43 replies to this topic

[sub2000]

I've seen new topics and posts with complains about Tier 1 games.
First reminder about how MM actually works:
it checks your Tier and registers weight class of your mech. It tries to match you to the existing teams. Since the chance is not that big for precise match due to low population, it waits for ~5 minutes and expands search borders. Hence outside of peak hours tier 5 can get play with or against tier 1 players. But MM still checks weight classes (so if your team has 6 assaults, which is the norm in tier 5-3 games, it means that the enemy has also 6 assaults) and of course MM tries to match tiers. It is not clear what mechanisms are used (averaging, sum, or actual tier score) but the result is rather OK. Higher tier players are assigned to lower tier teams. Balance is still present.

I have to remind that XP based Tier system was discussed and criticized to death on this forum.
The main complain is that bad players still arrive to the Tier 1 just much later than good players.
As an example I play one year on potatoe level (average MC is less than 200) but still I have arrived firmly in Tier 2..

The problem is that after getting Tier 1 either good or bad players are indistinguishable for the MM. It means in practice that existing MM easily assign few strong players to one team and noobs (or potatoes as they are called here) to another, because from point of view of MM they are of the same quality. From our POV MM goes mad, and simply stops working.
There is no MM in tier 1 games.

time or server are of no importance.

Edited by sub2000, 11 January 2018 - 11:31 AM.

[Verilligo]

So you're claiming that as soon as the MM is forced to assign a T1 player to a game with non-T1s, the side with the first T1 gets stacked with T1s because the MM loses its mind and tries to push all the T1s currently in the queue through? There are elements of that which do make sense, I'll admit. I've noticed that since I've moved from T2 to T1, I have a much higher likelihood of being matched with the same people as I had in the previous match and often end up on the same side as well.

[Bombast]

Actually, I'm pretty sure the Tier based MM collapsed or was turned off about a week prior to Christmas. I haven't checked stats in a week, but everytime I did confirmed what I suspected - Every game had people that could be, at best, T4. This was based on game count as well as performance - There was simply no what these people could have gotten to T3.

Anecdotally, playing today, I'm 99% certain this is still true - There is no such thing as a T1, or even a T1/2, match anymore. So its all moot. We're likely back at the T4/5, T1/2/3 system.

Edited by Bombast, 11 January 2018 - 12:28 PM.

[Verilligo]

Bombast, on 11 January 2018 - 12:09 PM, said:

Actually, I'm pretty sure the Tier based MM collapsed or was turned off about a week. I haven't checked stats in a week, but everytime I did confirmed what I suspected - Every game had people that could be, at best, T4. This was based on game count as well as performance - There was simply no what these people could have gotten to T3.

Anecdotally, playing today, I'm 99% certain this is still true - There is no such thing as a T1, or even a T1/2, match anymore. So its all moot. We're likely back at the T4/5, T1/2/3 system.

I can confirm for a fact that I've been matched with T3s and I know B33f has been as well. T4s and T5s usually don't mention their tier, so I have no idea if I've been in games with them or not.

[Cade Windstalker]

I'd like to propose something of a counter-scenario for what would happen if the MM was entirely based on an ELO-esque system, as it used to be in the past.

ELO rating systems, for those who may not be aware, are a way of gauging the relative skill of a population of players based on their performance against others in that population. The important piece of information for the point I'd like to make is that they tend to produce a bell-curve like distribution of players. The players at the very high end will generally be significantly outnumbered by those at the very low-end.

So what does this mean in practice?

It means that for a game with anything other than an extremely large population relative to the number of players in its matches you're going to have top-tier players in your games anyway but for those very top-tier players they're likely to experience longer wait times and still get into the same piss-poor quality of match, if not worse.

How could it be worse exactly? Lets say that you have someone who is *really* good, and has an ELO of 2500. The matchmaker waits 5 minutes and can't find someone to balance him out, so instead it grabs a few people with ELOs of 1500 and puts them on the other team so the weighted sum of the ELOs on both teams more or less balance out. This means that our 2500 ELO player is now probably in a harder match because while the *average* skill on both teams is the same but the *median* skill on his team is going to be lower since there are some walking-potatoes on his team to balance out his high ELO.

For our high-ELO player having literal random matchmaking would be better and improve the overall quality of his games, and the same goes for the low-skill players who end up being used as counter-weights in the matchmaker since they'll end up with a more even skill distribution instead of being more likely to fight intermediate players when high-skill players are present.

None of this is to say that the current system is *good* just that I'll take basically random matchmaking above a certain tier over playing "ELO Russian Roulette" where the 2500 ELO god-player is the bullet.

[Verilligo]

Cade Windstalker, on 11 January 2018 - 12:47 PM, said:

I'd like to propose something of a counter-scenario for what would happen if the MM was entirely based on an ELO-esque system, as it used to be in the past.

ELO rating systems, for those who may not be aware, are a way of gauging the relative skill of a population of players based on their performance against others in that population. The important piece of information for the point I'd like to make is that they tend to produce a bell-curve like distribution of players. The players at the very high end will generally be significantly outnumbered by those at the very low-end.

So what does this mean in practice?

It means that for a game with anything other than an extremely large population relative to the number of players in its matches you're going to have top-tier players in your games anyway but for those very top-tier players they're likely to experience longer wait times and still get into the same piss-poor quality of match, if not worse.

How could it be worse exactly? Lets say that you have someone who is *really* good, and has an ELO of 2500. The matchmaker waits 5 minutes and can't find someone to balance him out, so instead it grabs a few people with ELOs of 1500 and puts them on the other team so the weighted sum of the ELOs on both teams more or less balance out. This means that our 2500 ELO player is now probably in a harder match because while the *average* skill on both teams is the same but the *median* skill on his team is going to be lower since there are some walking-potatoes on his team to balance out his high ELO.

For our high-ELO player having literal random matchmaking would be better and improve the overall quality of his games, and the same goes for the low-skill players who end up being used as counter-weights in the matchmaker since they'll end up with a more even skill distribution instead of being more likely to fight intermediate players when high-skill players are present.

None of this is to say that the current system is *good* just that I'll take basically random matchmaking above a certain tier over playing "ELO Russian Roulette" where the 2500 ELO god-player is the bullet.

On the flipside of that, eventually the 2500 ELO player will be placed in enough bad matches that their ELO will plummet to something slightly less astronomical, averaging out with the ELO of their fellow divine beings. Or, in certain instances, that player will be so good that it doesn't matter if they carry for 11 other players. I have faith, however, that you could sufficiently refine the matchmaking criteria so that you aren't stuck with this type of situation.

[Alex Morgaine]

Verilligo, on 11 January 2018 - 11:44 AM, said:

So you're claiming that as soon as the MM is forced to assign a T1 player to a game with non-T1s, the side with the first T1 gets stacked with T1s because the MM loses its mind and tries to push all the T1s currently in the queue through? There are elements of that which do make sense, I'll admit. I've noticed that since I've moved from T2 to T1, I have a much higher likelihood of being matched with the same people as I had in the previous match and often end up on the same side as well.

I think hee meant that when you hit T1, whether you are legit or ride the xp ride it'll just throw T1 left and right because that's all it looks at, not kdr, wl, or unlocked skills/drops in the Mech you choose. A t1 with a brand new stock 3025 Mech pilot can be paired up with vs a meta ubertech pilot x12 times. Get enough coin flips heads and tails is going to have a bad time.

Hence the "mm gone loco!" posts

[Daurock]

Verilligo, on 11 January 2018 - 01:06 PM, said:

On the flipside of that, eventually the 2500 ELO player will be placed in enough bad matches that their ELO will plummet to something slightly less astronomical, averaging out with the ELO of their fellow divine beings. Or, in certain instances, that player will be so good that it doesn't matter if they carry for 11 other players. I have faith, however, that you could sufficiently refine the matchmaking criteria so that you aren't stuck with this type of situation.

To add to this. ELO systems make stats (I.E. KDR, Winrate, Dmg/Game, etc) for all but the extremely skilled players gravitate towards the server average. (I.E. your winrate always stays about 50%, but as you get better, you get pitted against better players, and/or grouped with dummies on your own team) For everyone NOT capable of carrying 10-12 others on their own, this will create a system where eventually, their stats will be indistinguishable (Except for that ELO rating itself) from a newbie shooting other noobs.

Oddly, I prefer the current system over such an ELO system. I rather like having a positive Winrate, and Positive KDR, things that would generally disappear with an ELO system that attempted to create "Skill balanced" matches.

Edited by Daurock, 11 January 2018 - 01:35 PM.

[Cnaiur]

I'm not buying that the MM is routinely mixing T4 & T5 players with T1. I have a T4 alt I play when I've been drinking and the names of the people I'm matched with in those games are different (and have significantly worse numbers on average), but more importantly, the level of play is far, far worse. You can get away with multiple blunders in T4/5 games that would instantly get you crippled or killed in T1, simply because most people in those lower tiers can't shoot accurately. Lord knows lurmboats are still a thing in T1, but I've been in multiple T4/5 matches with like 6-8 lurmboats on a side. Never seen that in T1/2.

I think it's more likely that the Xmas events brought out/back a bunch of mediocre and out of practice T 1/2/3 players. Their mechs are no longer meta, if they ever were, they may not have bothered with the skill maze, and a lot of them were probably pretty casual to begin with and just showed up for a month of free stuff. I'm sure to really good players, it felt like an invasion of low tier potatoes.

[Bombast]

Cnaiur, on 11 January 2018 - 02:05 PM, said:

I'm not buying that the MM is routinely mixing T4 & T5 players with T1. I have a T4 alt I play when I've been drinking and the names of the people I'm matched with in those games are different (and have significantly worse numbers on average), but more importantly, the level of play is far, far worse.

The level of play I witnessed during the described period was essentially exactly what I've seen in lower tier matches, and as I said, I checked - Many of the players in question literally hadn't played enough games to break out T4.

I wouldn't say they were the majority, still but they were there, along with a metric ton of confirmed T3 players, which should only show up when the MM gives up, not during matches that took half a second to put together. Unless PGI has rolled the MM back, which it almost certainly has, at the least.

I'm hoping PGI looks into the MM, or PSR, or whatever. The games become less and less playable over the last 30 days, to the point to where even winning with 400+ match score feels terrible because you know you didn't really earn it. For example, earlier today, we won a Grim-Dom match. From the bad side. 10-2. It was just... sad. I'd rather wait 8 minutes than play these games.

Edited by Bombast, 11 January 2018 - 02:17 PM.

[Water Bear]

I don't have a problem. Most of my games, every lance on both sides has at least two people getting 400+ damage. Most lances have at least one with 600+.

I also don't see very many games that go 12-2. They do happen, but I have seen far more games go, for example, 7-12, or even 10-12. I just assumed that at high T1, most people are pretty competent at the game and aren't getting blasted out in the open, failing to torso twist or find cover, etc.

[Jonathan8883]

Verilligo, on 11 January 2018 - 12:22 PM, said:

I can confirm for a fact that I've been matched with T3s and I know B33f has been as well. T4s and T5s usually don't mention their tier, so I have no idea if I've been in games with them or not.

Heh, I got matched with the B33f in my first two weeks playing. I had no idea who he was or anything until much later, but I remember people acknowledging him.

[UnofficialOperator]

Anecdotal but I am of the belief that overall WLR/KDR (I forgot which but probably WLR) is taken into account.

That is why I am no longer wary of "famous" good players on the opposite team because it means that they are carrying a few potato pilots on their team. And I would rather take 2x 1.0 WLR pilots than 1x 1.7 WLR + 1x 0.3 WLR pilots.

Don't believe me? Screenshot the pilot names after every match and check the leaderboard. The WLR should match. (I've done it a few months back but it got tiresome fast. That's why they should have a more transparent way to check ingame via mods etc)

PS: There is 1 exception though, you can sort of game it to encounter more lower tiers than usual but I won't say it here.

Edited by UnofficialOperator, 11 January 2018 - 03:47 PM.

[Bombast]

UnofficialOperator, on 11 January 2018 - 03:43 PM, said:

Anecdotal but I am of the belief that overall WLR/KDR (I forgot which but probably WLR) is taken into account.

I was going to test this, but Jarl's has bugged out on two matches in a row and I don't care enough to keep going.

[UnofficialOperator]

Bombast, on 11 January 2018 - 03:52 PM, said:

I was going to test this, but Jarl's has bugged out on two matches in a row and I don't care enough to keep going.

Yeah I'm going to try again, compiling the screenshots.

[Cnaiur]

Bombast, on 11 January 2018 - 02:17 PM, said:

The level of play I witnessed during the described period was essentially exactly what I've seen in lower tier matches, and as I said, I checked - Many of the players in question literally hadn't played enough games to break out T4.

I wouldn't say they were the majority, still but they were there, along with a metric ton of confirmed T3 players, which should only show up when the MM gives up, not during matches that took half a second to put together. Unless PGI has rolled the MM back, which it almost certainly has, at the least.

I'm hoping PGI looks into the MM, or PSR, or whatever. The games become less and less playable over the last 30 days, to the point to where even winning with 400+ match score feels terrible because you know you didn't really earn it. For example, earlier today, we won a Grim-Dom match. From the bad side. 10-2. It was just... sad. I'd rather wait 8 minutes than play these games.

I don't doubt your experience and there has absolutely been some truly awful play at the T1/2 level during the holidays, but I immediately noticed a significant difference whenever I'd swap over to my T4 alt. Most (not all, but most) T1/2 players can aim reasonably well. Enough to punish people that run around in the open in anything bigger than a locust or commando. T4/5 was full of people that struggled to hit any moving target faster and/or smaller than an atlas. The only real threat was usually another smurf account or the accumulated wear and tear of countless lurms.

Anyway, you can't rely on the Jarl's List to tell you how many games somebody has. I have several thousand games more than what's showing on the Jarl's List, from back before seasons were a thing.

[UnofficialOperator]

Just an example of how badly someone with a high WLR/KDR needs to carry. This is a famous pilot. So they need to give him weaker team mates to compensate for his stats. Of course like I said earlier, this is anecdotal. I'll compile more if I have the time. Well I'd rather have 2 average pilots than 1 exceptional pilot and a few potatos.

But overall, the MM does seem to try to equalize so for OP to say that there is no MM is false. Oh and also why pilots should stick to their usual mech which gave them their WLR. If you do well in Assaults with a high WLR and go down to playing Lights with a bad WLR, I think it might affect the MM (although I have no proof that MM accounts for this or does not account for this. Pretty sure we would have won this if I was in assault but I gave up my spot to a <1 KDR assault player)

Look at the KDRs... it is very telling that they gave the famous pilot more teammates who do 0.5 to 0.6 KDR. These people cannot kill.

PS: Of course I might be wrong too. I'm **** at stats so take this with a grain of salt

Edited by UnofficialOperator, 11 January 2018 - 05:39 PM.

[Grus]

To add. I've had a lot of QP matches where there are atlest a lance or more that didn't brake 100 damage for each player in said lance. Some so little damage, that combined were lucky to brake 100. It is frustrating to say the lest.

[UnofficialOperator]

This is another example of another famous pilot. The stats of this pilot is so high that it skews the MM. Again, I would rather have average pilots with average MS. This famous pilot who seems to pilot assaults mostly, goes down to a lower tonnage and frees up an assault slot for a potato. And this good pilot actually does 976 damage in a cicada (outscoring both teams) vs the potato assault pilot with 71 damage.

Seems like letting potatos pilot assaults is a good predictor of losses.

So again to say that MM doesn't do some sort of balancing is wrong. But is it the right way to balance is another question.

Edited by UnofficialOperator, 11 January 2018 - 06:19 PM.

[Shifty McSwift]

I love when developers leave their systems open to speculation rather than giving us simple details about how things work, oh wait no the other thing, I hate it. But yeah when they do it with important internal systems like crits (lacking any kind of in game explanation), why would it surprise me to see it elsewhere.

This is most definitely not an MWO specific problem, but it exists here for sure, and it's not just a rail on the devs for the sake of it, I defend devs constantly, but bad or lazy systems are still exactly that.