152 replies to this topic

[Belacose]

Mister Blastman, on 10 July 2016 - 07:48 AM, said:

The KDK-3 is exceptionally deadly in talented hands. Victoria is talented to begin with. Add KDK-3 and you have a nuclear arsenal without equal.

Not is he only using the Kodiak 3 for each and every one of those 285 matches, judging by that curb stomping 4.16 W/L ratio I'd also guess he's solely playing on an All-Star super team as well. He averages 2.8 kills per game.

Even with the KDK-3, wonder how his numbers would look if every game played was a QD pug?

BTW, if it turns out that his games were all QD pugs I'd be shocked, eat my words and want an autograph.

EDIT: Actually those numbers are impressive regardless of anything else. Wonder if he's wearing one of those new VR headsets which allow you to look out the side of the cockpit without having to turn your torso?

Edited by Belacose, 10 July 2016 - 10:52 PM.

[MrMadguy]

All, matchmaker tries to achieve - is W/L = 1, cuz it's simple ELO-based crappy matchmaker. My W/L is a little bit greater, than 1, as it has always been. Those, who still don't have W/L = 1 - haven't settled on their true rank yet or they're top tier players, who don't have skilled enough competitors. What is the problem? No matter, how low your real skill level is: if your W/L > 1 - then your rating rises. Just because it doesn't matter, how badly you personally play: if your team keeps carrying you - you PSR rises. Example: My K/D was 1.2 in the past. Now my K/D = 0.86 and keeps dropping. You can see my current stats. They aren't accurate enough due to low number of games played, but they show the trend - my short term K/D = 0.46 and average match score = 152. I'm bad player, I don't dispute it. But all of a sudden my rating keeps and keeps rising. Rising to a point, where game becomes completely unplayable for me. My teams essentially play without me, but...win anyway. And this can happen till the end of the world: team may be padded by bad players, but while teams have same amount of bads and W/L = 1 - MM doesn't see anything bad in doing it. And as W/L > 1, your rating also keeps rising, cuz PSR is biased toward increasing. And this process of rating inflating happens and happens (rating of bad players rises -> they all being carried all together -> their rating rises -> again they are being carried all together) -> your rating inflates to ridiculous values. I've even seen some famous streamers in my team - like Sean Lang. I try to find some players, with whom I usually being grouped, but I don't remember exact nicks. Player, I usually see, when I play on my usual rank, has nick, that is something like TACom, TACTom or TACTcom. Also the sneaky guy, who plays Lights and Mediums, and has long name like "XXXXX and the YYYYY".

Edited by MrMadguy, 10 July 2016 - 10:50 PM.

[iLLcapitan]

Don't think the leaderboards can be used for balancing purposes, tbh.
The amount of good players I've only seen in KDK-3s since leaderboards were introduced is sickening, but to each his/her own I guess.

[Ex Atlas Overlord]

iLLcapitan, on 10 July 2016 - 10:52 PM, said:

Don't think the leaderboards can be used for balancing purposes, tbh.
The amount of good players I've only seen in KDK-3s since leaderboards were introduced is sickening, but to each his/her own I guess.

Personally I'm doing it to demonstrate how grossly P2W that MWO is with every round of new mechs.

My best atlas KDR is 2.01
My best kodiak KDR is 3.55 and rising...

Same player, same MM, same teams....only difference is that I'm now using the latest cash mech.

[Lily from animove]

I can give you a ton of my match screenshots as I tend to screen them all, you can then make a list of all that pilots in them and do whatever you want with this data

[Champion of Khorne Lord of Blood]

I've noticed a trend
Everyone who has posted their results has had one thing in common. The team with more games played on average was the losing team.

I'd chalk it up to people who played less games have raised to their tier more efficiently, thus are better players than those who have played loads of games to get to the same tier. They might not have as much play time, but they gain experience at a faster rate, being faster learners they get better quickly. Chronicbads, as you've coined, will have played many games and moved up in tier from just grinding alone and will face off against these fast learning players throughout the tiers.

Not sure if anyone else noticed this yet since I've only read data posts and the OP.

[Lily from animove]

Dakota1000, on 10 July 2016 - 11:58 PM, said:

I've noticed a trend
Everyone who has posted their results has had one thing in common. The team with more games played on average was the losing team.

I'd chalk it up to people who played less games have raised to their tier more efficiently, thus are better players than those who have played loads of games to get to the same tier. They might not have as much play time, but they gain experience at a faster rate, being faster learners they get better quickly. Chronicbads, as you've coined, will have played many games and moved up in tier from just grinding alone and will face off against these fast learning players throughout the tiers.

Not sure if anyone else noticed this yet since I've only read data posts and the OP.

I wonder how you gather that data, PSR took matches MANY month into account, while the leaderboard is only tracking recent games. So I doubt there is a proper relation between matches played and how fats they rised in PSR. you also do not even know who palyed how long in which tier. what if Jeo played 1000games till T1 and also palyed another 1500. and Bob played 2000 till T1 but no more games?

[MrMadguy]

Dakota1000, on 10 July 2016 - 11:58 PM, said:

I've noticed a trend
Everyone who has posted their results has had one thing in common. The team with more games played on average was the losing team.

I'd chalk it up to people who played less games have raised to their tier more efficiently, thus are better players than those who have played loads of games to get to the same tier. They might not have as much play time, but they gain experience at a faster rate, being faster learners they get better quickly. Chronicbads, as you've coined, will have played many games and moved up in tier from just grinding alone and will face off against these fast learning players throughout the tiers.

Not sure if anyone else noticed this yet since I've only read data posts and the OP.

This also may be players, who: 1) Play on alt accounts to constantly reset their rating and stomp newbies 2) Haven't settled on their rank yet - therefore still have positive W/L or K/D. That's why current MM is also so terrible - it reacts too slowly. Sometimes I start thinking, that it's much better to have MM, that has empirical algorithm, than have "mathematically correct" one. I.e. sometimes it's better to have an illusion of matchmaker, that simply tries to provide as nice matches, as possible, than mathematically correct MM, that uses statistics, but can't react to fast situation changes and provide streaks of terrible matches instead, cuz it takes thousands of matches for player to settle on his true rating and it's extremely easy to trick it simply via creating alt accounts.

Edited by MrMadguy, 11 July 2016 - 12:40 AM.

[DemonRaziel]

Lily from animove, on 11 July 2016 - 12:09 AM, said:

I wonder how you gather that data, PSR took matches MANY month into account, while the leaderboard is only tracking recent games. So I doubt there is a proper relation between matches played and how fats they rised in PSR. you also do not even know who palyed how long in which tier. what if Jeo played 1000games till T1 and also palyed another 1500. and Bob played 2000 till T1 but no more games?

That's the funny thing about statistics - everyone can interpret them differently

Especially if there's a whole lot of factors affecting these stats that we can't really determine from the numbers alone.

For example, someone who levels 'Mech after 'Mech after 'Mech might have an issue with oftentimes running 'Mechs he's not too familiar with and that are lacking the skills, so he's at a disadvantage versus someone in a fully mastered and well equipped 'Mech. Thus his PSR might be quite low despite high amount of matches, but when he jumps into his favorite ride, he can do exceptionally well for his tier.

Then there's the issue of jumping the tiers too fast for your own good, such as LURMing your way through T5 and T4, only to realize you're suddenly way out of your league and a liability to your team.

[Lily from animove]

MrMadguy, on 11 July 2016 - 12:37 AM, said:

This also may be players, who: 1) Play on alt accounts to constantly reset their rating and stomp newbies 2) Haven't settled on their rank yet - therefore still have positive W/L or K/D. That's why current MM is also so terrible - it reacts too slowly. Sometimes I start thinking, that it's much better to have MM, that has empirical algorithm, than have "mathematically correct" one. I.e. sometimes it's better to have an illusion of matchmaker, that simply tries to provide as nice matches, as possible, than mathematically correct MM, that uses statistics, but can't react to fast situation changes and provide streaks of terrible matches instead, cuz it takes thousands of matches for player to settle on his true rating and it's extremely easy to trick it simply via creating alt accounts.

hundreds of games till T1 is just too much, that does not rate skill proper in a suitable time. a more harsh climb/fall with less needed games would be better to measure a pilots skill.

You can assume a true newbie who is well skilled will screw up his first matches reahcing T5 quickly, then when he knows how the game works, the way out of that hell is far too long and will make him be in the wrong brackets for a too long time.

DemonRaziel, on 11 July 2016 - 12:41 AM, said:

That's the funny thing about statistics - everyone can interpret them differently

Especially if there's a whole lot of factors affecting these stats that we can't really determine from the numbers alone.

For example, someone who levels 'Mech after 'Mech after 'Mech might have an issue with oftentimes running 'Mechs he's not too familiar with and that are lacking the skills, so he's at a disadvantage versus someone in a fully mastered and well equipped 'Mech. Thus his PSR might be quite low despite high amount of matches, but when he jumps into his favorite ride, he can do exceptionally well for his tier.

Then there's the issue of jumping the tiers too fast for your own good, such as LURMing your way through T5 and T4, only to realize you're suddenly way out of your league and a liability to your team.

yes and a more agile PSR would rerate him quicker depending on which mech he uses.

Edited by Lily from animove, 11 July 2016 - 12:44 AM.

[MrMadguy]

Lily from animove, on 11 July 2016 - 12:42 AM, said:

hundreds of games till T1 is just too much, that does not rate skill proper in a suitable time. a more harsh climb/fall with less needed games would be better to measure a pilots skill.

You can assume a true newbie who is well skilled will screw up his first matches reahcing T5 quickly, then when he knows how the game works, the way out of that hell is far too long and will make him be in the wrong brackets for a too long time.

MM, that has faster rating changing - isn't solution either. Yeah, it would react faster, but it won't be stable then - it would be affected by RNG too much then. We simply need smarter algorithm. One of the simplest, but effective, empirical smart algorithms - is simple anti-bad luck protection. There should be short term rating modifier, that kicks in, when player experiences 3 and more decent/terrible matches in a row. And in a long term some kind of skill profile should be built - statistics, that show, in what situations particular player has better or worse performance, and build his matches accordingly. For example: if he prefers to play brawlers - don't put exceeding number of snipers/LRM on the opposite team.

Edited by MrMadguy, 11 July 2016 - 12:57 AM.

[Lily from animove]

MrMadguy, on 11 July 2016 - 12:52 AM, said:

MM, that has faster rating changing - isn't solution either. Yeah, it would react faster, but it won't be stable then - it would be affected by RNG too much then. We simply need smarter algorithm. One of the simplest, but effective, empirical smart algorithms - is simple anti-bad luck protection. There should be short term rating modifier, that kicks in, when player experiences 3 and more decent/terrible matches in a row. And in a long term some kind of skill profile should be built - statistics, that show, in what situations particular player has better or worse performance, and build his matches accordingly. For example: if he prefers to play brawlers - don't put exceeding number of snipers/LRM on the opposite team.

faster could mean takign the last 100 matches into account. instead hundrets per tier to climb/rise.

dividing further into brawling and such, I doubt statistics can track this and MM handling this ontop.

[MrMadguy]

Lily from animove, on 11 July 2016 - 01:10 AM, said:

faster could mean takign the last 100 matches into account. instead hundrets per tier to climb/rise.

dividing further into brawling and such, I doubt statistics can track this and MM handling this ontop.

This would mean exactly what I say - bigger rating changes per game. +10 instead of +1 for example. MM can become unstable due to this - i.e. your rating may start to swing up and down. But at least per 'Mech rating is 100% needed. In AWS-8R and can deal up to 1K dmg without any effort - i.e. earn 500 match score easily in several matches in a row. But sometimes I can't deal even 100 dmg in some obsolete crappy Assaults with brawler builds, like Highlander. And I like to play different 'Mechs and change them quite often.

[Hit the Deck]

Belacose, on 10 July 2016 - 10:38 PM, said:

Not is he only using the Kodiak 3 for each and every one of those 285 matches, judging by that curb stomping 4.16 W/L ratio I'd also guess he's solely playing on an All-Star super team as well. He averages 2.8 kills per game.

Even with the KDK-3, wonder how his numbers would look if every game played was a QD pug?

BTW, if it turns out that his games were all QD pugs I'd be shocked, eat my words and want an autograph.

EDIT: Actually those numbers are impressive regardless of anything else. Wonder if he's wearing one of those new VR headsets which allow you to look out the side of the cockpit without having to turn your torso?

For a more realistic depiction of how the KDK-3 scores in the hand of a good player, you can search for UCCU on the leaderboard. He's an Oceanic player who has ben almost exclusively playing Dakka Dire Wolf and Kodiak (now only kodiak since its release). He pugs and a k/d score of around 4 would be fitting I feel (can't load up MWO right now so I just made a guess).

Edited by Hit the Deck, 11 July 2016 - 03:31 AM.

[Belacose]

Hit the Deck, on 11 July 2016 - 01:25 AM, said:

For a more realistic depiction of how the KDK-3 scores in the hand of a good player, you can search for UCCU on the leaderboard. He's an Oceanic player who has been almost exclusively played Dakka Dire Wolf and Kodiak (now only kodiak since its release). He pugs and a k/d score of around 4 would be fitting I feel (can't load up MWO right now so I just made a guess).

[color=#EEEEEE]No results found.[/color]

[MrMadguy]

I mean, that it's possible, that it would be more effective to simply match players with particular builds, instead of matching players with particular ratings.

But PGI simply ignores anything, that is related to improvements of matchmaker. I guess, cuz allowing players to play on 'Mechs, they like, instead of current P2W Meta ones - isn't in their interests.

[Tarogato]

imo, the whole effort is kinda pointless if you blank the names out. In order to verify that MM is working properly, PGI should check the tier/PSR of all pilots in an "unbalanced" match. If the PSR appear to be balanced, then the problem is PSR, not matchmaker.

[Idealsuspect]

This game would have a better "MM" if teams were builded randomly with 3/3/3/3 system.

Better than a PGI solution.

[Mazzyplz]

so winning teams always have better kdr?

why not match kdr average between teams when matchmaker creates match?

[RePlayBoy101]

im bored of canstantly carrying games -.-