81 replies to this topic

[627]

kill stealing and hanging back only works if you (or better, matchmaking calculations) find someone to carry the match for you.

But if MM decides you are the chosen one... well you're foxed because everyone will die and you'll face something like 10 enemies. Good luck with that.

In the end, those stats are meaningless. Pug life is what it is, a random game with random outcome. You can do something but only so much there. And if the odds are against you there's nothing you can do.

And group queue? That's a totally different game imho. With a small group, you have the same randomness but with a bigger group, say lance size or more you can not only carry the match but direct it to your favor.

Or you get even more crushed by your enemy because all of your buddies are drinking and laughing instead of playing serious. Oh, btw it is more fun in the drunken team, trust me.

[Adiuvo]

As you can see in this thread, the MWO forums aren't exactly the most competitive atmosphere out there. People try often to deny that stats actually mean something when there's no evidence to point that direction. There are many players out there with W/Ls far above 1 in the solo queue, and KDRs far above 2. It's not some mythical thing that only happens to the lucky.

Anyway, your stats are alright. In MWO the carry potential is much higher for an individual player compared to something like a moba, so while a 54% WL isn't the best thing out there it's not horrid either.

[N0MAD]

Adiuvo, on 08 April 2015 - 10:45 PM, said:

As you can see in this thread, the MWO forums aren't exactly the most competitive atmosphere out there. People try often to deny that stats actually mean something when there's no evidence to point that direction. There are many players out there with W/Ls far above 1 in the solo queue, and KDRs far above 2. It's not some mythical thing that only happens to the lucky.

Anyway, your stats are alright. In MWO the carry potential is much higher for an individual player compared to something like a moba, so while a 54% WL isn't the best thing out there it's not horrid either.

Bah W/L is one thing sure but in solo it means nothing, in group Q with the majority of teams being basically pubs any semi organised team can rack up good W/L ratio.
K/D is really again very exploitable, i know this because my son was leveling up an account, so one night i decided to hop on and do some games for him... after 10 or so games on his new account i had his K/D ratio at 5/1 in a Trial Stormcrow.
Im no leet player yet playing for points(being careful and calculated and ya stealing kills) i was able to get a 5/1 K/D? lol..

[Widowmaker1981]

Stefka Kerensky, on 08 April 2015 - 10:09 PM, said:

In 8vs8 your contribute will be 1/8. Still 7/8 is upon teamates.

BTW, in my previous post, neither dmg is a good indicator, because 600 dmg in a dire and 600 dmg in a commando... are 2 completely different skill levels.
That's why dmg/tons/match.

Disagree slightly because dmg/ton/match totally favours light specialists - there is no way in the world that an assault specialist in a Dire would have 3 times the dmg/match an equally skilled light specialist in a Firestarter would have. A little bit more maybe, maaaaybe as much as 1.5x - which is HALF the dmg/ton/match.

Just look at your own stats to see if you are improving.

Edited by Widowmaker1981, 09 April 2015 - 12:34 AM.

[El Bandito]

N0MAD, on 09 April 2015 - 12:02 AM, said:

Bah W/L is one thing sure but in solo it means nothing, in group Q with the majority of teams being basically pubs any semi organised team can rack up good W/L ratio.

Solo-Q WLR is currently the reliable indication of a person's skill. In group Q, win/loss always boils down to organized team or pug group. If one can consistently do well in the madhouse that is named Solo-Q, then one is undoubtedly good at MWO.

Edited by El Bandito, 09 April 2015 - 12:53 AM.

[dario03]

Adiuvo, on 08 April 2015 - 10:45 PM, said:

As you can see in this thread, the MWO forums aren't exactly the most competitive atmosphere out there. People try often to deny that stats actually mean something when there's no evidence to point that direction. There are many players out there with W/Ls far above 1 in the solo queue, and KDRs far above 2. It's not some mythical thing that only happens to the lucky.

Anyway, your stats are alright. In MWO the carry potential is much higher for an individual player compared to something like a moba, so while a 54% WL isn't the best thing out there it's not horrid either.

I think a lot of it comes from there being so many variables and difficulty in comparing stats. Is somebody that only plays solo and has a 5.0kdr with 75% win rate good? Yes of course.

But is somebody who only plays in large groups with only meta mechs and has a 4.0kdr and 60%w/l better than somebody who plays solo and groups with a mix of mechs with a 3.5kdr and 55%w/l? Not sure there.
And then there is the style of play and the game modes. One of the best ways to keep a high kdr is to simply not die. Even if you don't get many kills per game if you simply don't die much your kdr will be high. So somebody that plays skirmish mostly and goes out guns blazing at the end could have a lower kdr than somebody who mostly plays conquest/assault and hides when things go bad. Even if the guns blazing guy is a better player, the fact that the enemy must kill him and he doesn't drag out the game means his kdr will get hit more than the guy that can just hide and let the enemy easily finish the game by capping.

Now of course how much that really affects stats is up for debate and there are players that have high kdr without doing that but there are people that try to protect their stats like that and it causes some to have a negative view on kdr. I don't think as many people have issues with the W/L for comparison though. If you are winnng 60%+ in solo then you are probably doing something right. But I suppose people could still claim mech choices there.

Edited by dario03, 09 April 2015 - 01:09 AM.

[STEF_]

Widowmaker1981, on 09 April 2015 - 12:34 AM, said:

Disagree slightly because dmg/ton/match totally favours light specialists - there is no way in the world that an assault specialist in a Dire would have 3 times the dmg/match an equally skilled light specialist in a Firestarter would have. A little bit more maybe, maaaaybe as much as 1.5x - which is HALF the dmg/ton/match.

Just look at your own stats to see if you are improving.

Yep, it is esactly as you wrote.
But still it's a good way to read efficiency in the stats, at least comparing mechs in the same weight classes.

(yes, Lights are over quirked )

[Lily from animove]

Mcchuggernaut, on 08 April 2015 - 04:52 PM, said:

Really? A few hours and still no reply from anyone? I figured such an active forum community would have SOMETHING to say at least. Not even an obligatory #learntoplay troll post? Hell has officially frozen over...

k/d is pointless to measure in comparison, simply because it is not a skill indicator, let me run a fun build for like 50 matches and k/d is ruined.

k/d is also heavily influenced by the queue you play in, solo most time? or in small groups or in big groups? this affects k/d a lot more than your personal skill level. So when you compare your k/d with others, that play more big groups, probably 10-12 man all day, this is not really comparable at all. Do you play most missions with other good players, or maybe with bad players, becuase if you group with bad players, this often makes your k/d go worse. But yet it does not really judge your very own skills, because your k/d is not only provided by your performance.

If you want comparison, take screens of your matches, and look at the end where your performance in comparison to others is.

[JohnnyWayne]

When I group with bad players my k/d and w/l goes up, its way easier to carry.

Average damage and average damage taken (especially when you died) would also be nice stats to track.

Edited by JohnnyWayne, 09 April 2015 - 03:36 AM.

[Bawbagzz]

As long as I'm getting 4 kill games in my Hunchback G i'm happy.W/L in pug land is really out of your own hands so that isn't much of an indicator of skill.
I think having your stats hidden is good coz I've seen players getting pelters for being a tomato in WOT and the matchmaking in that game is as bad if not worse than MWO.

Edited by Bawbagzz, 09 April 2015 - 03:43 AM.

[Joseph Mallan]

I will assume you stand on teh ground. Unless you are indoors, then you will likely stand on teh floor.

Edited by Joseph Mallan, 09 April 2015 - 03:49 AM.

[STEF_]

El Bandito, on 08 April 2015 - 05:45 PM, said:

Just think of it like this. Your skill level is above 50% of the total players if you have more than 1:1 WLR. Probably not entirely accurate but it helps me sleep better at night, so what the heck.

Sorry but:



1) k/d ratio is not an indicator: I cored/killed only one by myself, here, and then I simply stool 2.
2) w/l ratio? Nope. Impossible. With 7 teamates doing less than 200 dmg... you can carry like a train, but still you will lose.

That's why, when this kind of threads appear, I end saying.. "PGI, please, throw away ELO MM...please...."

Edited by Stefka Kerensky, 09 April 2015 - 04:05 AM.

[Zordicron]

How to measure your performance as a pug:

Look at K/d, make sure it is in the green.
Look at W/L, same
Look at weapon accuracy, make sure nothing is abhorrent, keeping in mind the nuances of each.

Then, most importantly, look at individual mech stats. If you have some that you know you like to play, and feel you do well in, it should reflect here. If you know one of them is a stinker, it should also reflect there. Note, of course, different strokes for different folks. If you are really quite a good pilot, you mech stats shouldn't really deviate that much, with a few running higher then others, sometimes by a good bit. this part becomes easier to evaluate the more mechs you have run to master giving a bigger picture.

The mech thing is important as a pug player because it shows your ability to adapt and be aware of the field to apply individual mech's strong points accordingly. Overall stats show this very poorly IMO because they can be skewed so easily by those mechs that show as high runners on your list. Map stats can be useful also to see if you need to work on a given map or playstyle/tactical area of your game, usually this is related to movement/positioning and awareness also.


If you play in a unit, most of these things above do not apply. The unit play is basically 75% different. All it takes really is someone calling targets and 11 other guys focusing fire on it to see why. Pugs just don't call targets like that. As such, field awareness in a unit can be deferred to one or two guys on the team allowing the rest to just focus fire and position. In Puglandia, you need to watch your own team, and the enemy team, and react based on the actions of both- in a unit, you know your own team will react as the leader is calling for. That is a huge difference. A lot of unit guys don't adapt well to this at all in pug land despite all the chatter about l33tness etc. Pugs can get ripped also in unit play if they don't know how it works. IMO though, it is easier for a pug to move into a unit then the other way around. this is also why I think pug stats show your base skillset better overall, basically it's just you they are tracking and not your leaders(unless you are the leader but even then you are relying on the 11 other guys to focus harder then the other team)

Maybe PGI could release some leaderboard, but I think a lot of players would not be represented well there when you have a chunk of playrs that run EZmodeCatz all day and others that own 180 mechs and have stats for Spider 5v's etc weighed also. better to just evaluate yourself and not worry to much about it IMO.

[AThousandD]

Stefka Kerensky, on 08 April 2015 - 09:46 PM, said:

THE REAL parameter of effectiveness/efficiency is this, IMO: dmg/ton/match.

You piqued my interest, so I went ahead and crudely calculated the dmg/ton/match and set it against W/L records for the mechs. An observable pattern seems to be that light mechs will have better dmg/ton/match results than heavier mechs. Furthermore, dmg/ton/match doesn't seem to be closely correlated with either winning or losing a match (case in point: CN9-AH, which I love and usually have killer games in them - so what if statistically I lose more than I win, always fun to SMASH THINGS).

[Note: some of the data is not viable, due to the low number of games played]

Mech Damage Tons Games Damage/Tons/Games Avg dmg/match W/L
TBR-C 9094 75 22 5,511515152 413,3636364 2,67
DRG-1N 12178 60 32 6,342708333 380,5625 2,2
SDR-5K 9807 30 35 9,34 280,2 2,09
BNC-3M 33917 95 77 4,636637047 440,4805195 2
BNC-3E 21192 95 41 5,440821566 516,8780488 1,92
RVN-3L 19703 35 52 10,82582418 378,9038462 1,83
JM6-FB 54828 65 149 5,661125452 367,9731544 1,74
AS7-D 47108 100 105 4,48647619 448,647619 1,74
TBR-S 11529 75 31 4,958709677 371,9032258 1,73
VTR-9S 67168 80 161 5,214906832 417,1925466 1,62
CN9-AL. 67533 50 204 6,620882353 331,0441176 1,56
AWS-8Q 56139 80 140 5,012410714 400,9928571 1,55
AS7-D-DC 34436 100 76 4,531052632 453,1052632 1,52
JM6-DD 23427 65 58 6,214058355 403,9137931 1,5
CTF-3D 15835 70 41 5,517421603 386,2195122 1,5
ZEU-6T 4048 80 10 5,06 404,8 1,5
GRF-1N 46797 55 122 6,974217586 383,5819672 1,45
PNT-8Z 12249 35 42 8,332653061 291,6428571 1,41
DRG-1C 36562 60 100 6,093666667 365,62 1,39
WVR-6K 31503 55 71 8,067349552 443,7042254 1,38
HBK-4G 47277 50 139 6,802446043 340,1223022 1,35
CTF-1X 36132 70 103 5,011373093 350,7961165 1,35
SDR-5V 28793 30 90 10,66407407 319,9222222 1,33
WVR-6R 21733 55 59 6,697380586 368,3559322 1,32
TDR-9SE 76658 65 216 5,45997151 354,8981481 1,27
HBK-4P 73355 50 194 7,562371134 378,1185567 1,26
AWS-8T 42487 80 101 5,258292079 420,6633663 1,26
CDA-3M 9392 40 27 8,696296296 347,8518519 1,25
ENF-4R 19369 50 51 7,595686275 379,7843137 1,23
SDR-5D 26536 30 82 10,78699187 323,6097561 1,19
CDA-2B 27222 40 85 8,006470588 320,2588235 1,18
GRF-1S 51135 55 133 6,990430622 384,4736842 1,17
CN-9A 37380 50 103 7,258252427 362,9126214 1,13
HBK-4SP 49260 50 151 6,524503311 326,2251656 1,13
TDR-9S 59873 65 169 5,450432408 354,2781065 1,12
CDA-3C 12481 40 40 7,800625 312,025 1,11
ENF-5P 12967 50 41 6,325365854 316,2682927 1,11
RVN-2X 51089 35 147 9,929834791 347,5442177 1,1
RVN-4X 51089 35 196 7,447376093 260,6581633 1,1
PNT-10K 17544 35 69 7,264596273 254,2608696 1,09
TBR-PRIME 9384 75 23 5,44 408 1,09
GRF-3M 40573 55 112 6,586525974 362,2589286 1,08
WVR-7K 50101 55 118 7,71972265 424,5847458 1,07
BNC-3S 40384 95 107 3,972848008 377,4205607 1,06
PNT-9R 12097 35 40 8,640714286 302,425 1,05
JM6-A 13654 65 46 4,566555184 296,826087 1,05
CTF-4X 62733 70 183 4,897189696 342,8032787 1,01
AS7-RS 35727 100 96 3,7215625 372,15625 0,98
CN-9AH 37585 50 94 7,996808511 399,8404255 0,94
TDR-5S 66544 65 174 5,883642794 382,4367816 0,94
CN9-D 33769 50 99 6,822020202 341,1010101 0,92
AWS-8R 51291 80 109 5,881995413 470,559633 0,91
ENF-5D 26190 50 84 6,235714286 311,7857143 0,89
VTR-9B 14618 80 39 4,68525641 374,8205128 0,89
DRG-5N 27549 60 78 5,886538462 353,1923077 0,79
VTR-9K 21014 80 61 4,306147541 344,4918033 0,69
ZEU-6S 7136 80 22 4,054545455 324,3636364 0,54

Pastebin, more readable: http://pastebin.com/ni0Tw01w

Edited by AThousandD, 09 April 2015 - 05:18 AM.

[MeiSooHaityu]

Well, maybe it is myself just trying not to feel too bad, but I'd imagine anything solidly over 1.0 KDR is at least good.

It doesn't sound impressive, but anything solidly over 1 means (on average) you kill more people than you die. I figure that means chances are you will be a positive asset. If you are under 1, you die more than kill and chances are on average you will be a liability.

Plus even average KDR can be a bit misleading. If all I ever piloted was my Panther 10K, I would currently have a KDR of 2.48. If I strictly piloted my grasshoppers, I would also be over 2. But, I also pilot Locusts, Vindicators, Catapults, and now Urbies. Mechs like those post KDRs anywhere from 1.7 (Cat K2) down to 0.13 (Locust PB).

Basically, if I played only mechs I was REALLY good at, my KDR would be higher (currently 1.44). Even my average is misleading because if I happened to be working on my PB that match, I probably will be a liability.

Overall though, fun trumps KDR. If I was worried, I'd play the mechs that gave me the best KDR for bragging rights. Instead I play what I have fun with. I'm happier that way.

[STEF_]

AThousandD, on 09 April 2015 - 05:16 AM, said:

You piqued my interest, so I went ahead and crudely calculated the dmg/ton/match and set it against W/L records for the mechs. An observable pattern seems to be that light mechs will have better dmg/ton/match results than heavier mechs.

Because lights are quirked (and some over-quirked) to be viable in what we are playing: skirmish.

AThousandD, on 09 April 2015 - 05:16 AM, said:

Furthermore, dmg/ton/match doesn't seem to be closely correlated with either winning or losing a match.

Because it's always the team that win or lose, and a mechwarrior is only one with other 11.
This regarding Pugville, of course. Groups is another business.

dmg/ton/match is useful for reading pilot efficency while using different chassis, and, for testing efficiency of the same mech but with different loadout, too.

[Mawai]

Mcchuggernaut, on 08 April 2015 - 09:12 PM, said:

I don't think your right about how the matchmaker works. Everything I have heard and read before is that the matchmaker works by looking for an average skill level that is as close as possible for the two teams, and doesn't do it how you say. To my knowledge, it works by finding an AVERAGE skill level by adding up the skill levels of two teams. This is why you get so many teammates that don't do well after you have a particularly long streak of good games, and get a lot of competent teammates when you have been consistently doing poorly for a while. That explains a lot of the "streaks" of wins and losses a lot of players see, as well as those sets of games where you easily pull off 600+ damage a match but your team keeps melting and self-destructing on you.

Ok ... I am tired of folks who haven't actually read the dev posts on how the matchmaker works. You make up a story to explain your experiences instead. I have included links to dev posts below that cover a variety of aspects of Elo and the matchmaker. There are a lot of posts I have not cited.

Karl Berg was a great source of information about MWO. Unfortunately for us he has moved on to other opportunities, apparently at Amazon.

Link to main thread
http://mwomercs.com/...courtesy-phone/

Link to dev tracker so you can see his posts:
http://mwomercs.com/...serMode=content

Summary of how matchmaking works ...

1) Matchmaker selects a seed player or group - oldest in queue - this defines the initial Elo of the match.
2) Matchmaker begins adding players to both sides finding players within a narrow Elo band of the seed value (A number of 50 was cited in one of the posts but this is a variable parameter)
3) If by chance one side or the other starts to have a somewhat higher average Elo the matchmaker can bias to put the next higher Elo player on the side with the lower total Elo ... but still within the limited search parameter band.
4) If the match is taking too long to come up with two teams that are weight class and Elo matched then the matchmaker can relax the weight class matching and expand the acceptable Elo band.
5) Statistically both the standard deviation and difference between Elo for two teams in the solo queue is quite small though there are always outlier matches with larger deviations.

All of these statements are supported by comments, mostly from Karl, who was the person responsible for writing the matchmaker. I haven't cited all of his posts ... if you want to read them you can go through the thread linked above.

But ... please ... stop the BS about the matchmaker since you don't seem to understand how it works.

--------

Matchmaker assembles matches iteratively ... going from side to side ... picking players within constraints of Elo and weight class ... harder for group queue
http://mwomercs.com/...ost__p__3695838


Teams formed iteratively first one then the other ... both group and solo
http://mwomercs.com/...ost__p__3695793

Elo is self correcting over time ... you eventually get a correct Elo even though you are only 1 of 12 players on a team.
http://mwomercs.com/...ost__p__3571420

Elo converges to a value not dependent directly on your win/loss ratio - they are not the same.
http://mwomercs.com/...ost__p__3562350

Matchmaker does NOT specifically match a high and low Elo player vs two medium players ... if teams get really out of sync in Elo by chance it can expand range to bring them back to even.
http://mwomercs.com/...ost__p__3534326


---------------- Lots and lots of other useful posts ---------------


Elo updating:
http://mwomercs.com/...ost__p__3934714


Initial and growing Elo range for match making:
http://mwomercs.com/...ost__p__3769240

On fixed brackets
http://mwomercs.com/...ost__p__3710122

On matchmaker rewrtie and Elo adjustment/match
http://mwomercs.com/...ost__p__3710094


If matchmaker has let release valves out on Elo ... then if one team has a higher standard deviation then usually so does the other team
http://mwomercs.com/...ost__p__3698172

Seed Elo used to start matchmaking ... initial band about 50 (see other post)
http://mwomercs.com/...ost__p__3698163

Before matchmaker improvements ... outlining scenarios that can cause worse standard deviation in matches
http://mwomercs.com/...ost__p__3696790

Group matchmaking and hard vs soft constraints
http://mwomercs.com/...ost__p__3696769

[Crotch RockIt]

To OP, take a look at this website:

www.mechstats.com

Players can log in and upload their stats for public viewing and comparison. It can help show how you stack up against other players, but unfortunately, the sample size is a bit low.

[MischiefSC]

How long does it take to get to work in the morning?

You can give good estimate, can't you? What if you averaged your commute time over a month, assuming it's 40 days? How accurate would it be?

You don't control any of the hundreds of other drivers you interact with but the fast and slow ones average out over 40 days to give a good representation of how fast you drive to work.

Pug queue is the same thing. The quality of your team averages out vs the other team and the result is all you.

That is the real problem. People all want to think they are already amazing and the truth is they are average. The actual top tier players are fundamentally those who can recognize that and focus on changing what they do to get better. Everyone else wants to just do what they have always done and change the metrics to say they are better, or make excuses for why the metrics are wrong.

[El Bandito]

Stefka Kerensky, on 09 April 2015 - 04:03 AM, said:

Sorry but:



1) k/d ratio is not an indicator: I cored/killed only one by myself, here, and then I simply stool 2.
2) w/l ratio? Nope. Impossible. With 7 teamates doing less than 200 dmg... you can carry like a train, but still you will lose.

That's why, when this kind of threads appear, I end saying.. "PGI, please, throw away ELO MM...please...."

Presenting such argument with a screenshot of a single game is not gonna convince me. You need to realize that such games are not all that common and your pug WLR will average out in the long run due to MM doing its best to fit you with bad teammates as you gain more victories. Only those who are at least moderately skilled will continue to beat the odds in pugs and have more than 1:1 WLR. The rest are simply tumbling along MM's tender mercy.

Pug WLR is a good indicator of how well you stack up against other puggers (85% of the player base)--after few hundred matches or so.

Edited by El Bandito, 09 April 2015 - 06:06 AM.