Jump to content

Why Don't We Use W/l Ratio Instead Of Psr...?


62 replies to this topic

#21 VitriolicViolet

    Member

  • PipPipPipPipPipPipPip
  • The Corsair
  • The Corsair
  • 592 posts
  • LocationAustralia, Melbourne

Posted 02 June 2018 - 02:44 PM

View PostBWS2K, on 02 June 2018 - 11:32 AM, said:

Since the game is a shooter and role warfare is... not really done well... would it make sense to just let PSR equal an average of damage done per match or something? T1 would be the top damage-dealers, etc. Limit it to a single factor that aligns with what the game is supposed to be about (not my words - MWO calls itself a tactical shooter)? Maybe that leads to the same problem though. Hmm...


This doesnt work either unfortunately. not all damage is equal. 800 damage with LRMs or LBXs is entirely different to 800 damage with MPLs or Gauss rifles.) 800 damage with LRMs can be achieved just by shooting a single Cataphract over the match, 800 damage with Gauss (if you can aim) will often net 3 kills or more.

An Atlas has something like 900 health across its components but you can kill it by only doing about 150 damage into the CT, whereas spread weapons will probably do 400-800 damage before killing it. damage efficiency

Edited by VitriolicViolet, 02 June 2018 - 02:45 PM.


#22 BWS2K

    Member

  • PipPipPipPipPipPip
  • 215 posts
  • LocationUSA

Posted 02 June 2018 - 02:58 PM

The more I learn about PSR, the more I learn how complex solving this issue really is, lol

#23 JudauAshta

    Member

  • PipPipPipPipPipPip
  • Big Brother
  • Big Brother
  • 264 posts

Posted 02 June 2018 - 03:05 PM

View PostBWS2K, on 01 June 2018 - 03:53 PM, said:

I'm not totally sure I understand what you mean but if a person is always playing to win, whatever the cost (and I realize that's not always the case), then wouldn't W/L represent that better than the PSR? Maybe it doesn't help for MatchMaking though...


what he means is that in team based games individual skill doesn't matter that much, the amount of baboons on your team do!!

enemy team has more lead drinking kids then congrats you win
your team got more super stupids tough luck you lose

this is true for other games like lol and wot as well

#24 Fleeb the Mad

    Member

  • PipPipPipPipPipPip
  • Urban Commando
  • Urban Commando
  • 441 posts

Posted 02 June 2018 - 04:09 PM

In games like WoT that don't have skill-based matchmaking win/loss rate really is the best metric to measure player skill, though it's skewed like everything else by factors like grouping.

It's not pointless because you're one of twelve people on your team. Over a large number of games (hundreds or thousands), the difference between win/loss rates from 50/50 are the ones where you were the difference in swinging the outcome. In WoT the average player had 49% wins, 49% losses and 2% draws. An elite player might only have a 55% win rate.

That's because the performance of the whole team is a landslide that one person struggles to influence once way or the other, but in a minority of near-balanced matches, that player and only that player was the deciding factor between victory and defeat for the team.

Problem with applying that to MWO is that once you try to equalize the skill pool at various levels, people can't maintain high win/loss ratios as the competition gets tougher. That's why you see people on alts topping leaderboards by farming in the low tiers. Win/loss alone isn't enough when you try to divvy up the skill pool. You need skill brackets similar to tiers as well as a rating system to determine whether players move up or down through them.

#25 S O L A I S

    Member

  • PipPipPipPipPipPip
  • Giant Helper
  • 390 posts
  • LocationFlorida

Posted 02 June 2018 - 04:36 PM

View PostNightbird, on 02 June 2018 - 02:41 PM, said:


You mean number of matches played over skill right? Or number of wins if you put it another way, without regard to win rate (percentage).


No I mean you go up even if carried in a win. Makes no sense to give someone a psr increase if they so happen to win while doing a hundred damage.

#26 Nightbird

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • The God of Death
  • The God of Death
  • 7,518 posts

Posted 02 June 2018 - 08:53 PM

Nope, with the current system, skill > winning in that even if you get the losing team all the time, you can still go up the ranks via personal performance. The problem is only that the PSR rating changes awarded are net positive in that a player no matter how bad will go up the rankings with sufficient matches. I would prefer a system where you can lose as much rating on a bad performance on a loss as you can gain with a good performance on a win, so that someone who isn't learning the game never leaves their tier.

Edited by Nightbird, 02 June 2018 - 08:56 PM.


#27 Nothing Whatsoever

    Member

  • PipPipPipPipPipPipPipPipPip
  • Bad Company
  • Bad Company
  • 3,655 posts
  • LocationNowhere

Posted 02 June 2018 - 10:49 PM

View PostBWS2K, on 02 June 2018 - 02:58 PM, said:

The more I learn about PSR, the more I learn how complex solving this issue really is, lol


I've missed some three years or so, due to needing to get my hardware in order with many RL changes, like my gear aging, not to mention moving, and caring for my 94 year old grandma, so yeah she was way more important than something like MWO.

With that stated sure, the devs did have lots of collected data on matches, and had many vocal players on various topics, where I still feel that not enough has been done in Heat Capacity, Heat Dissipation, Target Interlock Circuits, aspects of Battletech that could be of benefit.

Ideas that could have helped MWO grow and go beyond the original format and the previous MechWarrior games that also grappled with transitioning the rules to make alive action game like MWO, where the MechWarrior is the active player and not a commander telling MechWarriors what to do and where to go. And yes, I'm gonna save enough to eventually get the new BattleTech game.



I want to believe that MWO can still see these areas as aspects that evolve with how MWO currently works, at least in a Public Test arena.

Edited by Nothing Whatsoever, 02 June 2018 - 11:02 PM.


#28 Dogstar

    Member

  • PipPipPipPipPipPipPipPip
  • Bad Company
  • Bad Company
  • 1,725 posts
  • Twitch: Link
  • LocationLondon

Posted 03 June 2018 - 12:28 AM

The matchmaker matches players with similar experience not similar skill because tier is nothing to do with skill. Changing to a skill based matchmaker would make no significant difference to the number of stomps (other than possibly increase them) because stomps are a design feature of quickplay.

It's all explained here.

#29 Sjorpha

    Member

  • PipPipPipPipPipPipPipPipPip
  • Philanthropist
  • Philanthropist
  • 4,477 posts
  • LocationSweden

Posted 03 June 2018 - 01:50 AM

PSR could work fairly well as it is if it wasn't biased towards increasing, the main problem is that it's way too easy to get high PSR.

#30 MischiefSC

    Member

  • PipPipPipPipPipPipPipPipPipPipPipPip
  • The Benefactor
  • The Benefactor
  • 16,697 posts

Posted 03 June 2018 - 06:33 AM

If only we had some mathematical way to solve for the value of 1 constant in a team of 12. Some sort of, I dunno, math for statistics and stuff. You could gauge predictions and accuracy and such. We could solve all sorts of things like how to build long bridges or tall buildings, or astrophysics. Someone could invent epidemiology! We could make a big financial market like a stock market using that sort of math. Expand medical and chemical research via statistical analysis of changes in chemical states.

Wow, wouldn't it be cool if we had a real understanding of mathematics and could figure out what 1 divides by 12 was as a percent and then measure the relative impact of that percent over a growing sample size?

Unfortunately we obviously don't and so all results of any matches are totally incomprehensible and actually proof of magic and irreducible complexity, invalidating all science and human progress. I'm waiting for the arrival of varies and unicorns to fix the worlds ills.

Or, you know, you can just look at the leaderboards. Do you see the part where better players have a better average win/loss? That's because better players drive more wins on average than bad players. Because, scary I know, math is actually a thing.

#31 BWS2K

    Member

  • PipPipPipPipPipPip
  • 215 posts
  • LocationUSA

Posted 03 June 2018 - 07:02 AM

View PostMischiefSC, on 03 June 2018 - 06:33 AM, said:

If only we had some mathematical way to solve for the value of 1 constant in a team of 12. Some sort of, I dunno, math for statistics and stuff. You could gauge predictions and accuracy and such. We could solve all sorts of things like how to build long bridges or tall buildings, or astrophysics. Someone could invent epidemiology! We could make a big financial market like a stock market using that sort of math. Expand medical and chemical research via statistical analysis of changes in chemical states.

Wow, wouldn't it be cool if we had a real understanding of mathematics and could figure out what 1 divides by 12 was as a percent and then measure the relative impact of that percent over a growing sample size?

Unfortunately we obviously don't and so all results of any matches are totally incomprehensible and actually proof of magic and irreducible complexity, invalidating all science and human progress. I'm waiting for the arrival of varies and unicorns to fix the worlds ills.

Or, you know, you can just look at the leaderboards. Do you see the part where better players have a better average win/loss? That's because better players drive more wins on average than bad players. Because, scary I know, math is actually a thing.

MischiefSC, I knew you'd show up eventually. I will never think about PSR, or statistics, ever again without thinking of you (there was a thread a while back where someone was putting up a stink and you were laying some serious maths down), lol. ;)

I wonder what the average match score is in each Tier though? The PSR is not weighted for advancement below 250pts. If you aren't consistently scoring that much, you aren't going to 'float up'. Is a puzzlement.

#32 MischiefSC

    Member

  • PipPipPipPipPipPipPipPipPipPipPipPip
  • The Benefactor
  • The Benefactor
  • 16,697 posts

Posted 03 June 2018 - 07:45 AM

View PostBWS2K, on 03 June 2018 - 07:02 AM, said:

MischiefSC, I knew you'd show up eventually. I will never think about PSR, or statistics, ever again without thinking of you (there was a thread a while back where someone was putting up a stink and you were laying some serious maths down), lol. ;)

I wonder what the average match score is in each Tier though? The PSR is not weighted for advancement below 250pts. If you aren't consistently scoring that much, you aren't going to 'float up'. Is a puzzlement.


PSR would work better if we had 100k players. Any MM would. Currently PSR though is broken in that because you don't have tiers broken up among each other as they're supposed to be so individual matches have a higher variance - as in really good and really bad matches. Your average is still going to be pretty accurate but each match will be more extreme. The problem there is that PSR has a strong upward bias so your good matches gain plenty and your bad matches only lose a bit. So unless you're absolutely terrible you'll eventually move up to T1.

It is also blind to the caliber of who you're playing agaimst, so bearing a horrible team by tons gains you more than barely beating a team that should have beaten you, even if the latter took more skill.

It's why an individual scoring system will never be accurate and why in reality only w/l matters. How you drive a win is irrelevant. What matters is who won and who they beat. That, over time, gives the only accurate reflection of who is more likely to beat who and that's all the matchmaker cares about.

PSR and match score is irrelevant to what a matchmaker tries to do. It wants to make a match balanced around who is going to win. How they win doesn't matter, that they win and against who does matter.

#33 BWS2K

    Member

  • PipPipPipPipPipPip
  • 215 posts
  • LocationUSA

Posted 03 June 2018 - 08:07 AM

Posted Image

Is this graphic accurate then, with the understanding that dealing more damage offsets needing to score in other ways (the last sentence)? Overall, PSR is weighted to help players rise in rank... if they are scoring more than 400pts consistently (and the values in change are equal to their worded counterparts). Or is there something else going on that I'm missing?

Because it feels like the explanation that there are some folks who can't escape T5, or who are plateauing earlier than T1, is that they can neither get more wins than losses or consistently score 400+ in MS and both of those get more difficult as one rises through the tiers. Presumably the T1's who don't feel like they belong up there are being carried through wins and not consistently scoring 400+ anymore?

As an aside, I get the sense that Team Damage is weighted pretty severely. It's hard to have a match where you don't get at least a point of it simply because you were walking when someone ran up and walked into you from behind or right from spawn. Intentional Team Damage should be punished but a couple points is basically par for the course. That's just a feeling I have though. I try to help folks understand this system as much as possible, even if it is pretty broken, because I'm a guy who likes internal consistency and accuracy. Probably ought to mosey on over to another game, lol.

#34 Revis Volek

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • The 1 Percent
  • The 1 Percent
  • 7,247 posts
  • Google+: Link
  • Facebook: Link
  • LocationBack in the Pilots chair

Posted 03 June 2018 - 08:33 AM

View PostEl Bandito, on 01 June 2018 - 07:29 PM, said:

GQ stats should really be separate from SQ stats. Silly PGI...



Aside from the fact that things like Dmg and kills are much harder to come by there is also the nature of team cohesion of a large unit vs a few small ones doing there own things.

They dont understand stats or the weight they have on how to sort players and mechs in the game. They talk about "NUMBERS" but i think they assume they can look at everthing in a vacuum and the game itself didnt matter. SO what this Atlas is a bad mech it got 7 kill in this game where the rest of the team did all the work.


I just dont feel like they dont know what they are looking at all the time and it leads to them thinking stats are pointless or less important then they are.

Edited by Revis Volek, 03 June 2018 - 08:33 AM.


#35 JC Daxion

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Liquid Metal
  • Liquid Metal
  • 5,230 posts

Posted 03 June 2018 - 08:35 AM

a better question is...

Why don't we use leader boards to set your tier.

It still should be harder to go down.. But there is no way a guy that hit top 1k is not a good player.. I know i'd probably be somewhere in t2, i just don't have the twich skills of those top guys. I also don't constantly play meta/top mechs.. At the same time a person that is listed constantly at 30k is not so hot. It should also not drop fast, so players don't just tank a few matches a month to drop lower..

If you can hit 2k-5k... and then a bunch of months you get 25k, it's not cause you suddenly suck. Your either screwing around (not a bad thing which i think battle value would help), or tanking matches


I also wouldn't mind seeing a battle value added into the mix. I'm not saying a T1 in a crap mech should be counted as a 5 but some sort of system might help things a bit.

Edited by JC Daxion, 03 June 2018 - 08:40 AM.


#36 Revis Volek

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • The 1 Percent
  • The 1 Percent
  • 7,247 posts
  • Google+: Link
  • Facebook: Link
  • LocationBack in the Pilots chair

Posted 03 June 2018 - 08:37 AM

View PostBWS2K, on 03 June 2018 - 07:02 AM, said:

MischiefSC, I knew you'd show up eventually. I will never think about PSR, or statistics, ever again without thinking of you (there was a thread a while back where someone was putting up a stink and you were laying some serious maths down), lol. Posted Image

I wonder what the average match score is in each Tier though? The PSR is not weighted for advancement below 250pts. If you aren't consistently scoring that much, you aren't going to 'float up'. Is a puzzlement.



Problem is majority of the player base DOES in fact score between 200-250 but if they get a win on a 200 W or L they still dont go down. IT needs to be just as steep of a downward slope on the other side of the 250 (or current player avg should change by season) if you are doing worse then the majority of the playerbase.

More maths...

https://leaderboard.isengrim.org/stats

#37 Asym

    Member

  • PipPipPipPipPipPipPipPipPip
  • Nova Captain
  • 2,186 posts

Posted 03 June 2018 - 09:15 AM

View PostJC Daxion, on 03 June 2018 - 08:35 AM, said:

a better question is...

Why don't we use leader boards to set your tier.



Because Leader Boards are about numbers a group helps produce... It's not about what "I" did but about what "we" did... Some would argue that S7 rankings might be more accurate because it is a 1x1 fight...

The other issues is that there is no "average amount of gameplay".... Some people play 1000 hours and some less than 10 a month..... Another issue is "category of play".....are you in a fully optimized mech or just starting to master it???

Mech effectiveness, pilot effectiveness and team effectiveness all combine for a skill ratio for each pilot based on the mechs they want to use! A Light pilot and an Assault can both be great pilots and yet, where would you see their skill? Kills? KMDD? Assists? something else? It's a complex calculation that renders a combat effectiveness metric a pilot earns....

What we have now is worthless because if lacks dimension and exists on one plane of existence: win or Kill.....both are binary in a game of multifaceted complexity. And, it could be fixed and fielded ! All of the metrics we need are available; just, not used or thought about.....sigh.

#38 JC Daxion

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Liquid Metal
  • Liquid Metal
  • 5,230 posts

Posted 03 June 2018 - 09:23 AM

View PostAsym, on 03 June 2018 - 09:15 AM, said:


Because Leader Boards are about numbers a group helps produce... It's not about what "I" did but about what "we" did..



I beg to differ.. Look at the leader boards top players.. they are all very good players. good players kill, and they win.. Bad get carried along.


Anyone can go a win streak of good teams and get a positive win/loss.. Your not going to be at the top of the leader boards with many matches played if your not good.

I've had tons of matches with 1k+ damage and 5+ KMMD's and 8 kill matches, but i never broke into the top of the boards, those guys are just better. To me it just means the people on the other team were worse than me. I can often tell must by looking at the pre-match scoreboard and pick out names and know which team is going to win. They are always at the top of the boards and great players.

#39 BWS2K

    Member

  • PipPipPipPipPipPip
  • 215 posts
  • LocationUSA

Posted 03 June 2018 - 09:25 AM

Like, I'm reading these responses and I'm thinking "Oh, that seems fixable! Just... oh wait, [insert other variable]. Never mind."

What I find the most often is that people will dismiss the PSR saying it's something it's not or that it does something it doesn't and then the debate devolves from there (in-game or something, not this thread). What do other similar games do - are they also in a bind like MWO? Maybe they just have a lot more players?

#40 Tarl Cabot

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Tai-sho
  • Tai-sho
  • 7,770 posts
  • LocationImperial City, Luthien - Draconis Combine

Posted 03 June 2018 - 10:47 AM

View PostBWS2K, on 03 June 2018 - 08:07 AM, said:

Posted Image

Is this graphic accurate then, with the understanding that dealing more damage offsets needing to score in other ways (the last sentence)? Overall, PSR is weighted to help players rise in rank... if they are scoring more than 400pts consistently (and the values in change are equal to their worded counterparts). Or is there something else going on that I'm missing?


400+MS (points) is the extreme on rushing through tiers and max out Tier 1. Damage divided by half is applied to the MS (I believe this is still valid). A player playing most drops in Group queue while being carried and winning most of the time can have a much lower, overall MS while being a team player. But get that same player in Solo queue on a tier 1 drop with people who are not playing as a team, with no communication, running non-optimized builds, etc, would likely be completely out of his element.

Take myself as an example. After 23 seasons, I have an average MS 240, average 1.12 W/L with almost all matches in Solo queue except for maybe 10 in Group (most drops as a group/unit have been in FP). And I hit Tier 1 in Nov 2016 (Season 5) and was max out 5 seasons/months later. I play well in a group setting, ie a soldier, but I am not even close to being in the same caliber range as some others. If the PSR movement was not as bias as it is now, I would likely be in Tier 2. When PSR was introduced I was seeding as a high Tier 3 from Elo stats when PSR went live Aug 2015. So 14 months to move from approx Tier 3.7 to Tier 1 then another 5 months to max out T1.

Quote

Because it feels like the explanation that there are some folks who can't escape T5, or who are plateauing earlier than T1, is that they can neither get more wins than losses or consistently score 400+ in MS and both of those get more difficult as one rises through the tiers. Presumably the T1's who don't feel like they belong up there are being carried through wins and not consistently scoring 400+ anymore?

As an aside, I get the sense that Team Damage is weighted pretty severely. It's hard to have a match where you don't get at least a point of it simply because you were walking when someone ran up and walked into you from behind or right from spawn. Intentional Team Damage should be punished but a couple points is basically par for the course. That's just a feeling I have though. I try to help folks understand this system as much as possible, even if it is pretty broken, because I'm a guy who likes internal consistency and accuracy. Probably ought to mosey on over to another game, lol.


For those who cannot escape T5-T4 (and some who equip/play most games to stay in said tiers) it is due to generating sub-175 MS for a variety of reasons - potato system, health reasons, a combination of those two, or they have made personal choices in mech construction (example lore type folks - stock mechs) and the choices they make or do not make on the field.

Take someone who has played primarily assaults for the last 2 years while remembering that 50% of damage is applied to MS. If a player only does 150 damage in an assault that converts to 75 MS.

As for what other games do? Do most other games allow mixing of faction/tech for their standard gameplay? One can read other game boards and there are always complaints about the MM and what information it uses to determine group makeup, and still have stomps. MWO simply does not have the player base to smooth things out + way too many different queues for said playerbase.

Edited by Tarl Cabot, 03 June 2018 - 10:51 AM.






3 user(s) are reading this topic

0 members, 3 guests, 0 anonymous users