56 replies to this topic

[Duke Nedo]

PGI could/should mine their already huge set of data in their match logs to adjust the MM, introducing ecm and mech tiers. Problem is that it takes quite a lot of effort... I mean preferably a neural-network should be trained to predict the impact of ecm/mech tiers, which would be a big investment only for solo-queue matchmaking so I am not holding my breath...

The main goal is CW I assume, which doesn't need MM, so...

There is however one thing I would really like if they gave a shot. If they would select the 24 players just like they do now, then before launching the game just let the MM check a few iterations of swapping players between the teams and see if balance could be improved by 1. this time ignoring Elo, just assume that all players picked by the MM for this match have comparable skill (Elo should be wildly inaccurate anyways when it comes to predicting skill), and 2. factor in mech tiers and ecm. It would not affect the matchfinding time (this would take milliseconds at most) and could be used to quite crudely balance ecm and mech selection between the teams, still maintaining mech class distribution.

Edited by Duke Nedo, 04 March 2015 - 02:13 AM.

[PurpleNinja]

For this challenge only I switched to the following guidelines (high rate of qualifying matches).

• Pick a mech with high alpha or a mech which can sustain constant damage output (DWF and KGC are my favourites).
  
• Hit everyone with a couple of laser sweep (sometimes I can hit as much as 6 mechs with a single laser fire).
  
• Patiently wait behind your teammates and let then absorb incoming damage.
  
• As soon as you see an open CT, begin shooting, non stop, through your teammates (they should be aware of your firepower).
  
• Blame your teammates for standing right in front of you when you opened fire.
  
• After you see "KILLING BLOW" and "KILL ASSIST" shout "LEEEEEEEEEROY" in global chat and then charge the rest of enemy team with shutdown overridden.
  
• After you die, state how awful your team is for shooting your back and let you die on your final heroic charge (this is important to create a sense of guilty in other players so they are less likely to do the same thing you do).

Edited by PurpleNinja, 04 March 2015 - 02:43 AM.

[Joseph Mallan]

PurpleNinja, on 04 March 2015 - 02:42 AM, said:

For this challenge only I switched to the following guidelines (high rate of qualifying matches).

• Pick a mech with high alpha or a mech which can sustain constant damage output (DWF and KGC are my favourites).
  
• Hit everyone with a couple of laser sweep (sometimes I can hit as much as 6 mechs with a single laser fire).
  
• Patiently wait behind your teammates and let then absorb incoming damage.
  
• As soon as you see an open CT, begin shooting, non stop, through your teammates (they should be aware of your firepower).
  
• Blame your teammates for standing right in front of you when you opened fire.
  
• After you see "KILLING BLOW" and "KILL ASSIST" shout "LEEEEEEEEEROY" in global chat and then charge the rest of enemy team with shutdown overridden.
  
• After you die, state how awful your team is for shooting your back and let you die on your final heroic charge (this is important to create a sense of guilty in other players so they are less likely to do the same thing you do).

[Brother MEX]

PurpleNinja, on 04 March 2015 - 02:42 AM, said:

For this challenge only I switched to the following guidelines (high rate of qualifying matches).

• Pick a mech with high alpha or a mech which can sustain constant damage output (DWF and KGC are my favourites).
  
• Hit everyone with a couple of laser sweep (sometimes I can hit as much as 6 mechs with a single laser fire).
  
• Patiently wait behind your teammates and let then absorb incoming damage.
  
• As soon as you see an open CT, begin shooting, non stop, through your teammates (they should be aware of your firepower).
  
• Blame your teammates for standing right in front of you when you opened fire.
  
• After you see "KILLING BLOW" and "KILL ASSIST" shout "LEEEEEEEEEROY" in global chat and then charge the rest of enemy team with shutdown overridden.
  
• After you die, state how awful your team is for shooting your back and let you die on your final heroic charge (this is important to create a sense of guilty in other players so they are less likely to do the same thing you do).

I am already doing some of the things you suggest, or similar ... especially the OVERHEAT, to deny the enemy team the satisfaction of killing me ... because I am so UPSET about the unfair MM !

Edited by Brother MEX, 05 March 2015 - 01:44 AM.

[Sjorpha]

Roughly the following info was given by a PGI MM programmer interviewed in one of the NGNG podcasts.

The matchmaker already tries to reach as low standard deviation of ELO as possible, which means it tries to put you with equally skilled players. But it often has too few players to work with.

The more extreme the ELO of a player, whether extremely high or extremely low, the harder it is for the MM to find two teams of similar skill. There is simply not enough players that bad or good to make complete matches. Therefore these players experience the "carry harder" or "fifth wheel" phenomenon, as at a certain extreme level of good you start being dropped with increasingly bad players, and vice versa. This isn't ideal of course but fixing it would require either a much bigger population or less player choice, as described later in this post.

There is also two very persistent myths about the matchmaker that should be dispelled.

The first is that "roflstomps" are a sign of bad matchmaking, that is not true. The somewhat counter intuitive truth is that the chance of stomps goes UP the more even in terms of skill the teams are internally. This is (plausibly) because teams of even players tend to succeed of fail together, either they manage to coordinate or not, and the team that clicks will roll the other. Whereas in very uneven teams the good players on both sides will just kill the bad players on the other side and the match ends up looking more "even" in terms of kills.

The second myth is that streaks of wins or losses are a sign of bad or "rigged" matchmaking. The opposite is true, any random system of sufficient size will feature predictable quantities of extreme results. If no one ever got extreme streaks of wins or losses then that would be evidence of a rigged matchmaker. So either we want a rigged matchmaker or we accept the occurrence of extreme win/loss streaks. You can't have the cake and eat it in this regard.

The last thing to ponder it that as player choice goes up, matchmaking quality goes down. There is no avoiding this tradeoff and MWO currently has a lot of player choice in terms of game mode, group size, builds and two separate ques. There is only so much that can be done with matchmaking while combining extensive player control with a small population. If we gave up game mode selection and limited group que sizes to 4, 8 and 12 the quality of matchmaking would increase considerably. I am personally in favor of those changes, but many aren't.

Edited by Sjorpha, 05 March 2015 - 02:30 AM.

[Rossario x Vampire]

PurpleNinja, on 04 March 2015 - 02:42 AM, said:

For this challenge only I switched to the following guidelines (high rate of qualifying matches).

• Pick a mech with high alpha or a mech which can sustain constant damage output (DWF and KGC are my favourites).
  
• Hit everyone with a couple of laser sweep (sometimes I can hit as much as 6 mechs with a single laser fire).
  
• Patiently wait behind your teammates and let then absorb incoming damage.
  
• As soon as you see an open CT, begin shooting, non stop, through your teammates (they should be aware of your firepower).
  
• Blame your teammates for standing right in front of you when you opened fire.
  
• After you see "KILLING BLOW" and "KILL ASSIST" shout "LEEEEEEEEEROY" in global chat and then charge the rest of enemy team with shutdown overridden.
  
• After you die, state how awful your team is for shooting your back and let you die on your final heroic charge (this is important to create a sense of guilty in other players so they are less likely to do the same thing you do).

[Insects]

Duke Nedo, on 04 March 2015 - 02:11 AM, said:

PGI could/should mine their already huge set of data in their match logs to adjust the MM, introducing ecm and mech tiers. Problem is that it takes quite a lot of effort... I mean preferably a neural-network should be trained to predict the impact of ecm/mech tiers, which would be a big investment only for solo-queue matchmaking so I am not holding my breath...

.

Perhaps they have analyzed that data and found that ECM in fact makes no difference to team win/loss ratios?
Would be quite easy to generate a graph of matches with 1 ECM vs 0. 2-0, 3-0... And straight away see if there is an increased win rate.
Perhaps it is even an increased loss rate?
ECM is often on a lone light sniper offering no advantage to the team, very possible it has little influence.

I dont really get ECM MM complaining, there is actually a way to ensure your team has 100% chance of a friendly ECM providing mech every game, which should ensure victory if ECM as OP as gets claimed.
Guess what that secret trick is?

Edited by joelmuzz, 05 March 2015 - 02:56 AM.

[Duke Nedo]

joelmuzz, on 05 March 2015 - 02:54 AM, said:

Perhaps they have analyzed that data and found that ECM in fact makes no difference to team win/loss ratios?
Would be quite easy to generate a graph of matches with 1 ECM vs 0. 2-0, 3-0... And straight away see if there is an increased win rate.
Perhaps it is even an increased loss rate?
ECM is often on a lone light sniper offering no advantage to the team, very possible it has little influence.

I dont really get ECM MM complaining, there is actually a way to ensure your team has 100% chance of a friendly ECM providing mech every game, which should ensure victory if ECM as OP as gets claimed.
Guess what that secret trick is?

Yeah, perhaps. Would be interesting read however it looks.

Though, I know for a fact that I prefer to have 3 ecm on my team and 0 on the other rather than reverse. I am also 100% sure I prefer to have Direwolves and Crabs in the assault lance rather than Gargoyles and Highlanders.... not saying you never get surprised, but more often than not you get what you expect.

Edited by Duke Nedo, 05 March 2015 - 03:15 AM.

[IG 88]

after 2 min the matchmaker open his door wider.

[Insects]

Duke Nedo, on 05 March 2015 - 03:15 AM, said:

Though, I know for a fact that I prefer to have 3 ecm on my team and 0 on the other rather than reverse.

3 ECM Commando or MystLynx vs 3 Firefarters on a team?
That probably depends on the map, Caustic will always favor LRM and ECM countermeasures to them. Maps with more cover will be the FS9 playground.

[Norbaer HALL]

Another frustrating day where the UNFAIR matchmaker only allowed me A SINGLE WIN of all the games played today ( I stopped counting after 10 losses in a row ).

I think I will play other games now as it isnt fun that the matchmaker ALWAYS puts me into the loosing team

[Brother MEX]

I too dont play MW0 currently because of the matchmaker !

[Mudhutwarrior]

The deal is you have players whose skill level far exceeds the average dropping into solo only to club seals. They know who they are.

Its not MM as much as it is a solid core of people (use that term lightly) whose sole objective is to pad their scores to feel superior. They always end up on the same team most of the time too but were not to question that.

With over 20k matches its pretty obvious when it happens and you see not a mistake amongst them and perfect flanking and pushes. Usually their pug lance dies before anyone else then they mop up. Again its not MM but many gaming the system to their full advantage.

It was proposed some time ago that high elo players in solo should always go to the back of the line and be very randomized in placement. PGI could do that easily it was said. I would gladly wait longer if it would make it better for new players and casuals. It would also encourage better players to move up into group and cw.

[Creovex]

I think people assume too much...

If PGI posted damage state for every mech (35% armor left) on the closing match screen we would see less rants then this. The problem is the lack of focus fire in pubs. Trust me when I say I come out of stomps on the winning side and count at least 4 -5 guys who are under 40% but the enemy team never communicated to focus fire or didn't care. This is the biggest issue and has gotten way better since VOIP went in but still happens when idiots chose to turn off VOIP and not listen. ( Yes I mean idiots, I watched a guy swearing in text chat because people didn't read his text in combat however he never heard anyone who was trying yo tell him otherwise because he has VOIP turned off as not to listen to swearing.. wow right?)

Additionally, challenges that require wins are made for newer players to get free mechs... however the last one only required a kill and assist per a match so take an assault, bum rush a medium and you only had to nick another mech for the assist. Or take an LRM boat... It wasn't a hard challenge just time intensive and realistically, if you can't kill one enemy per a match on average, your ELO is not near high enough to have the issue you are describing.

Edited by Creovex, 07 March 2015 - 06:59 AM.

[Brother MEX]

Creovex, on 07 March 2015 - 06:59 AM, said:

... Or take an LRM boat... It wasn't a hard challenge just time intensive and realistically, if you can't kill one enemy per a match on average, your ELO is not near high enough to have the issue you are describing.

You overestimate the power of LRM's

Many competitive players rank LRM as low as FLAMERS !

And my problem is that my ELO is much too high, as I often play in groups with much better players, which means that I get paired with better players in the SOLO queue because they may have a lower ELO than me AND are more skilled as I so that I have a hard time to get the necessary kills or points for the challenges

Edited by Brother MEX, 18 March 2015 - 09:44 AM.

[km1710]

Sjorpha, on 05 March 2015 - 02:28 AM, said:

The second myth is that streaks of wins or losses are a sign of bad or "rigged" matchmaking. The opposite is true, any random system of sufficient size will feature predictable quantities of extreme results. If no one ever got extreme streaks of wins or losses then that would be evidence of a rigged matchmaker. So either we want a rigged matchmaker or we accept the occurrence of extreme win/loss streaks. You can't have the cake and eat it in this regard.

why? a balanced MM would even out wins and losses not by streaks but by more distributed outcomes. If one keeps losing its Elo should lower till the point there is competition and not one sided stomps. Extreme results are predictable but the big size means it will be distributed following statistical laws. This MM simply cannot work due to variable teams and averaging out the small player base.

[Escef]

Matchmaker is equally unfair to everyone, which is the most fairness one can realistically expect.

I mean, look, when I find myself in drops with people that derp around in severely underweight mechs (a stock engine Catapult with 3xML, an LRM20, and no jets? where's all the tonnage?), or straight up bizarre builds (a Kit Fox with a UAC2, 2xERML, and an LRM5?), or, better yet, my personal (mis)favorite, the guys that do nothing most of the match and try to KDR-mine at the end... When I see all that over the course of only a couple days, yeah, cry me a River City, we all get screwed, roll with it.