Jump to content

Tier System Reset


48 replies to this topic

#41 Athom83

    Member

  • PipPipPipPipPipPipPipPipPip
  • The Death Wish
  • The Death Wish
  • 2,529 posts
  • LocationTFS Aurora, 1000km up.

Posted 12 September 2017 - 03:44 PM

View PostWalker Pryde, on 12 September 2017 - 01:00 PM, said:

I purposely played poorly for a lot of games on Saturday in hopes to drop my Tier score down because I am almost to tier 3. Sadly it failed. I am not a bad player, but I am not a great player. This tier system is going to force me to start playing against a level of players I will not be able to keep up with.

Back when I tiered up from 4 to 3, I could feel the IQ of my team decrease. When I flew right through T3 and got into T2, I felt the teams get better again. T3 is not good. T3 is where sanity goes to die.

#42 kesmai

    Member

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

Posted 13 September 2017 - 01:44 AM

Remove tiers.
Install a "savezone" for beginners (~100 matches).

But kes, i don't want to play tier 1 players.
See there is no tier 1, so you wont meet them.

Set the leaderboard to solo q only.

And stop whining and biatching about every little fugg.



#43 Tiewolf

    Member

  • PipPipPipPipPipPip
  • Bad Company
  • 408 posts
  • LocationHessen

Posted 13 September 2017 - 02:22 AM

This hole Tier-System solves nothing. And my guess is that stomps and bad gameplay will stay the same, which is not surprising given the nature of this game.

You will allways have bad games due to the nature of the complicated and inconsistent interactions between loadout/mech, skills, what your team does, kontext (gamerules,map+gamemode) and the behavior of the enemy team.


Your performance has a lot of factors that interact with another:

1. Your skill
...given that you preform in all mechclasses, all weapontypes, on all maps, in all gamemodes, against every enemy strategy and all the time the same

2. Your Mech&Loadout
...some mechs are simply better but not on every map, with every loadout and against the strategy the enemy team plays (E.g. on some maps its better to have the bulk of your weapons mounted sidways than on the top or having indirect fire weapons on polar trench warfare, etc.)

3. Map and Gamemodes
...pritty obvious (Cold or hot maps, small or large maps, lot of brawling cover or open spaces and so on)

4. Strategy of your team and the enemy teams strategy
...obvious too (Fit of your loadout, skills and map+gamemode to the strategies of the 2 teams)

5. The overall performance of your team and the enemy team
...in the most cases if you pull out massive kills and damage you are not good but some of your teammates screwed up or got focused so you could pull it off. In some cases its the behavior of the enemy team too. If all on your team preforms well you can´t pull off 1k+ dmg.

some minor points are not listed here, but i hope my point gets clear why a simple tier based system solves nothing at all and is wasted development time for pgi that they could use better to create new maps/kontext.

Edited by Tiewolf, 13 September 2017 - 02:28 AM.


#44 LordNothing

    Member

  • PipPipPipPipPipPipPipPipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 17,895 posts

Posted 13 September 2017 - 03:11 AM

i had a log grueling climb out of t3 and it took me a lot of back and fourth to get into t2. then i rapidly surged all the way to t1. i find that its easier and more profitable (cbill wise) to pug around fp than it is to play qp. as when i drop in qp im forced to play with high skill players, while in fp i get a more random sampling while usually also filtering out the hyperpugs and stat padders that plaque qp. i dont even want to play in qp now that the match maker requirements are tightened.

everyone and their mother seems to be t1 nowadays. i guess pgi's plan all along was to let everyone bubble up to t1 and then it would be random again, and by that time population would be so low that it wouldn't matter. meanwhile everyone is getting their ego falsely inflated. all it really does is make the game less fun. randomize the match maker.

#45 H Seldon

    Member

  • PipPipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 214 posts
  • LocationChicago

Posted 13 September 2017 - 06:43 AM

Since the change matches have been much better for me. My WL is above 1.0 for once, my KD has been floating over 2 and my Average match score is the highest it's been. There have been fewer stomps and shorter losing streaks as well.

#46 Lily from animove

    Member

  • PipPipPipPipPipPipPipPipPipPipPip
  • The Devoted
  • The Devoted
  • 13,891 posts
  • LocationOn a dropship to Terra

Posted 13 September 2017 - 07:01 AM

Hawken had once a nice post about their MMR system, and I am very sure nearly every games skill distribution of players looks extremely similar

See the topic here

and fancy pic here

Posted Image


So lets ignore the VERY bad rating system MWO has which even allows bad players to get high and good players to stay too long in the lower ranks.

Now if MWO had a proper rating system the skill courve very much would look like this too.
So tell me how would you cut tiers? Unless having 2million concurrent gamers online the t1 and t2 gamers will not be enough for proper matchups, especially when you consider they distribute into FP, soloqueue and group queue (with it's rather difficult way having to match various groupsizes).

I doubt it would ever work to seperate these people properly enough without a massive increase of population.
5% of the players are the light blue, in that graphic, lets consider them the "best of the best" steam stats say 1349 people were playing MWO as peak on the 11th of september. thats just 67 of T1 being online at the same time maybe if non Steam gamers are 2x as much as Steam we would have like 180T1 gamers. Thats not even 8 games filled with T1. and they divide into people in FP, in group play, into Solo and those messing around in the mechlab, or being AFK. What do you think would the average T1 have as a queue time if you only make them getting matched with T1's?

Then comes the above average crowd (orange in the graphic) thats another 10% so true T1 and T2 material basically is compsed of 15% of the playerbase. And thats hard to match them up properly.


you can also take other games skill rating charts and nearly all speak similar numbers, so thats a very given distribution no matter what game you speak about, and then you have your playerbase, with which you need to make sure, that even at the lowest playercount times matchmaking gets properly done.

And from that view, 461 was the low number on the 11th. sept. so theres like 69 people of T1 online at that point of the day. thats not even 3 parallel T1 matches given they were all queueing up in the same queue.

So theoretically building a rating system and a MM is rather easy, but then going into the state of your game will quickly show you how feasable this will be. And then many not so mainstream games will have population issues to make these MM work.


But at this point look again carefully at the chart, the light blue is 5%, the orange is another 5% BUT it actually divides again betwen top of the tops and lowets of the lowest as well as above average and below average. So that 5% and additional 10% I told you are not the real numbers, the t1+t2 would actually be even less, but hard to tell from the graphcis how much they are of that mass. So if you have questions about that graphic ask me, it's not that easy to read first.

So what will your reset do? without fixing the rating system first, you reset will create nothing than potatoes amongst french fry cooks ready to be sliced. But it will not improve the games that happen. And even if the rating system would be perfect on low number population hours your 210 T1+T2 material gamers won't enjoy seeing the same people over and over and over again.

Edited by Lily from animove, 13 September 2017 - 07:34 AM.


#47 Mawai

    Member

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

Posted 13 September 2017 - 07:12 AM

View PostEast Indy, on 09 September 2017 - 07:51 AM, said:

Plain and simple, the problem is that PSR isn't accurate.

Having some kind of exclusive top tier with the 17 best players in the world is a gamer prestige fantasy, not a solution for effective matchmaking.

But if ranking were actually accurate, players could expect that teams would at least be balanced (e.g., each team has 8 true Tier 2 players and 4 true Tier 1 players, or a group of effective players with a few dominant players).

A while ago, PGI boasted of gathering all this "telemetry." This is what I think devs probably can, and definitely should capture and compare to determine player rank:
  • Time until first 30 points of non-strike damage dealt to enemy. Good players are aggressive. (Lower is better.)
  • Average interval between receiving non-strike damage from enemies. Good players are constantly engaged. (Lower is better.)
  • Total damage of non-strike dealt. Good players maximize opportunities. (Higher is better.)
  • Total solo kills. Good players pull their weight. (Higher is better.)
  • Playerbase average score of variant used. Good players excel regardless of tool advantage. (Lower is better.)
Wins should not be recorded. Almost every athletic event since the dawn of time has valued individuals by individual performance, especially when statistics exist that can meaningfully rank one contestant over another.

Metrics like these, taken together, are impossible to fake. You can either do it all or you can't. Use them, and player skill should be accurately reflected in tier assignment.


LOL.

The many different ways to rank people have been discussed many times. Unfortunately, in my opinion, the list you supply is one of the worst I have seen.

1) Time to engage. Good players work WITH their team. The solo dude who runs out to engage quickly often dies quickly, or ends up stringing out their team as some folks follow and some don't. This type of behaviour is often the sign of a bad player, not a good one in my opinion.

2) Taking more damage should be rewarded? Good players try to minimize the damage they receive while maximizing the damage inflicted. Walking out to try and take a little damage to boost your player rating ... I can just see the strategies now ...

3) Total damage dealt. This is what got us into the issues with PSR in the first place. Damage is the biggest single component of the current rating. LRMs are one of the best ways to inflict damage while increasing your survivability (at least if used correctly). I've also seen lots of long range laser builds that have hopeless aim ... they do 1200 damage to targets during a match spread all over several targets. Is that good effective play? No it is spray and pray. On the other hand, the same mech in the hands of a good player will do 600 focused damage, take limited damage in return and virtually kill 4 to 6 mechs by coring key components.

The problem? Raw damage doesn't distinguish good player from bad player.

4) Total solo kills. LOL. It is a team game. The MOST effective way to win, the most EFFECTIVE way to be a good player, is to focus fire down opponents at the first and every opportunity. If someone peaks over a ridge then all 2-5 mechs on the other side should be focus firing them down. The idiot who runs off to find solo duels so they can get a solo kill just hurts their team. It takes longer to kill a target solo, you take far more damage doing so even if you succeed, and while you waste your time getting the solo kill you could have been helping your team mates focus down two or three opponents and win the match.

5) I don't even know what you are referring to with "Playerbase average score variant".

Anyway, the biggest issue with all of these ranking strategies is that they do NOT support the game play. They support a meta game play to obtain an epeen rank that doesn't reflect how good the players are at actually WINNING the game ... which is the entire point in the first place.

With your suggested ranking system folks will run to engage, try to take light damage while hiding behind whatever they can find and doing maximum damage to individual opponents. There is no ranking for team play that actually wins the games in the first place.

The problem is that for matching purposes we want a system that ranks players based on their contribution to WINNING the match which is where all the Elo based systems like the Microsoft Trueskill player ranking system come from.

https://www.microsof...ranking-system/

With 2 teams and 8 players/team it takes about 91 matches (though as many as about 300 depending on conditions) to estimate the player skill reasonably accurately. With 12 players it is a lot more. In an MWO context, this probably wouldn't include light/medium/heavy/assault dependencies since players may be better with one class/build/playstyle than they are with another which makes it even more challenging to estimate player skill.

P.S. I'd also add that your suggested elements to player skill are based only on the one fast brawling type play style and don't address sniper, hit and run, light mech pack tactics or any number of other good player winning strategies ... just the get into battle and do damage archetype.

#48 Thiccacuga

    Member

  • PipPipPip
  • Giant Helper
  • 92 posts

Posted 02 June 2019 - 11:26 PM

One, why did you bump this ded thred wew.

Two, take your meta and shove it cuz I built my own dank stuff and went from T5 to 3/4 T2.





2 user(s) are reading this topic

0 members, 2 guests, 0 anonymous users