69 replies to this topic

[Mergatroid Skittle]

Problem: Too many games end up 12 - 0

Solution: Get rid of ELO-based matchmaking

Most people seem to agree with this and I just wanted to voice it on the forums. Peace out.

[MischiefSC]

No they don't agree with that. If they do they are wrong.

We had Elo-less matchmaking before. It was horrible.

We're never going back there.

Peace out.

[Mystere]

MischiefSC, on 15 August 2014 - 11:16 PM, said:

No they don't agree with that. If they do they are wrong.

We had Elo-less matchmaking before. It was horrible.

We're never going back there.

Peace out.

Honestly, I preferred the excitement of the old non-matchmaker we had.

[MischiefSC]

Mystere, on 15 August 2014 - 11:21 PM, said:

Honestly, I preferred the excitement of the old non-matchmaker we had.

I miss having a 90%+ win/loss and a KDR of 3 or 4 with pretty much any mech I wanted to take, because dropping in a 4man even without VOIP was like driving a steamroller through a herd of ducklings.

It's alright that you miss that. It was just bad for new and old player alike otherwise though.

[Aresye]

Mergatroid Skittle, on 15 August 2014 - 11:03 PM, said:

Problem: Too many games end up 12 - 0

Solution: Get rid of ELO-based matchmaking

Most people seem to agree with this and I just wanted to voice it on the forums. Peace out.

The stomps usually have nothing to do with Elo. Believe it or not, most of it is teamwork.

Now, there's other things that factor in, such as the shared targeting information, which makes it pretty much impossible to make a comeback if you're the last guy because once you fight the first mech you see, his whole team knows where you are.

People don't stay alive. A lot of times people just try to fight it out when they get into a bad situation instead of trying to get away. Sure, they might get lucky and do some decent damage, but no matter what, you can do MORE damage if you can escape, regroup with the team, and continue from there. If you watch any competitive team, you'll see that as soon as one of them gets crit, they fall back and allow another player to replace them. Meanwhile they fall back behind the group and continue to do damage.

[Kiiyor]

I just want closer games. Things have gone downhill since 4x3.

[MischiefSC]

Kiiyor, on 15 August 2014 - 11:32 PM, said:

I just want closer games. Things have gone downhill since 4x3.

How did you get a comparable pre/post sample size, and what were the population breakups?

One thing that is true - when you remove 100 tons or so from each team in total tonnage you create a more significant cluster effect.

[Kiiyor]

MischiefSC, on 15 August 2014 - 11:39 PM, said:

How did you get a comparable pre/post sample size, and what were the population breakups?

One thing that is true - when you remove 100 tons or so from each team in total tonnage you create a more significant cluster effect.

I collated around 100 matches worth of data a while ago, and did an analysis. The community pitched in and now i've got a fairly decent number of matches. I've almost finished the comparisons of the new data, but I haven't fully doublechecked the numbers yet, as i've already found an error with one of the charts above.

[JediMechwarrior]

Kinda disagree 100% on that one good sir. No elo is not the solution.

[Kiiyor]

Oh, class balance. Here:

[VonFrundsberg]

BV 2.0

[MischiefSC]

Kiiyor, on 15 August 2014 - 11:46 PM, said:

I collated around 100 matches worth of data a while ago, and did an analysis. The community pitched in and now i've got a fairly decent number of matches. I've almost finished the comparisons of the new data, but I haven't fully doublechecked the numbers yet, as i've already found an error with one of the charts above.

My concern is that balance changes over the last 3 months would have a huge swing in populations. Was this all post-Clan release for the pre-4x3? Clans introduced a big upswing in IS lights for example - killing Clan lights and Dires was something they could do and do well. AC nerfs changed weight populations. Also the drive to 4x3 shifted populations.

I'm just curious about the population changes and what drove them.

[Kiiyor]

MischiefSC, on 15 August 2014 - 11:58 PM, said:

My concern is that balance changes over the last 3 months would have a huge swing in populations. Was this all post-Clan release for the pre-4x3? Clans introduced a big upswing in IS lights for example - killing Clan lights and Dires was something they could do and do well. AC nerfs changed weight populations. Also the drive to 4x3 shifted populations.

I'm just curious about the population changes and what drove them.

Sorry, posted above.

Though things appear more evenly spread across the weight ranges, the Clans have stayed relatively top heavy, while the IS has more lights and mediums.

IS lights compete well with Clan lights (i've got damage figures also) but poorly against the legions of TW's and DW's making up the bulk of Clan forces.

IS mediums.... are death traps. Death traps armed with nerf guns, and armoured with polystyrene. They have the highest mortality rate and the lowest contribution of all the classes.

More IS mediums means more easy kills for heavier mechs, both IS and Clan.

Edit: Yep, the pre 4x3 match sample was taken after the clans dropped.

Edited by Kiiyor, 16 August 2014 - 12:04 AM.

[MischiefSC]

Kiiyor, on 16 August 2014 - 12:02 AM, said:

Sorry, posted above.

Though things appear more evenly spread across the weight ranges, the Clans have stayed relatively top heavy, while the IS has more lights and mediums.

IS lights compete well with Clan lights (i've got damage figures also) but poorly against the legions of TW's and DW's making up the bulk of Clan forces.

IS mediums.... are death traps. Death traps armed with nerf guns, and armoured with polystyrene. They have the highest mortality rate and the lowest contribution of all the classes.

More IS mediums means more easy kills for heavier mechs, both IS and Clan.

Edit: Yep, the pre 4x3 match sample was taken after the clans dropped.

Figured. Back before Clans you couldn't find a Medium with a dowsing rod and Lights were either 4man wolfpacks or nothing.

Also, the pug/team queues split. That shifted a lot of weight away from the optimal builds (read - heavier) and made sub-optimal builds (read - mediums and most lights) more common by a population percentage.

IS Mediums are a special sort of bad, thats' absolutely true.

[RedDragon]

MischiefSC, on 15 August 2014 - 11:25 PM, said:

I miss having a 90%+ win/loss and a KDR of 3 or 4 with pretty much any mech I wanted to take, because dropping in a 4man even without VOIP was like driving a steamroller through a herd of ducklings.

Which has nothing to do at all with Elo but with the different queues.

Elo is a terrible balance factor in a team game and that should be quite obvious. How can you improve your Elo when you never drop with a group? What exactly are the chances that you can reliably affect the outcome of every match (or at least most matches) when you are only 1/12th of your team's force? The chances that you are paired with exceptional dumb/good players is far higher than the chance that you really have a meaningful impact in the game, thus skewing the results.

[TheCaptainJZ]

ELO works fine for me.

[Kiiyor]

MischiefSC, on 16 August 2014 - 12:19 AM, said:

Figured. Back before Clans you couldn't find a Medium with a dowsing rod and Lights were either 4man wolfpacks or nothing.

Also, the pug/team queues split. That shifted a lot of weight away from the optimal builds (read - heavier) and made sub-optimal builds (read - mediums and most lights) more common by a population percentage.

IS Mediums are a special sort of bad, thats' absolutely true.

Yep. They do have some VERY competitive chassis, but they are weighed down by the ludicrous badness of many of their variants.

[MischiefSC]

RedDragon, on 16 August 2014 - 12:26 AM, said:

Which has nothing to do at all with Elo but with the different queues.

Elo is a terrible balance factor in a team game and that should be quite obvious. How can you improve your Elo when you never drop with a group? What exactly are the chances that you can reliably affect the outcome of every match (or at least most matches) when you are only 1/12th of your team's force? The chances that you are paired with exceptional dumb/good players is far higher than the chance that you really have a meaningful impact in the game, thus skewing the results.

Do you really need me to explain this to you? I've done it 30 or 40 times if not more on a variety of threads. I should probably just write up a token rundown of it and copy/paste it for these threads.

Short version - you're going to get dropped with and against nubs either way. The only difference Elo makes in that is creating a degree of balance between teams.

Your impact on any given match are ~8%. What you make of that 8% is up to you. Elo serves only to keep your 8% within a similar group of players skill levels, so that you're not stomping nubs or getting rolled by Golden Master Race Elites.

Pug or premade was significant - but even without that it meant the good players, regularly dropping against nubs, could easily rock a KDR of 7 or even 12. Why? Because you've eliminated any skill balancing in the population which, in turn, limits the ability for people to gain skill. They're not playing better people, they're either seals getting clubbed or they're clubbing seals, match after match.

Kiiyor, on 16 August 2014 - 12:39 AM, said:

Yep. They do have some VERY competitive chassis, but they are weighed down by the ludicrous badness of many of their variants.

Fundamental game design flaw for IS -

You have to play in food-grade trash variants to level up the decent ones. It inherently creates purpose-built bad mechs on the field and people having to grind them.

It's a bad concept all the way around.

[Duke Nedo]

sneeking, on 15 August 2014 - 11:31 PM, said:

I suspect at times solo droppers are been fed to teams who are on ts, it doesn't always seem like it just at times I feel kind of exploited.

I can attest to that as a 100% solo dropper you at times get totally abused by the mm. This weekend i went from +100 wins vs. losses down to +50 wins vs losses. 10 match long loss streaks, in total some 25% wins and about 100 drops. I can guarantee that i did not suck worse than usual this weekend and it was not fun. This happens from time to time and it has to do with the mm somehow. NB solo drops only. Most of these games i was top 3 on my team and tried to give useful pointers in chat for what its worth... Subjectively, there could possibly be a correlation with which side of the map i dropped, maps like alpine, tourmaline, crimson, river city, forest colony and frosen are rather one sided... Anyways, this happens too often to be random bad luck by any statistical significance....

On a side note, I have a feeling that bad side balance on several maps reinforce the # of stomps in this game but its rarely discussed...

[Flyby215]

New players have lower Elo to begin with?

I've found myself in Elo hell occasionally... I actually have screenshots where only 3 players on my team actually break the 100 dmg mark... myself exceeding 500 obviously

The absurdity of couse is that weight classes share Elo. My Timberwolves can easily break 1,000 dmg and 3-5 kills without even really thinking. Yet, for some strange reason, the Timberwolf Elo is shared with my Dragon Elo! Uh-huh... so a Dragon is apparently equal in capability to the Timberwolf... according to the Elo system that is. ???

I suppose the plus side of all this is that by leveling my Dragons I am dragging my Elo WAY down, and so when I play my Timberwolf (strictly for c-bills since it has long since been mastered) I can completely obliterate all foes that stand before me.