284 replies to this topic

[The6thMessenger]

Wil McCullough, on 25 November 2016 - 02:47 AM, said:

now if you have nothing to add other than your "s/s or it didn't happen" rhetoric, once again i'll ask you to keep quiet unless you'd like me to keep demolishing your ridiculous arguments.

What arguments?

Edited by The6thMessenger, 25 November 2016 - 02:53 AM.

[Duke Nedo]

Zergling, on 25 November 2016 - 02:33 AM, said:

As far as I can tell, there isn't any attempt to balance teams by PSR; I suspect PSR is just a way to bracket players, not balance teams.

How can you tell? I seem to remember that the numerical PSR value was somehow used in MM, not just brackets for selection. Don't remember the source though, and it's a long time ago.

[Wil McCullough]

The6thMessenger, on 25 November 2016 - 02:53 AM, said:

What arguments?

the argument that the OP's in-game performance holds any weight when it comes to his point that players who consistently do <200 damage are bad.

[Zergling]

Duke Nedo, on 25 November 2016 - 03:09 AM, said:

How can you tell? I seem to remember that the numerical PSR value was somehow used in MM, not just brackets for selection. Don't remember the source though, and it's a long time ago.

No way to know for sure without seeing something from the Devs, but the battle outcomes just feel way too random for there to be any skill matching going on, and it doesn't feel like there's any weighting shifting players back to 1.00 W/L; there are too many good players maintaining above 1.00 W/L, and too many bad players below 1.00 W/L.


Hell, look at the correlation graphs here.



The 'average' point is 200 match score and 1.00 W/L. Yet players that average 400 are pulling off over 2.50 W/L; they are winning more than twice as many battles as average, for just twice the average match score.

If skilled players were being matched against one another by the matchmaker, I think there would be a much lower impact of skill on W/L, as such a skill matchmaker pushes above/below average players back towards the average.

Edited by Zergling, 25 November 2016 - 04:03 AM.

[Duke Nedo]

Zergling, on 25 November 2016 - 03:58 AM, said:

No way to know for sure without seeing something from the Devs, but the battle outcomes just feel way too random for there to be any skill matching going on, and it doesn't feel like there's any weighting shifting players back to 1.00 W/L; there are too many good players maintaining above 1.00 W/L, and too many bad players below 1.00 W/L.

Hell, look at the correlation graphs here.

The 'average' point is 200 match score and 1.00 W/L. Yet players that average 400 are pulling off over 2.50 W/L; they are winning more than twice as many battles as average, for just twice the average match score.

If skilled players were being matched against one another by the matchmaker, I think there would be a much lower impact of skill on W/L, as such a skill matchmaker pushes above/below average players back towards the average.

This data is mixed group queue and solo queue, taken from the leaderboards. It basically just shows that the more you win, the higher your average score is. It doesn't say that your individual score drove the win, rather the opposite I'd guess. It's easier to score high if you're dropping with a group and win a lot more than you lose.

[Zergling]

Duke Nedo, on 25 November 2016 - 04:19 AM, said:

This data is mixed group queue and solo queue, taken from the leaderboards. It basically just shows that the more you win, the higher your average score is. It doesn't say that your individual score drove the win, rather the opposite I'd guess. It's easier to score high if you're dropping with a group and win a lot more than you lose.

Most players play solo queue though, even the higher scoring players.


Also, see this battle screenshot, with (slightly rough) player skill estimations based on this month's leaderboard stats for each player:



My team had:
1x top 5% player
2x 'good' players with stats between top 25% and top 10%
2x 'above average' players with stats between the top 50% and top 25%
2x 'average' players with stats near the 50% mark
3x 'below average' players with stats between 50% and bottom 25%
2x 'potato' players with stats around bottom 25% or less


Enemy team had:
1x top 0.1% player
1x top 1% player
1x 'good' player
5x 'above average' players
2x 'below average' players
2x 'potato' players

If the matchmaker is trying to balance players based on PSR, it is doing a terrible job at actually balancing battles, as the enemy team definitely had a large skill advantage that battle, and honestly should have won.

Edited by Zergling, 25 November 2016 - 04:57 AM.

[meteorol]

Deathlike, on 24 November 2016 - 04:20 PM, said:

That is why group queue play is more deterministic.

If only playing MWO in small (especially with only 2 guys) groups wouldn't one of the most torturous things i have ever encountered playing online. Sadly, most of my buddies had to stop playing due to IRL schedule... We usually don't get more than 2 guys online at a time.

We usually decide against playing MWO at all instead of going to the groupqueue with just 2ppl. It's just so painful. Shoutout to all those 4 mans using like 1/3 of their available tonnage. Especially those guys dropping with 2 locusts, a spider and a hunchback (think it was a HBK, could be mistaken), ending on our team 5 times in a row, scoring ~32 dmg per person. Thanks guys, you lost your teams 5 matches in a row the moment you clicked on the search button.

The few times i played in groupqueue the last few months massively crushed my stats across the board and always made me take a MWO break.

[Chados]

Mech Croissant, on 24 November 2016 - 10:03 AM, said:

I love my Executioners, especially the one with 3 Lpl's and 9 serl. Lpl's for long and mid range and the small er-lasers for short range. It rocks a**!

Yeah? Isn't that like OMG hot? I might have to try that. I'm dying here-just can't find *that* build.

[The6thMessenger]

Chados, on 25 November 2016 - 05:12 AM, said:

Yeah? Isn't that like OMG hot? I might have to try that. I'm dying here-just can't find *that* build.

I found more success with 2x ERPPCs.

[Duke Nedo]

Zergling, on 25 November 2016 - 04:25 AM, said:

Most players play solo queue though, even the higher scoring players.


Also, see this battle screenshot, with (slightly rough) player skill estimations based on this month's leaderboard stats for each player:



My team had:
1x top 5% player
2x 'good' players with stats between top 25% and top 10%
2x 'above average' players with stats between the top 50% and top 25%
2x 'average' players with stats near the 50% mark
3x 'below average' players with stats between 50% and bottom 25%
2x 'potato' players with stats around bottom 25% or less


Enemy team had:
1x top 0.1% player
1x top 1% player
1x 'good' player
5x 'above average' players
2x 'below average' players
2x 'potato' players

If the matchmaker is trying to balance players based on PSR, it is doing a terrible job at actually balancing battles, as the enemy team definitely had a large skill advantage that battle, and honestly should have won.

Doesn't look that bad to me... remember that Tier 1 is skill-capped so Good and above will most likely all be Tier 1 with maxes XP-bar, thus counting the same. Above average to potato could be anything between Tier 1 and Tier 4.... so if you look at it like that it could very well be the same number for T1 players on both teams.

[PhoenixFire55]

Appogee, on 24 November 2016 - 04:16 PM, said:

Or do you instead believe that you're so good a player that you're single-handedly carry games and make your win:loss positive?

I have 7,701 / 4,280 W/L ratio in overall current stats, thats 1.8 W/L. From nearly 12.000 matches played there are maybe 200-300 played in a group, even if we assume all were wins and deduct them it'll still be 1.73 W/L. I highly doubt that you can call 12k matches a "lucky streak" of winning, i.e. the only common factor in all of them is me. I also happen to have about 200 fully mastered mechs, pretty much anything from Locust to Kodiak.

So yes, one person can single-handedly carry games, or rather sway a lot of close games in his favor. And I'm not even near the best players in this game skill-wise.

If you can't do it, it doesn't mean that someone else (like Zergling for example) can't do it either.
#timetogitgud

[PhoenixFire55]

Duke Nedo, on 25 November 2016 - 03:09 AM, said:

How can you tell? I seem to remember that the numerical PSR value was somehow used in MM, not just brackets for selection. Don't remember the source though, and it's a long time ago.

PSR value is irrelevant, as PSR gain is asymmetric and biased towards increase, i.e. everyone who is an average or better player will eventually reach fully maxed T1, thus the current tier-one-potatoe matches. All that PSR currently does is filter our complete terribads and griefing trash from everyone else.

[Zergling]

Duke Nedo, on 25 November 2016 - 05:29 AM, said:

Doesn't look that bad to me... remember that Tier 1 is skill-capped so Good and above will most likely all be Tier 1 with maxes XP-bar, thus counting the same. Above average to potato could be anything between Tier 1 and Tier 4.... so if you look at it like that it could very well be the same number for T1 players on both teams.

Well, the top 5%er on my team was me, and I'm only Tier 2, not 1.

Still, even if it does balance teams by PSR, that is definitely a terrible way to balance teams, given how widely skill levels vary in PSR Tier, and how much of a difference there was in that battle.

For my team to have won that battle is just a sheer fluke tbh.

Edited by Zergling, 25 November 2016 - 05:42 AM.

[STEF_]

Zergling, on 24 November 2016 - 04:07 PM, said:

Sorry, but you're wrong about that.

This month the only mechs I've been playing are those I've been leveling; my Steam Heavy pack mechs (Tbolt 9SE and Timby C) and Clan Hero mechs (the Kit Fox, Adder, Nova, Stormcrow, Timby and Warhawk).

I exclusively play the solo pug queue and I'm not using meta builds, yet I still have decent stats.

My overall leaderboard stats for this month: 1.3 W/L, 1.69 K/D, 285 average MS
Lights = 1.52 W/L, 1.44 K/D, 280 average MS
Mediums = 1.17 W/L, 1.60 K/D, 275 average MS
Heavies = 1.10 W/L, 1.76 K/D, 279 average MS
Assaults = 1.88 W/L, 2.73 K/D, 338 average MS


Basically, a decent player will produce decent stats even when leveling bad mechs in the solo queue.
When they are playing good mechs, they will just score even better.

Interesting stats.
However, I feel match score not the best factor to evaluate pilots.

You win when you kill the enemy. So at the end, I like to observe those 2 ratios: k/d and w/l.

I don't care about my 264 average match score....but I do care to improve my 2.38 w/l ratio and 2.21 k/d ratio.

Edited by Stefka Kerensky, 25 November 2016 - 05:45 AM.

[STEF_]

Zergling, on 25 November 2016 - 04:25 AM, said:

Most players play solo queue though, even the higher scoring players.


Also, see this battle screenshot, with (slightly rough) player skill estimations based on this month's leaderboard stats for each player:



My team had:
1x top 5% player
2x 'good' players with stats between top 25% and top 10%
2x 'above average' players with stats between the top 50% and top 25%
2x 'average' players with stats near the 50% mark
3x 'below average' players with stats between 50% and bottom 25%
2x 'potato' players with stats around bottom 25% or less


Enemy team had:
1x top 0.1% player
1x top 1% player
1x 'good' player
5x 'above average' players
2x 'below average' players
2x 'potato' players

If the matchmaker is trying to balance players based on PSR, it is doing a terrible job at actually balancing battles, as the enemy team definitely had a large skill advantage that battle, and honestly should have won.

Are those stats about match score?
Just asking because, as I said before, average match score is pretty irrilevant.

[Zergling]

Stefka Kerensky, on 25 November 2016 - 05:44 AM, said:

Interesting stats.
However, I feel match score not the best factor to evaluate pilots.

You win when you kill the enemy. So at the end, I like to observe those 2 ratios: k/d and w/l.

I don't care about my 264 average match score....but I do care to improve my 2.38 w/l ratio and 2.21 k/d ratio.

It was just rough work, but I accounted for W/L and K/D in addition to average MS. If they had bad W/L despite a decent MS, I rated them lower than their MS would indicate.

[Duke Nedo]

Zergling, on 25 November 2016 - 05:40 AM, said:

Well, the top 5%er on my team was me, and I'm only Tier 2, not 1.

Still, even if it does balance teams by PSR, that is definitely a terrible way to balance teams, given how widely skill levels vary in PSR Tier, and how much of a difference there was in that battle.

For my team to have won that battle is just a sheer fluke tbh.

Aye, I am not saying it's succeeds, I just believe it tries... In prime time that works quite OK in a way, there are enough high tier players in every match to average the huge skill disparity in T1.

Problem is when there are fewer players on and it starts to wildly mix tiers to fill up the matches and (if it tries to match PSR between teams) you are matched only against one or a few potentially very good T1 players on the other team... then you'd better carry your load and carry hard... at least that's what I tell myself when I get a loss streak every time I try to basic a new mediocre mech and am in need of being carried.

[STEF_]

Zergling, on 25 November 2016 - 05:51 AM, said:

It was just rough work, but I accounted for W/L and K/D in addition to average MS. If they had bad W/L despite a decent MS, I rated them lower than their MS would indicate.

Thanks.
I'm also interest in understanding if we really have a MM
If you have time, could you do the same evaluation, but taking into account w/l and k/d only?

I'm asking this because, imo, it's a very bad designed MM, if it is based on PSR (always up, never down) which is calculated in a bad match score system (which is, imo, irrilevant)

[NoiseCrypt]

Wil McCullough, on 25 November 2016 - 02:47 AM, said:

no, you numpty.

whether the OP is able or unable to carry his own weight by his own standards has no bearing on whether the benchmark he came up with is fair or not.

there's no fallacy here other than an ad hominem one. please find out what ad hominem means. if the word is too advanced for you, ask your parents to explain it.



all that posturing and still no attempt to participate in the actual point brought up by the OP.



thank you.

now if you have nothing to add other than your &quot;s/s or it didn't happen&quot; rhetoric, once again i'll ask you to keep quiet unless you'd like me to keep demolishing your ridiculous arguments.

The beautiful paradox of a guy wading in fallacies inorder to win his own version of the discussion. You are a true case study

I hope that you one day can move on from this methode

[Duke Nedo]

Stefka Kerensky, on 25 November 2016 - 05:44 AM, said:

Interesting stats.
However, I feel match score not the best factor to evaluate pilots.

You win when you kill the enemy. So at the end, I like to observe those 2 ratios: k/d and w/l.

I don't care about my 264 average match score....but I do care to improve my 2.38 w/l ratio and 2.21 k/d ratio.

May I ask how much you play in the group queue? I am curious how good w/l ration you can have in the long run as a 100% solo player.

I play about 95% in the solo queue mixing wildly between good mechs and underdogs, and when we do group up it's mostly just to yolo around in mixed builds drinking whiskey without much drop commanding going on. I rank like 12500 in w/l ratio but ranked 2994 for average score so it's a bit of a disconnect there.

Curious what makes that disconnect. Lack of pug-bossing? Too few drops in organized groups? My best meta-mechs have 1.5-2.0 w/l ratio but the sample is so small that doesn't have to mean anything...

Can you maintain a 2.0+ w/l ratio only dropping solo?

Edited by Duke Nedo, 25 November 2016 - 06:25 AM.