63 replies to this topic

[The6thMessenger]

Maybe this has been discussed, but what if there's no "=" in tier? Basically just Up if you win, Down if you lose?

Then all you need is to have a positive WLR, and it will eventually get you up.

Tiers would require a different amount of maintaining WLR -- i.e. T5 = 0.6, T4 = 0.80, T3 = 1.0, T2 = 1.20, T1 = 1.4.

That means, you only need to win 50% of your matches to attain/retain T3. You need to win 70% of your matches to attain/retain T1, and you only need to win 40% of your matches to attain/retain T4.

Every 10 matches is measured, and adds or takes away from the Tier Bar.

Edited by The6thMessenger, 08 April 2018 - 01:23 AM.

[El Bandito]

We talked about zero sum PSR system for two years. PGI aint doing anything about it.

[Zergling]

It wouldn't be perfect, but it sure as hell would be better than what we have now.

[MrMadguy]

Again. You don't understand, how MM should work. First thing. What you talk about - is pure ELO. And WLR isn't good indicator of skill in team-based game, as team performance =/= personal performance, especially if teams are random. ELO works perfectly for 1vs1 or premade vs premade situation only. Second thing. Higher WLR/AvgMS for higher tiers - is wrong concept. It comes from very common misconception, that higher skill = higher stats. And you should understand one simple thing: player A can have high stats only at the expense of some other player B, having low stats. I.e. if you have KDR = 2, then some other player have KDR = 0.5. As simple, as that. And this simply means, that player B is weaker, than player A and therefore they shouldn't be matched together. MM's main goal - to put all players into equal conditions. That means, that only thing, that should be different in different Tiers - is skill. Nothing more.

PSR is actually almost perfect. Things, that should be fixed: 1) Bias towards increasing - rating should drop as fast, as it rises. 2) Too large weight class/'Mech/build<=>performance dependence (i.e. some per 'Mech modifier should be implemented).

Edited by MrMadguy, 08 April 2018 - 02:07 AM.

[Zergling]

MrMadguy, on 08 April 2018 - 01:45 AM, said:

Again. You don't understand, how MM should work. First thing. What you talk about - is pure ELO. And WLR isn't good indicator of skill in team-based game, as team performance =/= personal performance, especially if teams are random.

Actually, if teams were random W/L would be the best indicator of skill, because then everyone would have the same degree of team quality over a large number of battles played.

The primary reason W/L isn't an ideal indicator right now is because leaderboards mix solo and group queue together.
The secondary reason is the PSR Tier system, which inflates the W/L of lower tier players and deflates the W/L of higher tier players.

Edited by Zergling, 08 April 2018 - 02:17 AM.

[eminus]

Dont fix what is currently broken

[MrMadguy]

Zergling, on 08 April 2018 - 02:15 AM, said:

Actually, if teams were random W/L would be the best indicator of skill, because then everyone would have the same degree of team quality over a large number of battles played.

The primary reason W/L isn't an ideal indicator right now is because leaderboards mix solo and group queue together.
The secondary reason is the PSR Tier system, which inflates the W/L of lower tier players and deflates the W/L of higher tier players.

My stats - are best proof, that WLR is completely meaningless for pure QP player.

[Nema Nabojiv]

What the point of fixing it if there is only enough players for 1-3 and 4-5 tier buckets? To feel truly leet with your tier one when you still matched with lower tier players?

[Spare Parts Bin]

The Tier at system penalizes buying new Mechs/ Omnimechs and trying new builds and using lighter or heavier machines.

To expect me to get as many kills or do as much damage in my Kit Fox Purifier as I do in a Ebon Jaguar or Supernova is counterintuitive.

[Zergling]

MrMadguy, on 08 April 2018 - 02:28 AM, said:

My stats - are best proof, that WLR is completely meaningless for pure QP player.

My stats - best proof that WLR is entirely meaningful for a pure QP player.

If you can't win more battles than you lose in solo QP, then I'm sorry, but you just aren't playing well enough.

Edited by Zergling, 08 April 2018 - 03:19 AM.

[Nema Nabojiv]

Spare Parts Bin, on 08 April 2018 - 03:08 AM, said:

The Tier at system penalizes buying new Mechs/ Omnimechs and trying new builds and using lighter or heavier machines.

Dont see any connection here. How exactly are you penaltized?

New mechs may hurt your statwhоring, but that is only if you care about stats, which is your choice.

[Bombast]

MrMadguy, on 08 April 2018 - 02:28 AM, said:

My stats - are best proof, that WLR is completely meaningless for pure QP player.

Your stats are a perfect example of why WLR means a ton when it comes to comparing players.

As an aside, check your Match Score per Season to your WLR per Season. You'll see that the more you did (That the game tracks), generally the better your WLR. Kinda seems to indicate the ability to do work in game increases WLR, doesn't it?

Edited by Bombast, 08 April 2018 - 03:58 AM.

[Alexander of Macedon]

MrMadguy, on 08 April 2018 - 02:28 AM, said:

My stats - are best proof, that WLR is completely meaningless for pure QP player.

Your stats are excellent proof that WLR is a good metric of player performance.

Also, note what Bombast said. Your higher W/L seasons tend to map pretty well onto seasons where you had higher matchscore. Even though it's 12v12, every single iota of effort from every single player helps improve the team's chances of winning. If you regularly underperform, you're one of the ones making the team a "bad" team.

[MrMadguy]

No. There is simple explanation for my "improvements" - holiday events.

What I see - is that there no correlation between WLR and personal performance. When I have WLR = 1.0, I can also have KDR and AvgMS values, like 0.48 + 134, 1.03 + 231 and even 1.39 + 250. First below normal, second around normal and third - above normal. Or, surprise, I can have higher WLR, but lower personal stats. What I see very often - is that I die very early with almost 0 contribution towards victory, but my WLR still = ~1. Yeah, I know, that it's against COC, but we could even make an experiment - to simply AFK in all matches. Guess what? WLR still would be around 1. This phenomena is called "carrying". And this phenomena is sole reason, why WLR can't be used as skill measure.

As you can see, correlation is way too weak:

But please remember, that MM itself tries to keep players at WLR = 1 and AvgMS = 200 level. Without it's effort there wouldn't be any correlation at all.

Edited by MrMadguy, 08 April 2018 - 05:19 AM.

[Jman5]

The thing that broke Elo was the group queue. I wouldn't mind a return to Elo if it only measured your solo queue games and did not carry over to when you played group queue.

Group Queue's a bit of a mess since your premade's strength is so variable from day to day. It probably does work better with a matchscore based ranking. (albeit an improved one)

[Mystere]

El Bandito, on 08 April 2018 - 12:40 AM, said:

We talked about zero sum PSR system for two years. PGI aint doing anything about it.

We had Elo, but we have what we have now because people cried about the former.

[El Bandito]

Mystere, on 08 April 2018 - 07:00 AM, said:

We had Elo, but we have what we have now because people cried about the former.

Which could mean that the former had serious faults as well.

[Mystere]

MrMadguy, on 08 April 2018 - 01:45 AM, said:

- is pure ELO.

I really can't take you seriously on this topic when you can't even type "Elo" right. The system is named after a person and is not an acronym.

El Bandito, on 08 April 2018 - 07:01 AM, said:

Which could mean that the former had serious faults as well.

Or people just really wanted an experience bar.

On a (slightly) more serious note ...

And what would those so-called "serious faults" be that the current system "fixed"?

Other than the people above, does anyone think the current system is even remotely better than Elo?

Edited by Mystere, 08 April 2018 - 07:14 AM.

[Bombast]

MrMadguy, on 08 April 2018 - 04:55 AM, said:

But please remember, that MM itself tries to keep players at WLR = 1 and AvgMS = 200 level. Without it's effort there wouldn't be any correlation at all.

You keep saying this.

That. Is. Not. What. MWO. MM. Does.

At no point has PGI ever said that's what (Current) MM has done. At no point has any intensive examination of their matches indicated this. At no point has a stat check of the population indicated that this happens. Why do you keep saying thats what the Matchmaker does?

MrMadguy, on 08 April 2018 - 04:55 AM, said:

This phenomena is called "carrying". And this phenomena is sole reason, why WLR can't be used as skill measure.

You cannot in the same breath say personal performance doesn't matter, and then invoke the power of being carried as proof of that. The very concept of carrying relies on personal performance mattering.

Edited by Bombast, 08 April 2018 - 07:07 AM.

[yrrot]

Group queue + Elo (+class balancing) was just slow to matchmake. Like, take the worst matchmaking time you get now in group queue and set that as the normal.

People get kind of hung up on WLR. If you are only matched in perfectly balanced games, your WLR would be 1. You could be just awful at the game and have a WLR of 1 just by being matched against other awful players. High WLR ratio is just as easily an indicator that you were matched against too easy of an opponent.

In an Elo system, your change in score on a win or loss is dictated by the Elo difference between the two players. If you were matched against several mediocre players in a row as a pro, for example, you could have a crazy high win/loss, but not go up at all in rank--or even go down in Elo if you lost just one game to an inferior opponent.

Elo is also designed for 1v1 (originates from chess). Its ability to form fair matches for groups is iffy, depending on implementation. Even assigning a score to a playing in a group game is a problem if you don't account for their contribution to the win/loss.

Part of me wonders if it would be better to try to balance teams based on the players Solaris 1v1 Elo for the division of mech they are in for QP, if PGI wanted to go to an Elo-based system there again.