110 replies to this topic

[Aidan Crenshaw]

Sniper09121986, on 23 March 2020 - 02:49 AM, said:

Well, who do I meet then? If I am now in T2, then how did these folks get here with that kind of play-style? Their score was less than 100, so no tier advancement up at all. For more or less consistently good players the tier system is just an XP bar, but for these people it would actually function exactly as intended. Have they bought their accounts or something? It just does not add up.

That's easy to answer. In Tier 2 you can drop with Tiers 2-4 or 1-3. Since it needs exactly 1 somewhat good match (400+ MS) to raise a new account from T5 to T4, you can theoretically drop against a player on his second match. Nonetheless, T3 is easily achieved and will get matched with T1 every once in a while. Thus, all sorts of bad players, the oblivious T3's and the bad T1's can be spotted in "high ranking" matches, alas, there is no such thing as a "high ranking" match outside of comp finals.

[IamGumbyDammitt]

Biomechtric, on 23 March 2020 - 03:16 AM, said:

It does function but not over a long period of time which is exactly the problem we have been trying to get across.
The PSR system needs reset at least twice a year. This would have a huge positive impact on match quality for all.
To say it's not working at all just isn't true.

Edit.
I must add that tha matchmaker is working fine, it's the PSR that's the problem.

That depends. If the only goal of the matchmaker is to put together games of 24 people then sure it's working fine. I'd argue that the matchmaker should also have a goal of putting together good matches. In that it fails, even if PSR were better because I don't believe the matchmaker is even attempting to balance games between T1 and T3s on each side (as in each side should have an equal number of T1-T5s in lights-assaults).

Edited by IamGumbyDammitt, 23 March 2020 - 10:58 AM.

[Brauer]

IamGumbyDammitt, on 23 March 2020 - 10:57 AM, said:

That depends. If the only goal of the matchmaker is to put together games of 24 people then sure it's working fine. I'd argue that the matchmaker should also have a goal of putting together good matches. In that it fails, even if PSR were better because I don't believe the matchmaker is even attempting to balance games between T1 and T3s on each side (as in each side should have an equal number of T1-T5s in lights-assaults).

You can believe what you want, there's no real evidence that the 2 tier rule has been tossed by the wayside and plenty of evidence that permabad players can (and DO!) end up in tier 1. Also, with PSR as it is it doesn't matter THAT much if the 2 tier rule has been opened up because being tier 1 means essentially nothing.

[IamGumbyDammitt]

Brauer, on 23 March 2020 - 11:19 AM, said:

You can believe what you want, there's no real evidence that the 2 tier rule has been tossed by the wayside and plenty of evidence that permabad players can (and DO!) end up in tier 1. Also, with PSR as it is it doesn't matter THAT much if the 2 tier rule has been opened up because being tier 1 means essentially nothing.

You misunderstand me. I didn't say the 2 tier rule was dispensed, rather that the MM is not putting say 1 T1 in a light on one team and 1 T1 in a light on the other team, then a T2 in a light on each side, etc. etc. to try to "even" things out on each team. I've seen some very lopsided matches where I looked people up and I don't think it's doing even rudimentary balance amongst the population of 24 it picks for a given match.

[Brauer]

IamGumbyDammitt, on 23 March 2020 - 12:03 PM, said:

You misunderstand me. I didn't say the 2 tier rule was dispensed, rather that the MM is not putting say 1 T1 in a light on one team and 1 T1 in a light on the other team, then a T2 in a light on each side, etc. etc. to try to "even" things out on each team. I've seen some very lopsided matches where I looked people up and I don't think it's doing even rudimentary balance amongst the population of 24 it picks for a given match.

Were you looking people up based on tier, or just based on average match score or something like that? If it was just average match score you might just have been seeing the normal variation in average match score of tier 1 players created by this awful PSR system where a top .05% player whose username rhymes with Wowser is held equal to a 60th percentile player with a WLR and KDR below or just around 1.00

I may be remembering this wrong, but I think it does balance out what tiers the players on each side are in, I don't think it uses mech class in that part though. You'd have to dig into old matchmaker threads to see if there's more detail on that. Even if it isn't balancing the number of T1 players on each side I don't think that matters now with most players in T1.

[justcallme A S H]

SteamingManure, on 22 March 2020 - 03:32 PM, said:

It doesn't say if he starts with just say T1 first and then widen it by T1+T2 if there aren't enough T1s, to T1 +T2 +T3 (max separation allowed) if there aren't enough T1+T2.

It doesn't say because it does not apply.

It's T1 to T3 - all the time. It does not wait for Tier 1 and then open to Tier 2 and then Tier 3.

If it did - it would be in the post. A post that clearly was aimed at reducing search times. Not increasing it.

[justcallme A S H]

Sniper09121986, on 23 March 2020 - 02:49 AM, said:

Well, who do I meet then? If I am now in T2, then how did these folks get here with that kind of play-style? Their score was less than 100, so no tier advancement up at all. For more or less consistently good players the tier system is just an XP bar, but for these people it would actually function exactly as intended. Have they bought their accounts or something? It just does not add up.

The answer is clearly explained HERE. Why have you not read it? This is maths that a 5yo can do.

+/- 2

So Tier 2 can meet anyone in Tier 1 - 3 - 4.

[justcallme A S H]

Sniper09121986, on 23 March 2020 - 02:20 AM, said:

Our world being what it is, PGI is not a charity. I think they took a cold hard look at their player-base and decided that the lambs make up small enough percent of it to be thrown to the wolves without significant risk of PR backlash, especially if they do not even announce the action. If said wolves make up the overwhelming majority of the player-base, they might as well cater to them so that they can find any matches at all. Because otherwise the game ceases to function and no-one is happy at all.

The highly skilled players number in the 100s. That's it.

Out of the 35,000 playing at the time (of the MM change) there was maybe 500-700 people capable of high performance.

The average match score of the entire playerbase is ~220.


All of this data is publicly available. You will see the gradual decline in player skill because the skilled player have mostly given up since Sept '19.

https://leaderboard.isengrim.org/stats

IamGumbyDammitt, on 23 March 2020 - 12:03 PM, said:

You misunderstand me. I didn't say the 2 tier rule was dispensed, rather that the MM is not putting say 1 T1 in a light on one team and 1 T1 in a light on the other team, then a T2 in a light on each side, etc. etc. to try to "even" things out on each team. I've seen some very lopsided matches where I looked people up and I don't think it's doing even rudimentary balance amongst the population of 24 it picks for a given match.

It's never done what you are describing.

Not can it.

Why? Because you'll never find a match. There is not the population to do this. One good player in a Assault is different to one good player in a light. Or a good player in Assault who is bad in a Light.

You're tightening the net far too hard by doing this and thus - it doesn't happen in MWO.

People forget there is not 200,000 people playing this game.

[IamGumbyDammitt]

justcallme A S H, on 23 March 2020 - 02:37 PM, said:

It's never done what you are describing.

Not can it.

Why? Because you'll never find a match. There is not the population to do this. One good player in a Assault is different to one good player in a light. Or a good player in Assault who is bad in a Light.

You're tightening the net far too hard by doing this and thus - it doesn't happen in MWO.

People forget there is not 200,000 people playing this game.

Fair enough, but I have seen games with 5 cadets and all 5 of them were on one side. As you would expect, they got rolled bad. I don't think it's too much to ask of the MM to try to at least split up cadets between teams as evenly as possible. I get that there is a bit more too it since it has to consider tier and weight class, but seriously 5 cadets on one team?

Edited by IamGumbyDammitt, 23 March 2020 - 04:43 PM.

[SteamingManure]

justcallme A S H, on 23 March 2020 - 02:37 PM, said:

The highly skilled players number in the 100s. That's it.

Out of the 35,000 playing at the time (of the MM change) there was maybe 500-700 people capable of high performance.

The average match score of the entire playerbase is ~220.

As a new player and mediocre at best but in Tier 3, I can feel the difference right away between a T1-T3 game vs a T2-T4 or T3-T5 game.

I should not be forced into games with T1s. Call me selfish but as bad as you say most T1s are, they are better than me and as a new player I shouldn't be forced into those games just so they get quicker matches. I'd rather wait longer and get into a game where I feel like I can contribute.

[CFC Conky]

SteamingManure, on 27 March 2020 - 10:51 AM, said:

As a new player and mediocre at best but in Tier 3, I can feel the difference right away between a T1-T3 game vs a T2-T4 or T3-T5 game.

I should not be forced into games with T1s. Call me selfish but as bad as you say most T1s are, they are better than me and as a new player I shouldn't be forced into those games just so they get quicker matches. I'd rather wait longer and get into a game where I feel like I can contribute.

I am a very average T3 player who eventually got to T2. Every time I would move up a tier I wold notice a ratcheting up of difficulty and I would get trashed for the first few dozen matches before things settled down. In T2 the difference is even more pronounced and so far I have been unable to get my w/l or kdr above 1:1 or my avg match score above 200.

That said, I have been un-learning some bad habits I acquired in the lower tiers and learned a lot of new techniques to improve my game.

....Eventually

Good hunting,
CFC Conky

[RickySpanish]

IamGumbyDammitt, on 23 March 2020 - 04:43 PM, said:

Fair enough, but I have seen games with 5 cadets and all 5 of them were on one side. As you would expect, they got rolled bad. I don't think it's too much to ask of the MM to try to at least split up cadets between teams as evenly as possible. I get that there is a bit more too it since it has to consider tier and weight class, but seriously 5 cadets on one team?

Cadets aren't always bad. Check out this guy for example, clearly he is an example of a superior human specimen!

[Temporary Axis]

There has long been a call for PSR to be reset and done so on a regular basis.

This would stop the tier bloat we have.

Furthermore, tiers levels should be set with higher matchcore requirements which must be sustained more strictly. Make the tiers much easier to lose. EG - Ten bad games, you fall.

A proper nominal distribution of players would be:

Tier 1 - 400+ Matchscore (proper elites of the game)

Tier 2 - 300+ Matchscore (Very good pilots)

Tier 3 - 200+ Matchscore (The average player, smack bang in the middle of the curve)

Tier 4 - 100+ Matchscore (Players still grappling with core mechanics)

Tier 5 - sub 100 Matchscore (RWtumbleweed)

[VonBruinwald]

Temporary Axis, on 28 March 2020 - 10:03 PM, said:

There has long been a call for PSR to be reset and done so on a regular basis.

This would stop the tier bloat we have.

Furthermore, tiers levels should be set with higher matchcore requirements which must be sustained more strictly. Make the tiers much easier to lose. EG - Ten bad games, you fall.

A proper nominal distribution of players would be:

Tier 1 - 400+ Matchscore (proper elites of the game)

Tier 2 - 300+ Matchscore (Very good pilots)

Tier 3 - 200+ Matchscore (The average player, smack bang in the middle of the curve)

Tier 4 - 100+ Matchscore (Players still grappling with core mechanics)

Tier 5 - sub 100 Matchscore (RWtumbleweed)

PSR does need a reset and if we reset it just once per year it be enough to separate players out based on some semblance of skill. Unfortunately we would inevitably see a certain portion of the community deliberately tank their first games so they start in a low tier for the purpose of stat-padding/jarls ranking.

Separating tiers by MatchScore doesn't work either, if you put two teams of 400+ scorers in a match the average MS is going to end up to be much lower, the same thing with the sub 100's, they'd see an increase in their average. This is an issue with using any per match performance related stat.

You would have to use a league system to:

Split the population into tiers 1-5 evenly based on their average MG. Then every month switch the top 10% of players in each tier with the bottom 10% of the adjacent tier based on their average MS. This pushes the good performers to T1 and pulls the bad players down without directly comparing T1 and T5 match scores.

[DontTaseMeBro]

VonBruinwald, on 29 March 2020 - 05:34 AM, said:

Split the population into tiers 1-5 evenly based on their average MG. Then every month switch the top 10% of players in each tier with the bottom 10% of the adjacent tier based on their average MS. This pushes the good performers to T1 and pulls the bad players down without directly comparing T1 and T5 match scores.

None of this will mean much if it's still going to do games as T1-T3, T2-T4,T3-T5. That is still too wide of a spread.

[PhoenixFire55]

Temporary Axis, on 28 March 2020 - 10:03 PM, said:

There has long been a call for PSR to be reset and done so on a regular basis. This would stop the tier bloat we have.

And how is regularly placing everyone from T1 to T5 into the same pot any better than tier bloat exactly?

Temporary Axis, on 28 March 2020 - 10:03 PM, said:

Furthermore, tiers levels should be set with higher matchcore requirements which must be sustained more strictly. Make the tiers much easier to lose. EG - Ten bad games, you fall. A proper nominal distribution of players would be: Tier 1 - 400+ Matchscore (proper elites of the game) Tier 2 - 300+ Matchscore (Very good pilots) Tier 3 - 200+ Matchscore (The average player, smack bang in the middle of the curve) Tier 4 - 100+ Matchscore (Players still grappling with core mechanics) Tier 5 - sub 100 Matchscore (RWtumbleweed)

I wonder sometimes if people who post these "suggestions" actually bother to think about em for like 1-2 mins before they do.

You took these arbitrary tier boundaries from where exactly? ... Have you perhaps considered that people with average matchscore of 400+ are less than 1% of population. Same for people with under 100 matchscore. Which means that you literally put 2% of population into two tiers, and remaining 98% into 3 tiers. Makes sense (no, no it doesn't).

The easiest way to separate people is to create dynamic boundaries between tiers. Player population is distributed along a near-gaussian curve, lets call it S(x), i.e. a number of players (S) with an average matchscore (x). Make tier 1 boundary x1 such that the integral of S(x) from +infinity to x1 nets (1/n)*100% of players, where n is the desired number of tiers. Tier 2 boundary would be such that the integral of S(x) from +infinity to x2 nets (2/n)*100% of players. And so on. Now these boundaries you can update every day or every hour or every damn 10 seconds if you want. Even with a population of 100k players it'll take like 2 seconds. Every tier would have a nearly exact same amount of players. Depending of the average times it takes to create matches you can then allow or disallow matching of players of different tiers.

Simple, right? Takes like 2-3 mins to figure out. Yet here we are, 7 years into the whole "MM phases", with the same old broken p.o.s. that makes no sense no matter how you look at it. Does take a special kind of effort to remain Minimally Viable^TM, doesn't it?

[DontTaseMeBro]

PhoenixFire55, on 29 March 2020 - 08:19 AM, said:

And how is regularly placing everyone from T1 to T5 into the same pot any better than tier bloat exactly?


I wonder sometimes if people who post these "suggestions" actually bother to think about em for like 1-2 mins before they do.

You took these arbitrary tier boundaries from where exactly? ... Have you perhaps considered that people with average matchscore of 400+ are less than 1% of population. Same for people with under 100 matchscore. Which means that you literally put 2% of population into two tiers, and remaining 98% into 3 tiers. Makes sense (no, no it doesn't).

The easiest way to separate people is to create dynamic boundaries between tiers. Player population is distributed along a near-gaussian curve, lets call it S(x), i.e. a number of players (S) with an average matchscore (x). Make tier 1 boundary x1 such that the integral of S(x) from +infinity to x1 nets (1/n)*100% of players, where n is the desired number of tiers. Tier 2 boundary would be such that the integral of S(x) from +infinity to x2 nets (2/n)*100% of players. And so on. Now these boundaries you can update every day or every hour or every damn 10 seconds if you want. Even with a population of 100k players it'll take like 2 seconds. Every tier would have a nearly exact same amount of players. Depending of the average times it takes to create matches you can then allow or disallow matching of players of different tiers.

Simple, right? Takes like 2-3 mins to figure out. Yet here we are, 7 years into the whole "MM phases", with the same old broken p.o.s. that makes no sense no matter how you look at it. Does take a special kind of effort to remain Minimally Viable^TM, doesn't it?

I like where you went with this. It makes sense to spread out the population into large enough chunks.

One thing I would perhaps add is that if 400+ scorers are 1% of the population, then perhaps they should get their own indicator. Possibly T1+ and the MM would need to evenly match those folks on either side as they can have a huge impact on a given game.

[Kodyn]

I feel like the issue with any of these MM/PSR discussions is they're now moot. We no longer have the population to support any kind of matchmaking whatsoever other than whatever we have. I've been in other low-pop games with the same issue. No matter how good the formula, if you don't have enough players, it's always going to be the same people facing the same people anyway.

Much like how we see nothing but the same names over and over now. Feels like window shopping or going to a strip club, all this talk of what would actually work better or make a decent MM formula....teasing ourselves with things that will never happen in MWO.

[PhoenixFire55]

DontTaseMeBro, on 29 March 2020 - 08:44 AM, said:

I like where you went with this. It makes sense to spread out the population into large enough chunks.

Thats literally the only point of tier or any kind of classification really.

DontTaseMeBro, on 29 March 2020 - 08:44 AM, said:

One thing I would perhaps add is that if 400+ scorers are 1% of the population, then perhaps they should get their own indicator. Possibly T1+ and the MM would need to evenly match those folks on either side as they can have a huge impact on a given game.

And why do 400+ scorers get this kind of special treatment while 390+ scorers don't? And if 390+ gets is, then why 380+ doesn't? ... MM should create balanced matches. How it should do that is debatable and needless to say there are limitless options in terms of how you balance that one 400+ guy and everything else for that matter. One way is to have another 400+ guy on the other side. But it doesn't mean its the only way. There are literally hundreds of well balanced combinations within a group of just 24 randomly drawn players. Times and times more within what MM actually works with. Needless to say a MM can have multiple matching algorithms as well a self learning mechanism that tracks "best" outcomes and what algorithm has been used to achieve those.

But hey, this stuff requires you to actually ... you know ... do your job as a programmer, and ain't nobody got time for dat in 21st century. Giving effort and even giving a damn is LosTech.

Kodyn, on 29 March 2020 - 08:49 AM, said:

I feel like the issue with any of these MM/PSR discussions is they're now moot. We no longer have the population to support any kind of matchmaking

It takes 24 people to create a balanced 12v12 match. Literally any 24 people.

[DontTaseMeBro]

PhoenixFire55, on 29 March 2020 - 09:10 AM, said:

And why do 400+ scorers get this kind of special treatment while 390+ scorers don't? And if 390+ gets is, then why 380+ doesn't? ... MM should create balanced matches. How it should do that is debatable and needless to say there are limitless options in terms of how you balance that one 400+ guy and everything else for that matter. One way is to have another 400+ guy on the other side. But it doesn't mean its the only way. There are literally hundreds of well balanced combinations within a group of just 24 randomly drawn players. Times and times more within what MM actually works with. Needless to say a MM can have multiple matching algorithms as well a self learning mechanism that tracks "best" outcomes and what algorithm has been used to achieve those.

There has to be a cutoff somewhere, and as you go higher up the avg match score the more impactful those guys get. Maybe 350 is the right number. Either way it's a small percentage of the population but if they are going to be put into games with other T1s, they need to be doled out carefully for reasonably decent matches.