167 replies to this topic

[xMEPHISTOx]

OP does have a point to some extent. W/L % is not the all inclusive factor in whether or not a pilot is good or not, and can easily be padded if play w/other good pilots.
A simple 1 dimensional method for calculated a pilots value is to simple if in fact that is how it is being calculated. If not going to create a proper algorithm then just make it (MM)completely random, it amounts to the same thing.

Edited by xMEPHISTOx, 08 November 2013 - 10:25 PM.

[Ghogiel]

Random you say?

FUN. Stomping back to 90% W/L ratio!

[Asakara]

Wispsy, on 08 November 2013 - 03:23 PM, said:

Well some of us manage to win an awful lot more then we lose...I guess that it is not really our contribution but just random matchmaking luck?

Once you start believing you cannot affect the outcome you have already lost. There is always a way to win, just have to dodge carry harder.

I agree though with a slight modification.

Adiuvo, on 08 November 2013 - 03:30 PM, said:

Assuming you're not using a useless mech, you can always carry your matches. Like Wispsy, I win a lot more than I lose even when solo dropping. It requires more effort than if I was grouped, but it's still possible and I mainly play lights. Play an assault mech and the whole thing is easier.

QFT.

[Wispsy]

Asakara, on 08 November 2013 - 11:08 PM, said:

I agree though with a slight modification.



QFT.

Dodge harder is what I tell myself every time I die in every game I play. As long as you do not get hit the rest will eventually fall into place!!

[Drasari]

MischiefSC, on 08 November 2013 - 03:59 PM, said:

I have the converse experience. I got the Overlord package and have tried to make the Phoenix Mechs perform anywhere close to where some of the better (not even top, just better) mechs in the weight class do. Result?

For my 'career' of around 5,000 drops my time spent with Phoenix mechs (BM, Tbolt, most Shadowhawks, Locusts) have dropped by win/loss from around a 1.35 or so down to a 1.21. Stretches of 10 or 16 straight losses.

Statistically impossible in a 'random' system at least in this number of drops I'm creeping past 50 more losses than wins in 100 drops, because the chassis are mostly absolute trash. My performance matters and when my performance suffers my win/loss rate suffers.

It goes both ways. Elo and a win/loss system is the best, most accurate system reasonably available for this sort of game.

What are the odds of getting heads when you flip a coin? 50%. the second time? 50%. 3rd time? 50%.

[JigglyMoobs]

Blurry, on 08 November 2013 - 05:11 PM, said:

hate to tell you but yes it is just matchmaking. whichever side fate puts you on. now as I grow as a player I have gotten significantly better as a player over the last month.

my seeing things is yes it really is up to the matchmaker 99% of the time. And now I do win a lot more and survive a lot longer and my conclusion stands.

If its up to the match maker 99% of the time then how is it possible for me to maintain win loss ratio above 1.5 in some of my mechs even after thousands of matches? Lots of others here have w/l ratios above 2 for some of their mechs. Surely you're not suggesting that this is the matchmakers doing?

[John MatriX82]

I've started a new account during this week; I don't care if it is against the ToS, but the idea was to have one secondary to make some school for the newcomers that otherwise -when they play with me- are pitted against tournament winners and aren't going either to enjoy the game nor to improve themselves..

• In the first 25 matches, I expected to see a plethora of trial (champion in the last week) mechs.

None of that.

Roughly up to 6 players (counting both teams) had trial mechs, many times I was the only one; the rest was already playing with property chassis and DHSs, XL engines and so on.

• I expected to find some dumbsters or strange builds. Besides a few issues with people shooting in your back or making dumb moves/charges, I've ran into few players standing still or shooting things like LRMs from above or under their min/max range; let's say 4-5 players in the whole first 25 matches and rarely two of those present in either one of the sides.

• I ran into 4 ERLL stalkers, UAC5 jagers, AC20+2PPC Victors or Highlanders, 733Cs, several hero/phoenix mechs, poptarts.. the usual competitive builds I run into daily; I expected not to run into players that I know (read: that I encounter at my elo level -which yes, it's very high-) but surprisingly I've ran into several "known" players that of course used to pwn everyone in the rounds I saw them into.

• I expected not to run in premades, right from the second match I was pitted against a 4man, at the 5th I ran into a 4+4 on the other side; it became a common sight in nearly every of the first cadet bonus matches I ran.

• Matches were most of the times totally one sided, with weight balance off in several matches. 6 out of 25 matches were "close". The rest were total streamrolls by either my team or the other team, often blessed by like 8 assaults vs 2 or 3.

This matchmaking system keeps failing even at the "average" ELO level of a beginner. It doesn't work. And that is alarming, my suspect it's that the player pool to chose from is rather limited = few people are actually playing the game.


To sum it up:
-weight imbalances (often huge)
-player level imbalances
-premades vs total pugs
-property mechs pitted against trial ones (and thankfully there were champions in the week I started off a new account)
-close matches (12-8 minimum): 6 out of 25 (the rest were one sided streamrolls).

Last comment: meh..

Edited by John MatriX82, 09 November 2013 - 02:40 AM.

[KinLuu]

The only constant factor in your team, is you.

[GalaxyBluestar]

it's a tough call when you think about it, i'm not realying on 12 guys being bad and 11 being good but i'm also hoping that the 12 are not good and that my eleven don't compeletly screw up.

in the end though it is up to you whether you count or not, decision, just one can mean all the difference in the outcome and it's just as much your one choice as apposed to what everyone else is doing.

so MM the way it is it's always 50/50 whether your team lets you down or their team is just OP, but in the end it's always your responcibility if you are apart of the letdown or apart of the struggle to salvage something, if you can master the latter a lot you will more likely win than lose each match. that's what pros do and the only reason i'm saying this because i've sen the best and the worst, i've had varying elo levels so i've seen how good players run like hell and duck and dive and go all over the place, and i've seen someone sit their taking damage and played worse than snorlax. so i guess W/L ratio with enough plays under the belt does show the best from the worst, meta's pending.

i'm not one of them pros BTW. casual is as casual does. only wins on a casual basis.

Edited by GalaxyBluestar, 09 November 2013 - 03:17 AM.

[Joseph Mallan]

Wispsy, on 08 November 2013 - 03:23 PM, said:

Well some of us manage to win an awful lot more then we lose...I guess that it is not really our contribution but just random matchmaking luck?

Once you start believing you cannot affect the outcome you have already lost. There is always a way to win, just have to dodge harder.

Adiuvo, on 08 November 2013 - 03:30 PM, said:

Assuming you're not using a useless mech, you can always carry your matches. Like Wispsy, I win a lot more than I lose even when solo dropping. It requires more effort than if I was grouped, but it's still possible and I mainly play lights. Play an assault mech and the whole thing is easier.

So if you two are carrying a team to victory you are artificially inflating your less capable team mates to a higher Elo they do not belong in.

Teams win a Team game, individuals contribute to that win. So unless you are killing all 12 by yourself, Its the team's win, not yours.

[Amsro]

While I don't particularly care either way, MM and Elo are and have been completely broken since GO. Anyone who talks themselves into believing its a fair and balanced system and keep dreaming.

There are too many factors (2-4 man, solo pug, solo newb, newb 4 man). PGI 's matchmaker will never account for all variables.

I'll continue to attempt to guide PUG's in game, that seems to be the largest factor in winning... you know the ability to keep 12 mechs organized~ish.

[Diablobo]

When one can get 4 kills and still lose the match, ELO is not appropriate.

[Wispsy]

John MatriX82, on 09 November 2013 - 02:38 AM, said:

I've started a new account during this week; I don't care if it is against the ToS, but the idea was to have one secondary to make some school for the newcomers that otherwise -when they play with me- are pitted against tournament winners and aren't going either to enjoy the game nor to improve themselves..

• In the first 25 matches, I expected to see a plethora of trial (champion in the last week) mechs.

None of that.

Roughly up to 6 players (counting both teams) had trial mechs, many times I was the only one; the rest was already playing with property chassis and DHSs, XL engines and so on.

• I expected to find some dumbsters or strange builds. Besides a few issues with people shooting in your back or making dumb moves/charges, I've ran into few players standing still or shooting things like LRMs from above or under their min/max range; let's say 4-5 players in the whole first 25 matches and rarely two of those present in either one of the sides.

• I ran into 4 ERLL stalkers, UAC5 jagers, AC20+2PPC Victors or Highlanders, 733Cs, several hero/phoenix mechs, poptarts.. the usual competitive builds I run into daily; I expected not to run into players that I know (read: that I encounter at my elo level -which yes, it's very high-) but surprisingly I've ran into several "known" players that of course used to pwn everyone in the rounds I saw them into.

• I expected not to run in premades, right from the second match I was pitted against a 4man, at the 5th I ran into a 4+4 on the other side; it became a common sight in nearly every of the first cadet bonus matches I ran.

• Matches were most of the times totally one sided, with weight balance off in several matches. 6 out of 25 matches were "close". The rest were total streamrolls by either my team or the other team, often blessed by like 8 assaults vs 2 or 3.

This matchmaking system keeps failing even at the "average" ELO level of a beginner. It doesn't work. And that is alarming, my suspect it's that the player pool to chose from is rather limited = few people are actually playing the game.


To sum it up:
-weight imbalances (often huge)
-player level imbalances
-premades vs total pugs
-property mechs pitted against trial ones (and thankfully there were champions in the week I started off a new account)
-close matches (12-8 minimum): 6 out of 25 (the rest were one sided streamrolls).

Last comment: meh..

Silly question but you were not grouped with anybody except perhaps other new accounts right? I did not experience this on my new accounts unless I was grouped with people much higher. Not everybody was some drooling ***** but I saw quite a few trial mechs and lots of very inefficient custom mechs, not necessarily running into walls (very often) but not necessarily making the right or the smart decisions in regards to what was going on in the battle, with amazingly poor aim on almost everybody.

[Joseph Mallan]

Diablobo, on 09 November 2013 - 07:40 AM, said:

When one can get 4 kills and still lose the match, ELO is not appropriate.

Elo is not based on your personal ability in your example.

[MischiefSC]

Wispsy, on 09 November 2013 - 08:12 AM, said:

Silly question but you were not grouped with anybody except perhaps other new accounts right? I did not experience this on my new accounts unless I was grouped with people much higher. Not everybody was some drooling ***** but I saw quite a few trial mechs and lots of very inefficient custom mechs, not necessarily running into walls (very often) but not necessarily making the right or the smart decisions in regards to what was going on in the battle, with amazingly poor aim on almost everybody.

I find that strange as his experience doesn't match with mine at all. If his experience is true then why do I only see new players once in a dozen or two dozen games?

I get the desire for people to want to feel that the matchmaker is responsible, that if they're not winning it's not because they made a mistake or over-extended themselves or left cover or chased a kill or went right on Canyon or some other nub mistake. I get that.

It doesn't make it true.

Elo is pretty accurate and it's based off of how well the player influences matches to drive a win for their team. That's it. Not your ability to swoop in and grab the kills that your allies weakened before dying, not your ability to stay alive by avoiding combat until the very end, etc. In fact a lot of the behaviors that benefit KDR (playing very cautiously, nabbing the easy kills) actually drive win/loss DOWN - if you're not in 12mans or seriously competitive play then playing cautiously is probably shafting your teammates, especially when the other team plays aggressive. You're not drawing attention from anyone and this allows the other team to focus their fire on fewer potential targets, increasing the damage your teammates take as individuals and causing that cascade of failure where 2 or 3 people on a team get focused down and then they get rolled.

At this point though it's been debated here a lot. All the details are out there. The MM is designed to keep the game challenging and that's uncomfortable for some people. They don't want a challenge, they want to feel like they're a winner.

The problem of course is that in any system where that's actually challenge oriented and not just handed out not everyone appreciates the effort winning entails.

[Ghogiel]

Drasari, on 09 November 2013 - 01:33 AM, said:

What are the odds of getting heads when you flip a coin? 50%. the second time? 50%. 3rd time? 50%.

Law of large numbers.

Also the whole notion that going AFK and you can win 90% of your games is hilarious. It's just luck after all.

[Bishop Steiner]

Drasari, on 09 November 2013 - 01:33 AM, said:

What are the odds of getting heads when you flip a coin? 50%. the second time? 50%. 3rd time? 50%.

depends on the coin. some (quarters i believe)are inherently unbalanced, making it like 50.1% you get tails.

Edited by Bishop Steiner, 09 November 2013 - 10:54 AM.

[mirrimon]

ELO does not look at w/l ratio. W/L ratio in not a very good indicator of your ELO score. From what PGI has told us the matchmaker makes 2 teams (somehow.) One of the teams is determined as the favorite to win (somehow.) If the favorite wins, ELO of all inviolved does not change much. If ithe favored team loses ELO changes based on how much the matchmaker thought the team should have won.

It is possible that many of a players wins and loses were the way the matchmaker predicted, thus their ELO is not very high, even with a high W/L ratio. While that is probably not very common, 2 people ith 1.5 w/l likely do not have the same ELO.

[Ghogiel]

mirrimon, on 09 November 2013 - 11:58 AM, said:

ELO does not look at w/l ratio. W/L ratio in not a very good indicator of your ELO score. From what PGI has told us the matchmaker makes 2 teams (somehow.)

It picks a target number and matches players with Elo rating closest to it. I think its been said it uses a player to seed the game, so it would use his Elo rating.

Quote

One of the teams is determined as the favorite to win (somehow.)

The team that has the higher Elo will be predicted to win a certain percentage of the time. Command chair shows the math how the probability is worked out.

Quote

While that is probably not very common, 2 people ith 1.5 w/l likely do not have the same ELO.

I think it is absolutely common. One player could have played thousands of more matches than the other. Outside of very slim odds the only way for that to happen is if those players only ever played the game together on the same side.

Edited by Ghogiel, 09 November 2013 - 12:26 PM.

[Diego Angelus]

ELO stat is laughable stat in mwo because it can only be true if it affects team not pugs. ELO is designed for chess so you get the picture how accurate it is in mwo when you are put in team of 12 people and base your rating around that. Best way to balance pugs is weight, how well does someone do in each mech (K/D,assists....) and number of matches played those are stats that MM should use.