23 replies to this topic

[Champion of Khorne Lord of Blood]

I know it won't happen because this is just another scream into the echo chamber of ideas, but what if on each team the top 4 players go up and the bottom 4 go down and the middle 4 stay the same win or lose?

It stops the worst guys on the team from being carried in PSR, lets people who are just doing average settle into a tier, and lets the people consistently top scoring move up. Its also pretty basic and rewards the best players even on a bad loss so you won't lose with over 500 damage done and a couple kills and only manage an equal because you couldn't carry the whole team on your shoulders.

Thoughts?

[C E Dwyer]

just actually altering the upward bias will have the same effect, as there would be more, equal and minus results.

Also what happens if they all have comparebly good games

[Davegt27]

seems like a good idea to me

[Davers]

Why don't we just cut out the middleman and replace PSR with damage, since that's the only stat players seem to be interested in?

[Jman5]

I wouldn't be against a complete zero sum ranking system. However I do think we need to address the matchscore weight bias first. Whether that means adding new stats, adjusting existing ones, or just adding a chassis score modifier, I think something needs to be done here.

[Mawai]

To do PSR correctly:

1) It needs to be conservative ... meaning that skill points are neither created nor destroyed in any match.
2) W/L still needs to play a role since damage isn't the only or even most important factor.

For example ... let's come up with a number for the total PSR +/- available in any match.

e.g. 100 points of PSR are available from any given match.

Give a bonus to match score to each player on the winning team (value to be determined but if you don't want to weight winning you can set this to zero).

Sum all the match scores for ALL players in the match. This method relies on Match score being some sort of meaningful representation of the player ability (including winning) which may not be the case.

Take the average match score - this is the zero point.

Assign the points earned or lost as:

Points_change = 100 * ( MS - Average)/ (sum_of_match_scores)

The farther above or below average you are in a match the more you make or lose.

If you sum the points change over N players you get ...

MS ... N players = sum_of_match_scores
N players * Average = sum_of_match_scores
Expanding the above you get the total points change for all players = 0.0 ... conservative ... so that what one person loses another person gains.

The total points available to be won or lost in a match should make a small but measurable change in rating.

Some checks and balances may be required
- avoid situations where folks serially disconnect or suicide to tank their ratings ... need something to prevent dropping rating being too easy.
- Need to also add checks for staged matches in which players on both sides collude to give rating to selected players.

Anyway, this sort of system allows folks to go up win or lose depending on how much weighting is given to winning the match and the average performance of all players in the match. Of course, if everyone does well then their ratings don't change much since they were in an even match up and thus should all be ranked comparably.

[Tristan Winter]

This would be a terrible idea, because match score doesn't reflect how good players are. Consider how easy it is to get a 400+ match score in an AWS-8R versus a Victor, for example. A guy trying to master Vindicators would end up in Tier 5 within a week. Furthermore, match scores are heavily biased towards heavier mechs. All the Tier 1 players would be assault jocks. And matchmaker would struggle to find enough Tier 1 light mech players for them anyway.

WLR is the most consistent metric of how good a player is. You can artificially increase your damage (and hence, match score) by playing certain mechs with certain builds (e.g. LRM boats) or playing certain styles (e.g. using your teammates as cannon fodder in order to inflate your personal score), but there's no way to artificially increase your WLR. In order to have a high WLR, you just need to be a good player.

In group queue, you can get carried, of course. But being carried will inflate your match score anyway, so the argument is moot.

[Mawai]

Tristan Winter, on 07 November 2016 - 02:25 PM, said:

This would be a terrible idea, because match score doesn't reflect how good players are. Consider how easy it is to get a 400+ match score in an AWS-8R versus a Victor, for example. A guy trying to master Vindicators would end up in Tier 5 within a week. Furthermore, match scores are heavily biased towards heavier mechs. All the Tier 1 players would be assault jocks. And matchmaker would struggle to find enough Tier 1 light mech players for them anyway.

WLR is the most consistent metric of how good a player is. You can artificially increase your damage (and hence, match score) by playing certain mechs with certain builds (e.g. LRM boats) or playing certain styles (e.g. using your teammates as cannon fodder in order to inflate your personal score), but there's no way to artificially increase your WLR. In order to have a high WLR, you just need to be a good player.

In group queue, you can get carried, of course. But being carried will inflate your match score anyway, so the argument is moot.

To be honest, I agree with you. Match score is not a good indicator of player skill. However, Match score and PSR were specifically introduced because the player base complained that win/loss data as utilized by an Elo based ranking system was also not a good indicator of player skill.

I think some of the flaws with the mostly damage based PSR system and LRMs could be remedied if LRM damage was discounted in the match score calculation so that doing a lot of damage with LRMs would be equivalent to doing less focused damage.

Anyway, I think there is little doubt that Match score could use some tuning. However, match score is the basis of the current PSR system which makes the current system just as flawed.

[MortZA]

Weight the match score's effect on PSR against ave. chassis and tier match scores (which means it can change over time and have a volatile period for new mechs).

[El Bandito]

Why do people try to make things more complicated? Keep it simple by just increasing match loss penalty for PSR, and halve the damage contribution to match score yet again, to 25%, while increasing match score for other tasks, especially successful caps. Also, separate solo-q and group-q PSR.

Over time, weaklings will be weeded out to lower tiers. PGI can do a soft PSR reset at start (let everyone fall two tiers) to make the process faster.

Edited by El Bandito, 07 November 2016 - 10:32 PM.

[Duke Nedo]

El Bandito, on 07 November 2016 - 10:21 PM, said:

Also, separate solo-q and group-q PSR.

This! Until this is done, PSR or stats have little meaning.

That said I don't think tweaking PSR or the MM will have much to say at this point, there are just too few players for good match making. Dropping solo now is always a mix of many tiers, if I understand the system correctly, a match seeded by a Tier 3 player can include all 5 Tiers? More often than not it feels like that anyways. In a case like that, who cares if I am maxed tier 1 or if I am middle tier 2? I'd still be in the same match against the same people... so as long as the MM can't afford tighter release valves... does it matter?

[El Bandito]

Duke Nedo, on 08 November 2016 - 12:41 AM, said:

This! Until this is done, PSR or stats have little meaning.

That said I don't think tweaking PSR or the MM will have much to say at this point, there are just too few players for good match making. Dropping solo now is always a mix of many tiers, if I understand the system correctly, a match seeded by a Tier 3 player can include all 5 Tiers? More often than not it feels like that anyways. In a case like that, who cares if I am maxed tier 1 or if I am middle tier 2? I'd still be in the same match against the same people... so as long as the MM can't afford tighter release valves... does it matter?

Sure, the release valve will eventually have to mix everyone if player count is very low, but at least this time tiers actually matter more in terms of skill-gauging.

[Sjorpha]

You only need to remove the bias towards increase.

Possibly have separate PSR for solo and group, but I'm not convinced that's necessary.

[Duke Nedo]

Sjorpha, on 08 November 2016 - 01:11 AM, said:

You only need to remove the bias towards increase.

Possibly have separate PSR for solo and group, but I'm not convinced that's necessary.

I would be careful if it was up to me though. If even more focus is put on personal score in order to build your Tier I would expect even worse team play (especially in the solo queue). IMO it's very important that winning the match is heavily weighted, and I think that someone doing something unselfish early in the game to help the team should not be punished. So how to separate that from being a potato and dying quickly because you did something stupid? Very difficult, so I'd not bother. The solo queue really needs people that dare to lead a push. We don't need more incentive to use your teammates as meat shields and try to farm high scores for yourself. There is more than enough of that already, I find myself more and more frustrated with the lack of team play when trying to drop solo, it's just no fun anymore.

It happens every evening that I forget I am in the solo queue (and bad tier one at that), end up too much in the front line, then the team backs away at first contact to save their own skins instead of taking advantage, and I die alone like a proper potato with 200 damage, sacrificed by my team mates. We then lose of course. So, next game what to do? Either not play (wait for unit members to log on) or do the same and thereby making it over all worse.

I am more for the old Elo to be honest, with group/solo queue separation. Or just keeping it as it is. It's good enough to separate Tier 3 from Tier 1, I am not sure we need more when the MM is anyways mixing Tiers, and the MM cannot even afford to take things like tonnage into consideration. I am more concerned about the game play that results from it, than the matchmaking itself.

[SnafuSnafu]

IMO, leave PSR the way it is, there's intrinsically nothing wrong with it, as it is simply a MM based on time invested into the game; a la the longer you play the more 'experienced' you are in how the game works, and thus get matched with other (roughly) similarly experienced ('EXPERIENCED'not 'SKILLED') players.

Tinkering with it further IMO is a very bad idea that will lead to nowhere good, in this case, selfless plays (plays that nets you no damage nor kills, a la a charging tanking Atlas [hopefully charging ahead to take heat off his advancing team] too busy twisting to get more than 2-3 salvos off) will be less of a factor.

My philosophy as to what are and aren't the MM's responsibility:

It is the MM's responsibility that new and less experienced* players do not get matched with seasoned players. And it is the MM responsibility that both teams are filled with equal amounts of players.

That's it.

On the other hand..

It's nobody's responsibility but the player's whether he or she wins all their games, or equally, looses all of them. It's nobody's responsibility but the player's that he or she may not be skilled enough to be a factor in being able to win said games. And it's nobody's responsibility, but the player's that he or she may perhaps be exceptionally unskilled enough to be a major detriment to his or her own team.

Equality in reality does not exist, not in life and especially not in this pvp game so don't try to achieve it, because it does not and will never exist. Trying to make it exist will simply be a waste of time and will only lead to more broken things to complain about.

ALL, WITH NO EXCEPTIONS, even the most 'dreamiest' and most 'perfect' MM solutions people may have are be BAD.

Edited by SnafuSnafu, 08 November 2016 - 03:11 AM.

[Willard Phule]

It's a great idea, if your intention is to separate players based on skill.

Keep in mind that separation is the farthest thing from PGI's mind, and always has been. They know this game has a brutal learning curve. So brutal, in fact, that player retention has always been an issue. Any form of matchmaking from PGI includes dropping experienced players with new ones....either to carry them or hoping that they'll learn something from watching.

The better you are, the more potatoes you carry.

[Mawai]

Duke Nedo, on 08 November 2016 - 12:41 AM, said:

This! Until this is done, PSR or stats have little meaning.

That said I don't think tweaking PSR or the MM will have much to say at this point, there are just too few players for good match making. Dropping solo now is always a mix of many tiers, if I understand the system correctly, a match seeded by a Tier 3 player can include all 5 Tiers? More often than not it feels like that anyways. In a case like that, who cares if I am maxed tier 1 or if I am middle tier 2? I'd still be in the same match against the same people... so as long as the MM can't afford tighter release valves... does it matter?

It is not supposed to work that way (tier 3 used as seed allowing grouping of all tiers). They have stated that tier 1 should never end up grouped with anyone below tier 3 for example. So the seed should not matter for that constraint to hold.

On the other hand, PGI hasn't published any remotely recent data on the MM performance and none since PSR was introduced so we really don't know how well it is working. The MM could be working perfectly and PSR isn't or you could be faced with a smaller player base resulting in challenging MM. At least at prime time, I don't think the player population is the issue but again without actual numbers from PGI it is impossible to say.

Finally, many folks complain about people .using LRMs having limited skill. However, LRMs are one of the easiest ways to advance your PSR and tier standing. There are a lot of tier 1 players who have and do use LRMs extensively, who sit in the back or use their team mates as cover. It can be very effective at both getting high match scores and progressing in tiers. In the end, though, I think it is the mix of strategies and time invested that all result in "tier 1" which cause problems for the matchmaker. The solo ECM sniper that doesn't play well with others, the LRM boats that survive to the end putting up huge damage numbers and match score, and the folks with good aim ... all advance and get grouped together by the matchmaker since they have the same PSR arrived at using different strategies.

However, none of these require good team work and team work is probably the single most important factor in winning or playing effectively. As a result, when you throw all of these folks who may or may not jell as a team together ... you can end up with a lot of matches that just feel like folks don't know what they are doing despite having comparable PSR scores.

[Xmith]

Remember this a team based game period. Just like all sports, the bench warmers still get trophies too if the team wins. The players that performs well will receive earnings based on performance. The players that do not perform well will get smaller earnings. But all team members should benefit from a win no matter how they actually performed.

Edited by Xmith, 08 November 2016 - 03:50 PM.

[RestosIII]

I just know that with my horrible skills, I'm still rising in PSR rating. That shouldn't be happening. When it isn't "rare" for me to have a 150 damage game because I'm bad at positioning and get shredded, I shouldn't going up and up towards T1. I'm running all lore builds for Kerensky's sake, I should be plummeting right now, not rising to the point where the only people I meet are running meta builds and have the twitch reflexes of a CS:GO competitive player on speed.

[Tarogato]

+1


This prevents people who score low, but "just enough" from keep climbing through the tiers, and rewards people who perform top of their team consistently.