43 replies to this topic

[A Large Infant]

Roadkill, on 04 April 2015 - 06:41 PM, said:

Close, but not quite.

Each player can have an Elo rating for each piece of equipment when used on every variant. Yes, that's a crazy number of Elo ratings for each player.

Your formula assumes that the combat value of a particular weapon is the same for everyone, which is obviously not true. I totally suck with the Gauss Rifle because I simply cannot get used to the charge mechanic. Lasers are another example that's pretty obvious - some people are very good at holding lasers on target for their entire burn time, while others use them to spray & pray.

Your formula would probably be better than what we have, but it wouldn't be as good as it could be. And if we're going to start changing code, we might as well go whole hog.

Although you may not be good with gauss, if you are a highly skilled player you will still understand it, employ it to its strengths and hence fall closer to the average gauss user than you might think. because some will be truly awful with gauss and bring its combat value down.

I'm not sure any system can be perfect but there has to be a way to separate

1) Truly great players with an awful build
2) Awful players with a great build
3) Great players with a great build
4) awful players with an awful build

and all inbetween

[Roadkill]

I don't remember who originally posted the idea, but one good way to help the matchmaker would be:

1) An overall Elo for every player. This is their "base" Elo and is used when no other Elo rating is available.

2) A weight class Elo rating for each weight class. Once you have 50 matches in a given weight class, that Elo rating is used instead of the overall Elo rating.

3) A chassis Elo rating for each Mech chassis. Once you have 50 matches in a given chassis, that Elo rating is used instead of the weight class Elo rating.

4) A variant Elo rating for each Mech variant. Once you have 50 matches in a given variant, that Elo rating is used instead of the Mech chassis Elo rating.

Only the most specific Elo rating is used for matchmaking purposes. However, every applicable Elo rating would be modified after every game. So if, for example, I have 50+ matches in the Stormcrow Prime, then each subsequent game would use my Stormcrow Prime Elo rating for matchmaking purposes and then modify my Stormcrow Prime Elo rating, my Stormcrow Elo rating, my Medium Mech Elo rating, and my overall Elo rating after the match.

How does this help? Well, the more you play the more accurate your overall Elo rating becomes as a general baseline for your skill. So when you buy a new Mech, you at least have that baseline to start from. But if it's a new Medium Mech and you play Mediums all the time, your Medium Mech Elo rating will be a better representation of your skill in Medium Mechs and so would be used while you're skilling up the new purchase. Or similarly if the new purchase is a new variant of your favorite Medium, the Stormcrow, you'd use your Stormcrow Chassis Elo rating until the new Mech had 50 matches of its own.

Perfect? No. You're using a well-developed Elo rating for a new (and unskilled) Mech whenever you buy a new one. But it'd be a lot more accurate than what we currently have, and shouldn't be that difficult to implement.

[Alistair Winter]

Meeso Thorny, on 04 April 2015 - 06:47 PM, said:

Does this not already occur?

Yeah, 100%. But if we're looking for a hypothetical solution (PGI will never do any of this anyway), then it should preferably not replicate existing weaknesses, in my opinion.

[jlawsl]

I would say to just drop the whole elo crap and simple go off of the amount of games played. If you suck by your 250-500th game, well, you deserve to be point bait. They should stop trying to match supposed skill of a single person to a team game. We see it all the time, guys you watch after you die, and you wonder two things-How am I already dead and this ground shooting, non hitting, overheating guy is still alive? Also, why can't I ever run into someone that terrible?(and we know them when we do).

Or, if you want to make it off of some sort of skill, base it on total c bills earned or xp earned or even average match score. All of those take into account your contribution to a match and not just a win/loss ratio. There is so much data on your profile that could be used to put you into a bracket of like skilled players other then this elo garbage.

[Sjorpha]

Meeso Thorny, on 04 April 2015 - 06:52 PM, said:

I'm not sure any system can be perfect but there has to be a way to separate

1) Truly great players with an awful build
2) Awful players with a great build
3) Great players with a great build
4) awful players with an awful build

and all inbetween

How would you program that? I think you are underestimating how hard it would be to create automatic build evaluation, I'm not sure it is possible to teach a computer how and why a build is currently good in a evolving metagame.

It would not only have to understand what weapon loadouts are effective and on which mechs, but also:
ammo placement, heat management techniques, weapon combinations, how ammon needs varies for different combination of weapons and for primary versus backup weapons (how does it know which is the primary weapon and which is the backup?), which builds are strong on their own and which are only good combined with certain other builds on the team, which engines are tonnage efficient for which build, which mechs are XL friendly and what builds can use an XL in otherwise XL unfrendly mechs, How zombie capability is valued in relation to other stats and a gazillion other parameters. Not even many human players are good at evaluating builds.

In my opinion there is no need for it anyways, building effective mechs is a skill in itself, so if a good player drops in a bad build he is either not that good, or he is handicapping himself for the sake of challenge (in which case there is no reason to compensate for it). There is no reason why the strength of different builds on the same chassis should not simply be considered a part of the players skillset, and therefore it's fine to leave it rolled into ELO.

[A Large Infant]

I have already posted two methods for the computer to learn which builds are stronger. It is not a preemptive system, it evolves with equipment usage.

[Sjorpha]

Meeso Thorny, on 05 April 2015 - 01:15 AM, said:

I have already posted two methods for the computer to learn which builds are stronger. It is not a preemptive system, it evolves with equipment usage.

I missed that, I've read them now. Not sure I'm convinced it would work but let's say for the sake of discussion that it would.

There are still two basic problems with the whole idea of compensating for strength of builds.

The first is that building and choosing the right mech for the job is every bit as much part of a players skillset as aim, torso twisting, positioning etc. What exactly is the difference between a stupid decision in the mechlab and a stupid decision on the battlefield? Why should one misstake be compensated for and not the other? Why shouldn't dropping in a XL stalker be as harshly punished as standing in the open or missing all your shots? It's an equally stupid thing to do after all.

The second is that it would obscure imbalances. If bad mechs are given easier matchups then they will seem better than they actually are, and that in turn would make it harder to know where buffs and nerfs are needed.

Now we can easily identify a bad mech chassis, like the awesome, precisely because the game can not. It is by demanding equal performance ton for ton from all mechs that we can push for better balance, not by introducing mech paralympics.

Edited by Sjorpha, 05 April 2015 - 02:56 AM.

[A Large Infant]

If it were to become clear the Awesome is an inferior chassis (which it currently is), a team could hypothetically get 3 timberwolves and 3 Awesomes to make up for it. To a developer it would still be clear the Awesome is an inferior chassis and realistically it would show on the battlefield as well.

[DrSlamastika]

Whats ELO? You mean in this game rly?

[Sjorpha]

Meeso Thorny, on 05 April 2015 - 03:11 AM, said:

If it were to become clear the Awesome is an inferior chassis (which it currently is), a team could hypothetically get 3 timberwolves and 3 Awesomes to make up for it. To a developer it would still be clear the Awesome is an inferior chassis and realistically it would show on the battlefield as well.

I'm not sure you understood my post, because I sure as hell don't understand how this answers it?

[Hotthedd]

Wouldn't this allow skilled players to start off in a lower tier Elo bracket every time they bought a new mech?

[Mawai]

Meeso Thorny, on 04 April 2015 - 12:31 AM, said:

the matchmaker needs to take into account combat value of the equipment brought, period.

if you choose to bring a 19kph, armorless, weaponless stalker you should have hypothetically guaranteed a spot for someone else on your team, who is the strongest player in MWO using the strongest equipment, to compensate

No.

1) I agree that it would be great if the MM could take fits as well as skills and ability into account by combining battle value, experience unlocks and Elo to obtain some estimate of combat effectiveness.

2) The Matchmaker has never combined high and low Elo players (except in very rare cases) and never should ... if you gimp your team with a terrible fit then you should find yourself grouped with folks who evaluate to equally terrible based on battle value, Elo and experience ... the matchmaker should not balance bad with good ... and it doesn't in most cases.

[Fragnot]

Very sound ideas OP. Those suggestions do a nice job of customizing Elo functions to this game's specific needs. I hope PGI sees this and thinks about it.

[InspectorG]

DrSlamastika, on 05 April 2015 - 03:14 AM, said:

Whats ELO? You mean in this game rly?

Its a system to push REALLY good players to be better by running them through a simulated Hell.

[LordKnightFandragon]

Roadkill, on 04 April 2015 - 07:04 PM, said:

I don't remember who originally posted the idea, but one good way to help the matchmaker would be:

1) An overall Elo for every player. This is their "base" Elo and is used when no other Elo rating is available.

2) A weight class Elo rating for each weight class. Once you have 50 matches in a given weight class, that Elo rating is used instead of the overall Elo rating.

3) A chassis Elo rating for each Mech chassis. Once you have 50 matches in a given chassis, that Elo rating is used instead of the weight class Elo rating.

4) A variant Elo rating for each Mech variant. Once you have 50 matches in a given variant, that Elo rating is used instead of the Mech chassis Elo rating.

Only the most specific Elo rating is used for matchmaking purposes. However, every applicable Elo rating would be modified after every game. So if, for example, I have 50+ matches in the Stormcrow Prime, then each subsequent game would use my Stormcrow Prime Elo rating for matchmaking purposes and then modify my Stormcrow Prime Elo rating, my Stormcrow Elo rating, my Medium Mech Elo rating, and my overall Elo rating after the match.

How does this help? Well, the more you play the more accurate your overall Elo rating becomes as a general baseline for your skill. So when you buy a new Mech, you at least have that baseline to start from. But if it's a new Medium Mech and you play Mediums all the time, your Medium Mech Elo rating will be a better representation of your skill in Medium Mechs and so would be used while you're skilling up the new purchase. Or similarly if the new purchase is a new variant of your favorite Medium, the Stormcrow, you'd use your Stormcrow Chassis Elo rating until the new Mech had 50 matches of its own.

Perfect? No. You're using a well-developed Elo rating for a new (and unskilled) Mech whenever you buy a new one. But it'd be a lot more accurate than what we currently have, and shouldn't be that difficult to implement.

Your overall ELO should be used at all times. The MM should then seek to put an equal number of similar elo ratings on both sides, rather then stacking the deck, putting 9 150s, and 3 100s vs a team of sub 30 elo rating players.

Rather it should be

6 100-150 elo rating players on each side)
2 80-100 elo players on each side

and so on, so each side gets a near equal number and range of skill and elo......then games of 1-12 might diminish a little.

RIght now its as if the NFL goes up against a high school football team and mercies them by the end of the 1st quarter.....cuz games in MWO rarely last longer then maybe 5 minutes...1-12, 2-12.....the like...its clear to see when the elo didnt epic fail. Then we see games of 7-12, 8-12, 10-12, they come down to the last 3-5 minutes and longer. People move together, focus fire, dont stupicide rush, dont bot, dont afk, dont connect then leave and never play.......MM needs to separate by elo as well...

just examples but like

0-75(scrub elo Q)
76-100(slightly less scrub Q)
101-150(Avg Q)
151-200(Good Q)
200+(MLGPRO Q)

And look to place players only in that Q. If it must compromise, it will search to put 1 scrub of similar elo per side, not just gimp 1 side with a trio of 9 elo rating players...

[Omaha]

Step 1) Kill some mechs.


Step 2) Play another match.


Step 3) Figure out whats wrong with your loadout.


P.S. Sorry was looking through an old toybox today, and I found this! At first I was like OMG!! (Teays of joy) Then I was like NOOOOOOOO! (Tears turn to blood) Reminds me of ELO. Had to share. GLHF out there!

[LordKnightFandragon]

Omaha, on 05 April 2015 - 07:24 AM, said:

Step 1) Kill some mechs.


Step 2) Play another match.


Step 3) Figure out whats wrong with your loadout.


P.S. Sorry was looking through an old toybox today, and I found this! At first I was like OMG!! (Teays of joy) Then I was like NOOOOOOOO! (Tears turn to blood) Reminds me of ELO. Had to share. GLHF out there!

Can I drive that in MWO? Load up 4 MPL on it and go to town. Is it actually a 50t medium? Cuz for w/e reason, in MW4 Mektek, the Rifleman was a medium mech, not a like 70t heavy...

[Bulletsponge0]

Agent 0 Fortune, on 03 April 2015 - 10:06 PM, said:

Elo per chassis not weight class, possibly per variant.
Elo modifiers for skill unlocks
Elo boundaries, limit the min and max Elo boundary in a match, if match cannot create, then drop to 8v8 as opposed to grabbing players outside the boundary.

Obviously this will only impact Elo Dependant queues, but considering 85% of the playerbase is PUGs, seems like a reasonable consideration.
Using Elo to offset bad mechs will work in tandem with quirks, so development will need to determine if every mech must be CW competitive level or if mechs will fall in the Elo bracket as determined by pilot skill and chassis.

I'd be in favor of a simple change...

instead of building teams with similar ELO averages, place players with similar ELO's on each team

but i fear the player base isn't big enpugh for that

[Omaha]

No its not a rifleman, hehe its a Raidar X!

[Weeny Machine]

Why an ELO at all? I mean it seems that I end up sometimes with great players and often I am in ELO hell where it is 0:6 before you can say "WTF BBQ?!" I doubt the absence of a matchmaking would make it worse. Especially since I like to swap my mech after a couple of games anyway and love to play non-meta mechs and builds and also test new stuff