will win more matches than you lose until you are matched up appropriately. (If your initial ELO is higher than your skill then you will lose much more than win until your ELO drops to where it should be). The matchmaker probably uses some sort of ELO multiplier for grouped players or queues groups separately from lone wolves. Also, as someone mentioned above .. if you play the mechs and builds you like then this will be factored into your ELO ... when trying new stuff out then you can expect to have a tougher time.

[Kassatsu]

I'd love to see a proper battle value system in place that weighs each mech's loadout, or at the very least, it's average performance in a match based on what variant it is and adds a value to the team based on that, which is then balanced out on the other team.

Purely as an example, let's say a Raven 3L has a rating of 10, you use your 3L and add 10 points to your team, so the enemy team needs another 10-point mech to balance it out. Someone else on your team uses a Spider, which has a point value of 2. Someone on the enemy team is 2, and so on.

Perhaps have a personal rating for each variant as well, that can increase or decrease your score based on your own past performance (either above, below, or about equal to the average performance of that variant).

@Mawai see my sig for my 4G build.

[valkyrie]

Kassatsu, on 26 February 2013 - 09:02 PM, said:

I'd love to see a proper battle value system in place that weighs each mech's loadout, or at the very least, it's average performance in a match based on what variant it is and adds a value to the team based on that, which is then balanced out on the other team.

Purely as an example, let's say a Raven 3L has a rating of 10, you use your 3L and add 10 points to your team, so the enemy team needs another 10-point mech to balance it out. Someone else on your team uses a Spider, which has a point value of 2. Someone on the enemy team is 2, and so on.

Perhaps have a personal rating for each variant as well, that can increase or decrease your score based on your own past performance (either above, below, or about equal to the average performance of that variant).

@Mawai see my sig for my 4G build.

This is probably the only way to get things in order without going to BV balancing, which PGI isn't going to do. This is far more of an issue than just "get gud scrub" like people want to believe. What about the poor ******** getting pulled "up" by high ranked players screwing around in new builds?

[Colin Thrase]

I believe the current ELO system is much improved. I felt like my teams were getting stomped constantly prior to the last revision, but now my kills/deaths ratio and wins/losses ratio have been consistently improving since the latest change. I realize some of those earlier games were due to learning curve (I'm not a hard-core gamer and I had a lot to learn early on), but I feel like I have now played enough games that I'm an average/fair pilot, and with the new ELO I feel like I'm seeing a roughly even number of wins/losses, so I'm probably on teams with and playing against other average/fair pilots.

I think the most likely problems with ELO are for completely New players and for Elite players:

New players: Because some new players may have no experience at all with shooters, and probably need a lot of practice, while other new players are probably pretty good, having played other sims & shooters and having good reflexes. ELO would probably rank both equally at first, since it has no data for the new pilots, and that player will be thrown in a group with players of dissimilar skill levels.

Elite players: I would imagine that once you get to a certain level of skill (top 5% or so), that the odds of other Elite players being online the same time as you are slim, so you'll get assigned to a group of players whose skill isn't on par with yours. I sympathize with them, because this is probably frustrating.

I don't believe PGI will be able to solve either of the above easily.

[Franchi]

EgoSlayer, on 26 February 2013 - 08:49 PM, said:

And the previous match making prevented these exact same scenarios how?

If 'Cheesebuilds' are so overpowered these player's ELO will rank them up to the point where you won't see them anymore and you'll be matched against other players who are not running these builds. It becomes self correcting.

Previous matchmaking limited this to a certain extent because if you play a medium the enemy team got a medium, still didn't balance 3L's versus 1d's or Catapults versus dragons, but you knew that whatever weight class you brought the enemy team was going to have one.