Jump to content

Doesn't 1.0 W/l Ratio Mean Matchmaker Is Working Well?


114 replies to this topic

#61 LordKnightFandragon

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • 7,239 posts

Posted 18 January 2015 - 06:09 PM

View PostEscef, on 18 January 2015 - 02:55 AM, said:


Because ELO is based on WINS AND LOSSES, NOT KILLS AND DAMAGE TOTALS.

How is that difficult?



that is why elo should take into account all the matters of a players performance.

Average damage per weight class, the higher the weight class, the more avg damage matters.
The player's average xp
Kills, deaths, assists, how often the player triggers those various role warfare rewards.

Just anything and everything pertaining to a player's performance over hte course of a match. it then takes all that and comes up with your elo rating, not just.....you have a 49% win rate, you get a rating of 100, then a the guy with a 59% gets a rating of 180 and go.

We should have a "Cadet Q", where for your first 100 games your elo is zero(0). During the first 100 games, it takes everything you do and at game 101 adds it up averages it and does the math to come up with your final elo. Beyond that, its up to you to improve it or degrade it.

#62 Egomane

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • 8,163 posts

Posted 18 January 2015 - 09:53 PM

View PostLordKnightFandragon, on 18 January 2015 - 06:09 PM, said:



that is why elo should take into account all the matters of a players performance.

Average damage per weight class, the higher the weight class, the more avg damage matters.
The player's average xp
Kills, deaths, assists, how often the player triggers those various role warfare rewards.

Just anything and everything pertaining to a player's performance over hte course of a match. it then takes all that and comes up with your elo rating, not just.....you have a 49% win rate, you get a rating of 100, then a the guy with a 59% gets a rating of 180 and go.

We should have a "Cadet Q", where for your first 100 games your elo is zero(0). During the first 100 games, it takes everything you do and at game 101 adds it up averages it and does the math to come up with your final elo. Beyond that, its up to you to improve it or degrade it.

Except, it wouldn't be Elo anymore. I can't even imagine a formula that could take all those things together and put it into a comperable value. Even if one exists, it would most likely spawn just as many threads declaring such a formula invalid for various reasons, just like they are there for Elo.

"But kills are way to heavily weighted!"
"My assists need to better valued!"
"Why are spotting assists rated so low, they should be more important!"
"Why are spotting assists rated so high, they should be less important!"
And so on ...

And I doubt that you'd get better matches out of it.

Edited by Egomane, 18 January 2015 - 09:54 PM.


#63 Rorvik

    Member

  • PipPipPipPipPipPip
  • Bad Company
  • Bad Company
  • 230 posts
  • LocationYYZ

Posted 18 January 2015 - 10:45 PM

View PostRocketDog, on 18 January 2015 - 04:49 AM, said:

Incidentally, one of the devs (Kar Berg?) said in a recent post that they actually see more 12-0 stomps when the team Elos are more closely matched than when they are not. I suspect that the mechanics of combat in MWO mean that if both teams are competent then it's very hard to come back from being even two mechs down - which is certainly my experience in pug land.


If that is true, then it CLEARLY shows that the MM is NOT working.

I disagree with the assessment that if there are equally skilled players on both sides that it would naturally lead to steam rolls. If you have 24 roughly equally skilled players on both sides, then losing one or two Mechs should not be the end of the match. I have been in matches where teams (either mine or the other team) have managed to come back from almost three Mech deficits, and those matches usually ended up being close, tough, satisfying matches. But they are the rare exception rather than the rule.


View PostTheodore42, on 18 January 2015 - 01:50 PM, said:

Awesome, I didn't know that. Makes perfect sense. If you've ever watched a chess match, and 1 player loses a pawn and can't take a pawn back, the commentators go crazy like 1 pawn is the end of the match. When I play chess it doesn't really matter if I am up or down a pawn, but a GM can exploit the weakness of losing just one pawn and turn it into a win.


It would make sense if there were two commanders going head-to-head controlling 12 Mechs each, but that's not the way it works. It's more like a Chess game where the pieces are all independent of one another and are incapable of coordinating due to no voic chat and no ability to send meaningful messages to their team.

Occasionally I can make a fair guess as to whether my team will win or lose simply by the way they move on the map and the positions they take up. In the occasions where my guess is wrong, it's often because the other team does something stupid or fails to capitalize on our mistakes.


View PostTheodore42, on 18 January 2015 - 01:50 PM, said:

Also, my lowest win ratio on a mech is 1.08, so MM definitely doesn't make losing matches to force even ratios.


How does that follow? How is it even relevant to the point?

#64 Rampancy

    Member

  • PipPipPipPipPipPipPip
  • The Blade
  • The Blade
  • 568 posts

Posted 18 January 2015 - 11:18 PM

View PostSpr1ggan, on 18 January 2015 - 01:59 PM, said:


True. That's why I left Lords. Not good enough for the solo queue let alone playing for one of the best teams in the game. I still score sub 100 damage games in solo on a regular basis.
Just wanted to hop in to say that you're a good dude and I'd gladly play with you again :)

#65 kesmai

    Member

  • PipPipPipPipPipPipPipPipPip
  • The Spear
  • The Spear
  • 2,429 posts
  • LocationPirate's Bay

Posted 19 January 2015 - 03:25 AM

Is there even a "gobal" elo?



i have the feeling it is at least weight based.
been mastering 4 cicadas on the weekend scoring a w/l of 4:1. Then i switched to level the trebs (all of them bought in the rasalhague sale). And bam with the trEbs unlevelled there was a complete different playerbase and playstyle. Maybe elo works even on a per chassis basis?

Edited by kesmai, 19 January 2015 - 03:26 AM.


#66 Egomane

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • 8,163 posts

Posted 19 January 2015 - 03:29 AM

View Postkesmai, on 19 January 2015 - 03:25 AM, said:

i have the feeling it is at least weight based.
been mastering 4 cicadas on the weekend scoring a w/l of 4:1. Then i switched to level the trebs (all of them bought in the rasalhague sale). And bam with the trEbs unlevelled there was a complete different playerbase and playstyle. Maybe elo works even on a per chassis basis?

It's per weight class (light, medium, heavy and assault), so you have four different Elo values.

#67 norus

    Member

  • PipPipPipPipPipPip
  • Philanthropist
  • 222 posts
  • LocationN.A.

Posted 19 January 2015 - 03:37 AM

View PostEgomane, on 18 January 2015 - 09:53 PM, said:

Except, it wouldn't be Elo anymore. I can't even imagine a formula that could take all those things together and put it into a comperable value. Even if one exists, it would most likely spawn just as many threads declaring such a formula invalid for various reasons, just like they are there for Elo.


WN8 in world of tanks. It bases your "elo" upon damage kills spotting etc per tank type and then assigns a number for how good your stats are compared to the stats of every other player. So someone who has extremely good stats in all their tanks has a very high overall wn8 value whereas someone who only has lots of damage and kills in op tanks will be compared against everyone else using the op tank too and won't have that high of a score since the average dmg etc for that tank is high to begin with.

EDIT: It's a completely player made system too, based off the stats that are public information so everyone can see. The makers frequently update it to make it even better at assigning numerical values to skill

Moar edit: In case you're wondering, yes it does work extremely well. The best WILL rise to the top and there's a distinct difference in the ability of someone at say 2000 and a complete potato who's at 800

Edited by norus, 19 January 2015 - 03:44 AM.


#68 LordKnightFandragon

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • 7,239 posts

Posted 19 January 2015 - 04:43 AM

View PostEgomane, on 18 January 2015 - 09:53 PM, said:

Except, it wouldn't be Elo anymore. I can't even imagine a formula that could take all those things together and put it into a comperable value. Even if one exists, it would most likely spawn just as many threads declaring such a formula invalid for various reasons, just like they are there for Elo.

"But kills are way to heavily weighted!"
"My assists need to better valued!"
"Why are spotting assists rated so low, they should be more important!"
"Why are spotting assists rated so high, they should be less important!"
And so on ...

And I doubt that you'd get better matches out of it.



WoT's XVM, WN7, 8 and w/e other system uses stuff like that, and it seems to be fairly spot on in player performance. They dont use it to create games, but when you do get in a game and your side has 12 "reds" and a "green" and maybe 2 "yellows", and on the other side you see 1 purple, 4 blues, 6 greens and a hodgepodge of yellows, you can almost bet your life that the 2nd team will win, and it wont even be a contest.

So, somehow, Im sure that if they came up with a system somewhere around WoT's XVM mods and used that to create a match, but not in the same rigged manner WoT does it, the matches would be better. If thegame started to pair about equally skilled players on each side, the games might be better. When you have about equally paired teams in WoT, it usually isnt a 15-2 wipe. Typically when the teams are about equally matched, barring RNG, its typically a 15-10 match with some base capping.

BUt when you see the one side with solid red and the other is a mix of better players, its a cake walk....

#69 SgtKinCaiD

    Member

  • PipPipPipPipPipPipPipPip
  • Overlord
  • Overlord
  • 1,096 posts
  • LocationBordeaux

Posted 19 January 2015 - 05:03 AM

View PostEgomane, on 18 January 2015 - 02:41 AM, said:

View Postmogs01gt said:

MM in MWO tries to equalize the matches by adding in playes way below your ELO

Why is this obviously false information still spread around as a truth?
Why do players not care to educate themself and keep spreading this stuff?


Not going to be rude but this exactly what the MM does to handle high and low ELO players => http://mwomercs.com/...ost__p__2633230

And my response from another topic about the MM => http://mwomercs.com/...ost__p__3973139

SgtKinCaiD said:

I must agree with Sandpit about the MM and the streaks of wins or losses : the MM does not work like that.
It tries to setup match where both team have a 50% chance of winnning (at least it tries to be as close as possible).

Now the real problem is how the MM assembles teams to get a match. Right now, it only considers the ELO sum up of each team to create a "fair" match, or it is not a good indicator of the actual skill of the team, especially for teamplay.
Plus, the rule of mixing high ELO players with low ELO players in order to have an average ELO easier to work with, is not working : right from the start, the team is already missing firepower from one (if not more) player which is a huge disavantage given the nature of the game and the snowball effect.
Another problem is that a player ELO is based on W/L only, or past a certain point, a low ELO player won't decrease anymore because basicaly he's caried by his team every time. When you see the last post of Kiiyor about (C) trial mech, it's terrible, i don't understand why there are still put in regular match while they don't know how to drive a mech and won't understand why there are getting killed so easily.

That's why i asked several times to reintroduce ELO buckets : use the player ELO score to determine in which bucket he is and put him with and against players of the same bucket. While the ELO sum up of each team will not be as close as before, the match will be better balance because every player in both team will have roughly the same skill. The real problem is that PGI has to rewrite the MM nearly entirely for that.

@Sandpit : the fact that you recommand to cancel the MM search given a certain period of time, is a good indicator that the MM is not working correctly. And the supposedly low population is a false excuse : actually, even in a TBR and a Heavy queue greater than 50%, you still find a match under 3 minutes.

@Mallan : trying to give advice to your team is helpful (it's a polite way to say guiding the herd) but what can you expect from a player asking "how do i save the game ?", another player asking "why do i power off with a red screen ?", or a player bringing a Victor LRM-boat ?


#70 Kensaisama

    Member

  • PipPipPipPipPipPip
  • Elite Founder
  • Elite Founder
  • 430 posts
  • LocationRedford, Michigan

Posted 19 January 2015 - 05:04 AM

No one likes ELO. Unfortunately nothing can be done about it. Thank the people who whinned about the unfairness of teams in a team based game. Team play trumps solo play, always has, always will.

#71 Spr1ggan

    Member

  • PipPipPipPipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 1,162 posts
  • LocationUK

Posted 19 January 2015 - 05:14 AM

View PostRampancyTW, on 18 January 2015 - 11:18 PM, said:

Just wanted to hop in to say that you're a good dude and I'd gladly play with you again :)


Cheers dude. Been a while since I've seen a jenner rekin face like you were.

And check your forum inbox. :)

Edited by Spr1ggan, 19 January 2015 - 05:15 AM.


#72 mogs01gt

    Member

  • PipPipPipPipPipPipPipPipPip
  • Shredder
  • 4,292 posts
  • LocationOhio

Posted 19 January 2015 - 06:02 AM

View PostKensaisama, on 19 January 2015 - 05:04 AM, said:

No one likes ELO. Unfortunately nothing can be done about it. Thank the people who whinned about the unfairness of teams in a team based game. Team play trumps solo play, always has, always will.

Actually I like ELO. The ELO system in LoL works perfectly(if you start at the beginning of the season). MWO issue is population. The less population you have, the higher the difference in ELO between players in a match. So the concern is wait time vs ELO difference.

#73 Mawai

    Member

  • PipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 3,495 posts

Posted 19 January 2015 - 07:41 AM

View Postmogs01gt, on 17 January 2015 - 04:53 PM, said:

It means the game is forcing you do take a loss. That isnt how ELO works. ELO is suppose to team you with equal level teammates and opponents. This will cause you to gradually gain more ELO to a point where you simply start to be 1.0 w/l.

MM in MWO tries to equalize the matches by adding in playes way below your ELO


No. It is not forcing you to take a loss.

No. It does NOT intentionally add players way below your Elo. This only happens under two conditions:

1) A truly elite Elo player enter the queue ... they are chosen as the basis for forming the match since they have been in the queue longest. The MM grabs whoever it can nearest that Elo level ... which are pretty sparse. You end up with a high Elo player mixed with more average players.

2) The queue is starved for players so after some time it expands the search window to larger Elo bands.


In general, if you play the same as always, are grouped with folks that are more or less near your skill level, are facing players more or less near your skill level THEN you will only WIN about HALF the time. The MM could care LESS about your individual record ... it tries to set up the best balanced matches it can ... it is then up to you and your team to play them.

On average, over the LONG term, unless you are getting better ... your W/L will be near 1 if you are ranked correctly.

PERIOD. No conspiracy. No MM grouping you with noobs to give you a loss ... it could care LESS.


Finally, MWO is a team game ... and once you lose a few mechs the odds of you winning drop significantly. One isn't that much of a big deal ... but if your side loses the first one then your odds drop a bit. If you are down by 3 then your odds have dropped a lot unless your team coordinates. That team mate who peeks and gets cored by the 6 opponents on the other side of the hill probably uses that technique effectively in many matches ... but they are killed after doing 20 damage in this match ... did you get grouped with a derpy Elo player ... or a good Elo player that made a mistake?


My archived W/L is something like 1.4 while my current one is a bit less than 1.0.

Why is it less than 1.0? This is because I have been leveling up and playing new mechs. I am not as good in a new mech. I haven't unlocked all the skills in a new mech. These are not reflected in my Elo so I get matched a bit higher and on average my W/L and Elo drop a bit as I train up a new mech. As I get better with that mech, I will probably start to earn back some of the losses and increase the Elo a bit.

That is how it works. Sometimes it feels like the MM has a personal vendetta (lol) but nothing could be further from the truth.

Edited by Mawai, 19 January 2015 - 07:41 AM.


#74 Amsro

    Member

  • PipPipPipPipPipPipPipPipPip
  • Overlord
  • Overlord
  • 3,436 posts
  • LocationCharging my Gauss Rifle

Posted 19 January 2015 - 07:52 AM

View PostMauttyKoray, on 17 January 2015 - 04:58 PM, said:

No, the matchmaker is meant to equally face you against stronger and weaker opponents. Its actually working as intended. If you face stronger opponents and lose you go down in W/L, if you face weaker opponents and win, you go up. That is the direct translation of ELO. A higher W/L ratio means higher ELO, which means stronger opponents, and vice versa. Stop trying to act like you know what ELO is supposed to make matchmaker do.


HAHA so you say, but in fact no this is all false. Your w/l does not equal your Elo. Nice try tho.

#75 Egomane

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • 8,163 posts

Posted 19 January 2015 - 07:55 AM

View PostSgtKinCaiD, on 19 January 2015 - 05:03 AM, said:


Not going to be rude but this exactly what the MM does to handle high and low ELO players => http://mwomercs.com/...ost__p__2633230

And my response from another topic about the MM => http://mwomercs.com/...ost__p__3973139

Not going to be rude, but that post is from before the rewrite by Karl last year. The current matchmaker was released some time after this post:
http://mwomercs.com/...chmaker-update/

#76 Mawai

    Member

  • PipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 3,495 posts

Posted 19 January 2015 - 07:56 AM

View PostLordKnightFandragon, on 18 January 2015 - 06:09 PM, said:



that is why elo should take into account all the matters of a players performance.

Average damage per weight class, the higher the weight class, the more avg damage matters.
The player's average xp
Kills, deaths, assists, how often the player triggers those various role warfare rewards.

Just anything and everything pertaining to a player's performance over hte course of a match. it then takes all that and comes up with your elo rating, not just.....you have a 49% win rate, you get a rating of 100, then a the guy with a 59% gets a rating of 180 and go.

We should have a "Cadet Q", where for your first 100 games your elo is zero(0). During the first 100 games, it takes everything you do and at game 101 adds it up averages it and does the math to come up with your final elo. Beyond that, its up to you to improve it or degrade it.


It would be interesting to have a performance based rating system ... but what does it really mean?

I spectated a Timberwolf in one match. His final score was something like 4 kills, 2 assists, 1100 damage ... game lost.

Sounds like he did a great job and was let down by the team?

However, I was spectating him ... he was sniping and hiding (in a Timberwolf) ... it came down to 8:1 ... the game was over ... our side had lost ... the 8 opponents were mostly badly damaged. I watched as this Timberwolf racked up tons of damage (his aim was pretty poor) while managing to kill 4 of the opponents.

His numbers were meaningless ... he did not contribute to winning the game in any way, shape or form ... and yet on a "performance" based ranking system he would have scored highly with lots of flank shots, hit and run, as well as kills, assists and total damage ... but there is no way I would want him on my team.

So ... this is the crux of the problem ...

- do folks who aim well, do 300 damage, work well on a team, get the occasional kill, but ENABLE the team to win ...

are they ranked lower than the folks who

- kill steal to pad their K/D, spray damage all over the target (1100 damage should be enough to kill 6 assaults if the person was aiming well ... think about it), don't necessarily work well with others ...


Overall, I am not convinced that an alternative performance-based ranking system to simply wins/losses as a team will actually produce any better results in terms of matchmaking. It would be interesting to perhaps develop a composite system that combines the two ... but the bottom line (I think) is that you want to group folks together who have similar winning contributions rather than just similar damage ouput.

#77 Appogee

    Member

  • PipPipPipPipPipPipPipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 10,966 posts
  • LocationOn planet Tukayyid, celebrating victory

Posted 19 January 2015 - 07:58 AM

View PostMauttyKoray, on 17 January 2015 - 04:58 PM, said:

If you face stronger opponents and lose you go down in W/L, if you face weaker opponents and win, you go up. That is the direct translation of ELO. A higher W/L ratio means higher ELO, which means stronger opponents, and vice versa. Stop trying to act like you know what ELO is supposed to make matchmaker do.
That's not actually how it works. I'm going to assume someone in the following four pages explained it for your benefit,

View PostMauttyKoray, on 17 January 2015 - 04:58 PM, said:

Stop trying to act like you know what ELO is supposed to make matchmaker do.
Quite an ironic statement, really.

#78 LordKnightFandragon

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • 7,239 posts

Posted 19 January 2015 - 08:07 AM

View PostMawai, on 19 January 2015 - 07:56 AM, said:


It would be interesting to have a performance based rating system ... but what does it really mean?

I spectated a Timberwolf in one match. His final score was something like 4 kills, 2 assists, 1100 damage ... game lost.

Sounds like he did a great job and was let down by the team?

However, I was spectating him ... he was sniping and hiding (in a Timberwolf) ... it came down to 8:1 ... the game was over ... our side had lost ... the 8 opponents were mostly badly damaged. I watched as this Timberwolf racked up tons of damage (his aim was pretty poor) while managing to kill 4 of the opponents.

His numbers were meaningless ... he did not contribute to winning the game in any way, shape or form ... and yet on a "performance" based ranking system he would have scored highly with lots of flank shots, hit and run, as well as kills, assists and total damage ... but there is no way I would want him on my team.

So ... this is the crux of the problem ...

- do folks who aim well, do 300 damage, work well on a team, get the occasional kill, but ENABLE the team to win ...

are they ranked lower than the folks who

- kill steal to pad their K/D, spray damage all over the target (1100 damage should be enough to kill 6 assaults if the person was aiming well ... think about it), don't necessarily work well with others ...


Overall, I am not convinced that an alternative performance-based ranking system to simply wins/losses as a team will actually produce any better results in terms of matchmaking. It would be interesting to perhaps develop a composite system that combines the two ... but the bottom line (I think) is that you want to group folks together who have similar winning contributions rather than just similar damage ouput.


I get the whole padding thing. Its basically what people do in WoT to become "blue" or "unicum", which is a really good player. They L2P on one acct, then few thousand games later, after they understand the game, they reroll and play only the good tanks, pad the damage and stuff up. I get there is padding and I get that you can make yourself look better than you are.

If we got a Performance based rating type, weigh winning and losing absolutely the heaviest. In WoT, the WN 8 rating's highest weight is in damage dealt. So you simply get a high damage tank and fire alot and do alot of damage.....

I ran across a set of "blues" in that game not to long ago, they were absolutely not blue at all. They couldnt destroy my T29, kept bouncing off my turret....and if you dont know Wot, that is the absolute best turret armor in the game, the American Heavy line, T29, T32 and the like, to shoot it there is to do absolutely nothing 95% of the time.

So, while there are loopholes, what we have now is just as derpy, it simply uses W/L record of the player. That leads to some players who have better lcuk then others getting on the winning team getting higher rating then their skill lvl. So, ultimately, they get on teams, suck royally and are actually a hindrance to the team. Where as, while it has flaws that can be milked, an Overall Performance rating system would be alot better. It would give you a more true depiction of the player's skills.

While in your example, the player was being a padding fool, and not helping much, he obviously could aim well enough to do the damage he did. While if he was a "red" player as per woT, he would have either been AFK the whole time, been so bad he would not have dealt any damage at all, or died instantly, being totally clueless and having no tactical sense about him at all. He knew to pull back, maybe drag the enemy team around, and bait them, and in the end, maybe it was a loss, but the game ended up alot closer then it otherwise would have. And also, with a performance based rating system, it should ultimately stick like skill players with and against other like skill players, not try to "balance" the ratings, which in the end, just hurts the good players, as since they weigh more in the MM they get stuck with more bads more often, which drags their own scores down and just pisses them off.

If it sticks truely good players with other truely good players, and the really bad with the really bad, you will get similar performance out of all the players, games would end up being closer overall.

#79 mogs01gt

    Member

  • PipPipPipPipPipPipPipPipPip
  • Shredder
  • 4,292 posts
  • LocationOhio

Posted 19 January 2015 - 08:08 AM

View PostEgomane, on 19 January 2015 - 07:55 AM, said:

Not going to be rude, but that post is from before the rewrite by Karl last year. The current matchmaker was released some time after this post:
http://mwomercs.com/...chmaker-update/

That link says nothing about MM and ELO. It's referencing 4x3 MM.

Edited by mogs01gt, 19 January 2015 - 08:09 AM.


#80 anonymous161

    Member

  • PipPipPipPipPipPipPipPip
  • FP Veteran - Beta 2
  • FP Veteran - Beta 2
  • 1,267 posts
  • LocationIowa

Posted 19 January 2015 - 08:23 AM

In conclusion the game is broken.





1 user(s) are reading this topic

0 members, 1 guests, 0 anonymous users