6 replies to this topic

[Agent of Change]

So as most of us agree the PSR system is not a good representation of player skill, which does not make for good match making.

It is in fact a glorified XP bar that basically only takes into account raw wins and it is possible to have a sub 1:1 WLR and advance to T1 through just playing a lot.

First off, we should also accept that in a game where the vast majority of matches are not played with the same 11 other players the wins and losses a player has matter very little in judging their personal skill and ability to influence a match. Even the best player is more or less done 9 times out of 10 when ten mechs on their team spontaneously explode.

So I've been discussing this and thinking about it at length with my unit and I'm gonna propose something. I don't expect everyone to like it, or agree with it or for it to be implemented, hell I wouldn't even be surprised if this has been suggested before but I'll go ahead and share so that i can say i tried.

Link a Players PSR to their average match score. That's it. No more progress bar, no more slow crawl towards up tier. Just average their match score and assign T1 -5 based on that, then only match +/- 1 tier because if you assign PSR based on average match score and you look at the distribution curve of those match scores you can ensure each tier is well populated.

I got this idea from using Jarl's list among other things as a reference but they have a shart that caught my eye:



That is a near ideal distribution skill curve. If PGI were to say re do the Tiers as something like: (Purely an example not meant to be final numbers)

• T1 = Average Match Rating 275+
  
• T2 = Average match Rating 250-275
  
• T3 = Average match Rating 215-250
  
• T4 = Average match Rating 190-215
  
• T5 = Average match Rating <190

You would have better distributed tiers and better matchmaking for skill if the population is distributed within these brackets.




In addition we known PGI tracks this particular stat (you can see it in teh leader boards) and they wouldn't even really need to change the UI... the current bar could basically just display where in the current bracket their average puts them. The End of match UI can still have teh UP, Down and Equals symbol for PSR because it's just comparing Match rating Vs. your current average.

This would have a lot of benefits for players not the least of which would be making skill level match-ups much better, but also giving a more accurate measure of where you stand in comparison to other players. I can't guarantee better match quality but you'd certainly have less opportunity to call Spud on a T1 player as you would be less likely to stumble into high average match score over time.


But what about new players you ask? Well you would have new players play 50 or so games as an assumed PSR 1 then average their game scores and place them in the tier they have earned.

What about progression from there? You adjust a players average every so many games until they hit a statistically significant number of games say every 10 games until they reach 100 then from there on it's a live average.

What if i get better... won't old games hold me back? well they might except there is already existing systems that can handle this. At the beginning of a new season your PSR is determined by the existing average you have at the end of the last season/ last time you played, then you treat it just like a new players progression from there. First 50 games sets your new Average (including the old average as a less weighted factor) then proceed from there. So even if you seem stuck in a lower tier than you deserve it would only really last until the beginning of the next season.

Are there ways that this can be improved? Sure off the top of my head I'd weight performance of the different weight classes differently much like Jarl's List does becasue judging a Locust and an Atlas by the same standard seems crazy and a little unfair.


I could go on but this is already a wall of text. the point is this COULD be implemented using almost entirely existing structures and UI with minor changes. The biggest issue would be the math in the back ground for determining averages but after you've determined averages and assigned PSR the MM structure wouldn't have to change and might be the better for it.

Please discuss.

*Edited for spelling/grammar/formatting*

Edited by Agent of Change, 20 June 2018 - 12:08 PM.

[Doomich]

Strong solution and easy to do
But we still have to play with the values

[Agent of Change]

I totally agree and it was already a wall'o'text, I have more suggestions and refinement ideas, but that's really getting into the weeds and really we'd need PGI to break down the population per "average" and create even PSR ratings by population.

I also see benefits like being less able to game this system very well because to intentionally tank your rating you'd have to not just lose but basically play very badly for a long time and then you'd be stuck low for just as long but you couldn't stay low and abuse the system because you'd rise quickly as you did very well.

Smurf accounts would rapidly get placed at the level of the player skill and it they are abusing it they'd basically end up back in higher tiers as soon as they hit the first threshold.

anyway I've been thinking about it for a bit, thanks for commenting.

Edited by Agent of Change, 20 June 2018 - 07:08 PM.

[Mister Bob Dobalina]

If I might play the devils advocat for a moment:

Since the match rating/score is more biased to dmg, kills, etc than to mission goals (capture etc.) it is to be excepted that the proposed system will greatly benefit heavy hitters and damage spewers against lights and fast mediums. I am suspecting there could be a shift towards using more and more Mechs with high Match Score potential.

All in all I am myself strongly for an overwork of the Tier system. I am a T2 on the brink to T1 and I simply do not belong up there. The gap of only +/- 1T might be to close for the MM to work properly though as the population isn't really that big to choose from. And these are distributed over 3 main server regions where 1 of which will be rather unplayable by ping to begin with.

[Agent of Change]

Hurbie, on 20 June 2018 - 11:51 PM, said:

Since the match rating/score is more biased to dmg, kills, etc than to mission goals (capture etc.) it is to be excepted that the proposed system will greatly benefit heavy hitters and damage spewers against lights and fast mediums. I am suspecting there could be a shift towards using more and more Mechs with high Match Score potential.

I completely agree which is why I mentioned above there should be weighting to the match scores in various weight classes to make it more fair.

Agent of Change, on 20 June 2018 - 11:57 AM, said:

Are there ways that this can be improved? Sure off the top of my head I'd weight performance of the different weight classes differently much like Jarl's List does becasue judging a Locust and an Atlas by the same standard seems crazy and a little unfair.

If you are unfamiliar with Jarl's list here is a link: https://leaderboard.isengrim.org/

Once there if you scroll to the bottom of the page in teh blue bar they explain transparently how their system works and how they weight scores. Now I don't think it's a perfect system, but it is well executed and at least worth looking at as a starting point for this concept. It goes a long way to evening out the differences to show that pilots that do well in lights (for being a light) routinely are rated fairly compared to those who do well in Assaults (for assaults)

Hurbie, on 20 June 2018 - 11:51 PM, said:

All in all I am myself strongly for an overwork of the Tier system. I am a T2 on the brink to T1 and I simply do not belong up there. The gap of only +/- 1T might be to close for the MM to work properly though as the population isn't really that big to choose from. And these are distributed over 3 main server regions where 1 of which will be rather unplayable by ping to begin with.

I also don't disagree here either. It's entirely possible the population isn't there. My initial suggestion is predicated on the reasonable assumption that while score distribution looks like a nice bell curve that PSR distribution Doesn't, that is is probably top heavy in regard to regular players because theoretically even an account that literally just hit launch and did nothing at all game after game would eventually end up in T1 because of aggregate wins. So my thought is that with players more evenly distributed among the PSR's through analysis of the average score populations you could tighten up the bands a little.

But again that's all essential nitty gritty implementation adjustments that don't exactly invalidate the idea itself, and while I know there are definitely better ways to do MM they also require much more effort and redesign. The Theory here is to give a suggestion that could be implemented with relatively little effort on PGI's part by comparison to more complicated solutions in order to provide an improvement to the Matchmaking in both skill matching and wait times by adjusting population distribution.

[Callsigntal0n]

I have the easiest solution of all. GET RID OF PSR.

Before i get too nailed for saying that lemme explain...

1:population- we simply dont have the population to run a tier system in Quick Play. I am T1 and out of maybe 10 games a day i run into the same pilots 7 out of those 10. (Not that i dont like playing against streamers) i just know my skill level is Not on par for the tier the only reason im t1 is because ive been playing for a good year year and a half. And honestly for quick play it should be just that. Throw everyone into the same pool. Think about it. No more "wait warrior online" yes the new players wont be able to hold their own vs the likes of proton or the myriad of streamers but at the same time the mix of tiers would actually help new players in the end by learning the calls, staying with the team etc

2:wait times- i have a t5 smurf account for times when my main acct is getting 2 to 3 minute wait times. Its a common complaint of the higher tiers. Personally my belief is if you notice you are waiting then you have been waiting too long. Example yesterday i wanted to take one of my mediums out for a spin the 2 matches i played both had wait tines of about 3 to 4 minutes. Which in my opnion is unacceptable for a "quick" queue. It actually pissed me off enough i actually went outside to do things (gasp! Heresy!)

But thats just my two cents and i can already feel it getting picked apart at the seams but i appreciate good debate so lemme know what you guys think.

[Agent of Change]

Callsigntal0n, on 21 June 2018 - 04:56 AM, said:

I have the easiest solution of all. GET RID OF PSR.

*snip*

But thats just my two cents and i can already feel it getting picked apart at the seams but i appreciate good debate so lemme know what you guys think.

I don't disagree at all. You make the same points that many others have and that i agree with. But quite frankly I don't think PGI is going to ditch the PSR system, so rather than wishlist a complex system there is no chance of use I figured giving them a potential way to make the system work better is a realistic approach... maybe. Y'know don't let the Great get in the way of the Good.

Theoretically, the problem with wait times is one of two problems (or a combination):

1. Low population - which means we are gonna have issues no matter what and if it's critically low whatever we do and any kind of MM is going to fail. The only way around this is getting more players or condensing the buckets.

Regarding Population it should be noted that per the steam charts while we have hit an all time low daily high of 1521 players that is still only about 300 players lower than teh 1800 some players that have been the average since May of 2016. It has also gotten sub 1600 in the past too so honestly the MWO player base has actually been fairly static on average for the last 2 years. Aside from "peak player" numbers since May 2016 the monthly average online players has been stable at about 1050 players +/- 150 players. So splitting the player base among game modes and queues may be the largest problem here. https://steamcharts.com/app/342200#All


2. Poor distribution of players in the tiers - I think this is definitely a problem as I've described. And exacerbated by having way too many buckets to split players into.

I'm not trying to fix everything, just looking at a better use of the existing system. That plus bringing in more players would likely improve things quite a bit.

*edited to add player number data and link*

Edited by Agent of Change, 21 June 2018 - 05:19 AM.