97 replies to this topic

[Wintersdark]

Also, as always:

Attempting to level mechs during events is always fraught with peril. In this particular event, you find a lot of people doing what I mentioned above - bum rushing to get a KMMD without concern to what happens after that. It's not as bad as the behavior seen in many other events, but it does impact gameplay.

[Leeroy Jenkin]

Here is my thoughts on the topic.....


Every game I drop, I kinda just talk crap in the all chat right out the gate. I kinda check the players out on both sides and see if I have dropped with them before....if I have a clue about their skill level as a pilot. If I don't notice any pilots that stick out as real threats with real skill, I assume we are playing against players that can in fact play(Never underestimate your enemy). If I see ones that I know are good pilots, I let my team know to focus them right out the gate.

If I don't hear any chatter back from my teammates, I hold back and out of site of the enemy team for a few minutes, and just wait to see how my team positions itself. If they are not watching out for UAV's, taking them down as soon as they go up, I know I am playing with pugs that don't have a clue or just don't care if they do. I will then make a choice, depending on the flow of the battle, the position of the enemy and my teammates, and either stay with them focusing down enemy mechs, or I will break off from them, and go completely around the enemy and flank them.

In one game I played last night, I was in a unmastered hunchback II C and noticed that none of my teammates was dropping the two UAV's that the enemy popped on us. We was on Polar, and the enemy had Three missile boats, a crow, a highlander II, and a stalker clear back from their teammates and was pounding our team with LRM's around 900 meters back from their front line. The three missile boats kept our team from pushing due to being pounded. With my Hunch back II C, only doing 67.9 km... I was able to go completely around them on the right side of the map, and came up on them from behind as if I had ECM. They didn't have a clue or expect it... and I ripped through the storm crow lrm boat.... killing it, moved behind the Highlander II, cored it from behind, and then proceeded to work the stalker over. I killed all three of their missile boats single handed, and them moved in on the rest of their team as my team finally started to push in and do something productive. I killed 6 mechs that match, and had over 800 damage.

The crazy thing about this match.... it ended with just me, and a chetta with two large lasers and ecm on the enemy team. I was legged.... and only had 65 rounds of AC/10 ammo and 2 med lasers. I was able to keep this chetta in my med lasers range after I ran out of AC/10 ammo, and the match ended after 4 mins of him trying to kill me, and shot after laser shot exchanged by us both. I used hills and ditches, buildings to move on him, (legged no less) and with only 3 seconds left..... used my jumps jets as he did the same coming over a building and I cored him. I was at 26%..... what a game!

Sometimes, when your teammates are noobs, you have to play your own game and break off from them and spread chaos and confusion from the flanks of the enemy. If you can pull it off...... you can break up the enemy team and just tear through them carrying your team to victory. Sometimes..... it fails, but when it works, its a beautiful thing indeed!






























p.s>>>>>> Yes, I am the guy that yell's out over com's everytime I kill a enemy mech or get a kdmm.....LEEEEERRRRROOOOOOY JENKKKKINS! Its not my fault. PGI really needs to make a Leeroy Jenkins warhorn, I know I will buy it!

Edited by Leeroy Jenkin, 08 October 2016 - 09:02 AM.

[Appogee]

Wintersdark, on 08 October 2016 - 07:46 AM, said:

And there can't be anything else going on. Are you seriously insinuating that the MM is somehow setting you personally up to fail?

Of course not. But I do suspect that, just like when Elo was the basis of matchmaking, there is some kind of weighting going on in the matchmaking that kicks in when you get too many matches ahead in wins.

In fact, I think the algorithm that was used under Elo to create teams with an equal average Elo, based on averaging personal Elo score per weight class, is still being used somehow.

Wintersdark, on 08 October 2016 - 07:46 AM, said:

All the MM has when making a match is PSR ratings and mech classes.

Says who? When they introduced PSR, they explained how teams would be drawn from the same PSR level, or after a time, 2 levels either side. But that doesn't preclude other mechanics or calculations also being used behind the scenes.

For example, maybe fine skill scores are still being kept per weight class, and are still being used to apportion players across the two teams into two teams of roughly equal average skill.

Given PGI's minimalist approach to any change, PSR may just be some kind of feel good overlay, or an additional gating factor, on top of the old system.

Bishop was saying that he too had noticed the "sine wave" effect of runs of wins and runs of losses. Maybe, I didn't notice the 13-win streak, just the 13-loss part of the sine wave.

Wintersdark, on 08 October 2016 - 07:46 AM, said:

But, you're taking 24 random people of wildly differing skill levels and cramming them into a match with basically no communication. Match outcomes are highly random.

Should be. But apparently, aren't. There were 8000 to 1 odds that I would get 13 matches in row where I lose despite making a high personal contribution to the probability of a win.

Sure, probability says that anything is possible. However, if it was truly random, then it was 7999 times more likely for me to have won a match somewhere in that 13.

Wintersdark, on 08 October 2016 - 08:04 AM, said:

Attempting to level mechs during events is always fraught with peril. In this particular event, you find a lot of people doing what I mentioned above - bum rushing to get a KMMD without concern to what happens after that. It's not as bad as the behavior seen in many other events, but it does impact gameplay.

Sure. But that applies to both teams. It has no bearing one way or the other on why I would lose 13 straight. In fact, as I said earlier, the build I was using had proven highly effective prior to this losing streak.

Incidentally, I have won 2 matches out of 11 today... again, despite getting a top-3 damage and kills score in almost every match. My win:loss over the past two days has dropped from 1.110 to 1.098.

Edited by Appogee, 08 October 2016 - 09:26 AM.

[Wintersdark]

Appogee, on 08 October 2016 - 09:17 AM, said:

Of course not. But I do suspect that, just like when Elo was the basis of matchmaking, there is some kind of weighting going on in the matchmaking that kicks in when you get too many matches ahead in wins.

This is completely wrong, and somewhat ridiculous.

The Elo matchmaking system was very clearly described by it's author. PSR works in exactly the same way building a match (a score is a score), the difference with PSR is in how scores are modified after the match.

In Elo matchmaking, the goal of the system seems to be to get you to a 1:1 win loss ratio. However, that is the (ideal) consequence of how the system works; it's the effect, not the cause. The actual goal is just to make as fair matches as possible with players at the minimal delta. It's just trying to get guys with the same score in a match together, but recognizing that's impossible it used the two teams average scores compared in calculating how to adjust scores after the match.

The matchmaker under PSR and Elo both has only two inputs it uses: The mech you're piloting, and your current score. It neither knows nor cares what your WLR is or what kind of streak you're on.

People liked to say the "matchmaker makes you lose if you win too often", but that's just someone who doesn't understand something making it up as he goes along. If you won too often, it would raise your score, which would result in being placed in higher ranked games, until such a time as you started losing matches, dropping it down. While this results in a "sine wave" style of movement in your score, it's both very slow (we knew exactly how fast the score moved in Elo in absolute terms, and we can visibly see your PSR score move via the position of the bar) and due to the nature of PSR it's very difficult to lose rating, so you don't see that happening "deliberately" here. Indeed, it didn't happen much in normal Elo play either, unless you played a lot in the group queue with a strong group, which would inflate your ranking significantly.

Quote

In fact, I think the algorithm that was used under Elo to create teams with an equal average Elo, based on averaging personal Elo score per weight class, is still being used somehow.

The matchmaker is the same; it hasn't changed. We only have one score, however, not 4 as per Elo. It DOES create teams as close to the seed ranking as possible, but it does not do any averaging. It simply takes the 24 players as close to the seed ranking as possible in valid weight class mechs to build two teams. That's it, no more, no less.

Quote

Says who? When they introduced PSR, they explained how teams would be drawn from the same PSR level, or after a time, 2 levels either side. But that doesn't preclude other mechanics or calculations also being used behind the scenes.

Karl Berg fully outlined how the matchmaker works. There are no other mechanics or calculations, nor would there be any reason to. It's kind of narcissistic to imagine the matchmaker is setting you up to win or lose. It didn't under Elo, it doesn't now. It doesn't care about you. You're just one totally irrelevant data point in a sea of data points.

The matchmaker is simple.

Quote

For example, maybe fine skill scores are still being kept per weight class, and are still being used to apportion players across the two teams into two teams of roughly equal average skill.

They're not; they specifically said they were dropping 4 scores and going to one. We HAVE a fine score - PSR rating (not tier; tier is purely visual for our benefit, being values equal to 1/5 of the total PSR range. The MM is using that.

Quote

Given PGI's minimalist approach to any change, PSR may just be some kind of feel good overlay, or an additional gating factor, on top of the old system.

This is true, but the problem is that you totally don't understand how the old system worked.

All PSR is is a new way to modify your scores at the end of the match, and a way to show players their score without showing specific numbers.

Quote

Bishop was saying that he too had noticed the "sine wave" effect of runs of wins and runs of losses. Maybe, I didn't notice the 13-win streak, just the 13-loss part of the sine wave.

Again, you're running way afoul of logic here in pursuit of seeing a pattern that doesn't exist. There's a name for that; it's a common thing, humans love to do it. But there's not a pattern here.

We know how the system works. It's been clearly explained. The very notion that it's "setting you up to lose" because you won too many, or "letting you win" because you've lost too many is absolutely ridiculous.

Quote

Should be. But apparently, aren't. There were 8000 to 1 odds that I would get 13 matches in row where I lose despite making a high personal contribution to the probability of a win.

Ultimately, matches are effectively a coin toss. Flipping a coin and getting heads 13 times in a row is not impossible; it's not even really unlikely. After all, each new game is unaffected by the last - just because you lost your last game doesn't make you more likely to win your next. What does affect each match most is 24 totally different people with completely uncontrolled things going on in their lives. Are they focused? Are the playing hard, or derping around? Because of how PSR works, you can tromp quickly to T1 with some solid play, then just derp forever in terribad mechs. You're still getting put into high end games, because even if you play so horribly badly that you lose rating, it'll take AGES for your score to dip so far you don't get pulled into all those high tier matches... For the ones that actually are high tier, for that matter: your as likely as a T1 to get pulled into mid-rank matches, because again 50 players to choose from.

Quote

Sure, probability says that anything is possible. However, if it was truly random, then it was 7999 times more likely for me to have won a match somewhere in that 13.

Sure. But that applies to both teams. It has no bearing one way or the other on why I would lose 13 straight. In fact, as I said earlier, the build I was using had proven highly effective prior to this losing streak.

Incidentally, I have won 2 matches out of 11 today... again, despite getting a top-3 damage and kills score in almost every match. My win:loss over the past two days has dropped from 1.110 to 1.098.

I have in no way insinuated that you were responsible for losses, so let's not go there, it's not constructive.

Note that in basically every loss (there's one in every match, remember) there are players getting top 3 scores (because of course!) and kills. 12-0 wipes are quite uncommon. Those guys are just like you. You are not special. The matchmaker doesn't care about you, any more or less than the other 23 players. The only thing the MM uses is your weight class and rating.

---

Here's the thing. Even if you want to propose that the author of the matchmaker flat out lied, and it's all a cunning experiment set to make specific players win and lose....

...It would still be impossible to achieve what you're proposing.

Because the vast majority of matches (I can link the post showing the math here if you really need it) are composed of players from a full third of the total range of the PSR spectrum. That's a HUGE skill delta. Given current player counts, there are roughly 50 players at best queued and looking for a match at any given time that are eligible to play together. The only way we get matches at all is because the spread valves cover such a wide range. As such, there simply are not enough players available ever to really achieve anything other than getting a very, very, VERY roughly put together match. (and that's ignoring the whole "xp bar" nature of PSR in the first place!)

Why even bother coding more conditionals into the matchmaker? For it to try to make specific players win or lose is basically impossible. To make you lose, it would have to deliberately (for unknown reasons, why YOU particularly?) put you with worse players, then fill the opposing team with better players. But it would have to do that for everyone, because I'm pretty damn sure PGI hasn't included a "Screw Appogee" toggle. That's just ridiculous, it's massively complicated and couldn't work. But still, they'd have to create that, then have the author of the system flat out lie about how it works while he describes the algorithm in detail, even after he left the company.... and even then, it still couldn't.

That's just silly.

Edited by Wintersdark, 08 October 2016 - 10:54 AM.

[Wintersdark]

A major consideration here is people assign a lot to "the matchmaker", but in fact the matchmaker only does one thing:

Takes scores and weight classes, and fits players into teams with that information only. It then passes those player lists to the dedicated server (the server that runs the actual battle). After passing the players off to the dedicated server, the matchmaker is completely done.

Modifying the score after the match has nothing to do with the matchmaker, that's separate, a return from the dedicated server, along with how much XP you gained and cbills won.

[MW222]

Appogee, seen you play. Come on you know that you can do it 100% right and still lose. If you don't know by now that sometimes you eat the bear and sometimes he eats you. From memory, you're the one doing the eating.




It's about the fun. Take a break, try a different mech.

Oh sorry, couldn't resist with the clip.

Edited by MW222, 08 October 2016 - 11:33 AM.

[A Shoddy Rental Mech]

Wintersdark, on 08 October 2016 - 08:04 AM, said:

Also, as always:

Attempting to level mechs during events is always fraught with peril. In this particular event, you find a lot of people doing what I mentioned above - bum rushing to get a KMMD without concern to what happens after that. It's not as bad as the behavior seen in many other events, but it does impact gameplay.

Not much of a choice when to level mechs.
4 days last month where there wasn't an event running
8 weekends since there wasn't an event running.

[Appogee]

Wintersdark, on 08 October 2016 - 10:47 AM, said:

This is completely wrong, and somewhat ridiculous.

You either don't know how the system worked, or misunderstood what I was saying. So don't describe 'reality' as 'ridiculous', kthx.

In MWO, players had 4 Elo scores, one per weight class. The matchmaker would assemble teams to try and get a roughly equal average Elo on both side. Individual Elos would go after the match, based on which side won the match.

A team that won the match despite having a lower average Elo would have its members' Elos increased. A team that lost despite having a higher average Elo would have its members' Elos decreased.

That is how the system was implemented in MWO.

Wintersdark, on 08 October 2016 - 10:47 AM, said:

The matchmaking system was very clearly described by it's author.

Movement up and down the tiers was described by its author. The way the matchmaker tries to create matches within each tier, and what it does when it (usually) fails due to lower player population, was also well described. I wrote a tutorial on it at the time.

While PSR pulls in players from Tiers, it was never explained how the 24 players are allocated to each team. I am suggesting that perhaps, beyond the 5 visible tiers, the 'individual player skill points' are still being used the tiers may be being used to allocate players between the two teams.

You seem to agree with this.

Wintersdark, on 08 October 2016 - 10:47 AM, said:

The matchmaker under PSR and Elo both has only two inputs it uses: The mech you're piloting, and your current score. It neither knows nor cares what your WLR is or what kind of streak you're on.

I already know this, and nothing I've said suggests otherwise.

IF individual skill tier points at a more granular level are being used to allocate players across the two teams, then that would explain how losing streaks could occur. Everyone in a match might be Tier 1 and 2, but some will be at a very high granular score. In addition, maxbar may be a hard numerical limit, or it could actually be open-ended for all we know.

If the matchmaker creates teams with a notional skill point seed number in mind, then you will get exactly what I stated - some players with very high individual skill points, counter-balanced by having low Tier 1 players allocated to their team. On the other side you might get 6 'a little higher than the seedpoint" players counterbalanced by 6 'a little lower than the seedbpoint" players.

This is in effect an attempt to create two teams of average skill. It fails because 3 high + 9 low isn't really the same as 6 a bit high +6 a bit low - or the many other possible combinations - when it comes to skill.

Wintersdark, on 08 October 2016 - 10:47 AM, said:

People liked to say the "matchmaker makes you lose if you win too often"

That could be shorthand for the process I described above. But I did not suggest that the matchmaker sets out to do anything other than create two teams within Tiers and 'escape valves' ... and then allocates those players to two teams.

I am saying that the effect of the way the matchmaker is that players get balanced out and in effect 'expected to carry'.

Wintersdark, on 08 October 2016 - 10:47 AM, said:

Karl Berg fully outlined how the matchmaker works.

He described how it worked under Elo. In fact, I am suggesting that the matchmaking code is doing what it always did, and the PSR system is just an extra gate before players are selected and then allocated to teams.

Wintersdark, on 08 October 2016 - 10:47 AM, said:

It's kind of narcissistic to imagine the matchmaker is setting you up to win or lose. It didn't under Elo, it doesn't now. It doesn't care about you. You're just one totally irrelevant data point in a sea of data points.

Are you a bit slow today? I have explained twice that I don't think this is about "me". It's offensive that you keep suggesting otherwise. I'd have to be a ******* idiot to think that.

For the third time: whatever is going on is going on with everyone. Bishop said he also noticed the same sinusoidal effect. There have been plenty of other people describing their long losing streaks.

Wintersdark, on 08 October 2016 - 10:47 AM, said:

This is true, but the problem is that you totally don't understand how the old system worked.

No I understood perfectly well. You apparently didn't understand what I was saying.

Wintersdark, on 08 October 2016 - 10:47 AM, said:

Why even bother coding more conditionals into the matchmaker?

But that's exactly what they did when they decided PSR Tiers would be an additional gating factor. And then they apparently added tonnage matching too, after Elo, IIRC.

Wintersdark, on 08 October 2016 - 10:47 AM, said:

included a "Screw Appogee" toggle. That's just ridiculous,

Yes, so ridiculous that I can't believe you would even suggest that I might think that.

[Zolaz]

When I get about 5 or 6 losses in a row, I take a break for an hour. Getting something to eat or doing something constructive while a different set of potatoes log on and are put on my team.

[Deathlike]

For the record, PSR is currently ONE rating for your character/profile. It doesn't matter what chassis you're running (whether it be Mist Lynx or Kodiak)... it's all the same.

That will almost always be problematic by design.

[Wintersdark]

Appogee, on 08 October 2016 - 12:36 PM, said:

You either don't know how the system worked, or misunderstood what I was saying. So don't describe 'reality' as 'ridiculous', kthx.

While I may well have misunderstood you, and if so I apologize, you did say:

Appogee, on 08 October 2016 - 09:17 AM, said:

Of course not. But I do suspect that, just like when Elo was the basis of matchmaking, there is some kind of weighting going on in the matchmaking that kicks in when you get too many matches ahead in wins.

Which was not the case then, nor now. Simply because, as I said, it only has as inputs rating, and mech. Nothing more. People imagined that to be the case, but it's not, for all the reasons I said in my post. Now, if you DON'T think there was some weighting going on based on wins, then I sincerely apologize, but I'm sure you can see how I thought that's what you were thinking?

Quote

In MWO, players had 4 Elo scores, one per weight class. The matchmaker would assemble teams to try and get a roughly equal average Elo on both side. Individual Elos would go after the match, based on which side won the match.

It would add players to a team, starting at the seed Elo and working outwards from it, take the average of that team then build another team in the same way. But yes, you had 4 (unrelated, weight class based scores); now you just have 1. That was one of the major points PGI made about the new system when it was implemented.

Quote

A team that won the match despite having a lower average Elo would have its members' Elos increased. A team that lost despite having a higher average Elo would have its members' Elos decreased.

That is how the system was implemented in MWO.

Yes, but that's not the matchmaker, and isn't important. Elo is well established, and the math behind it is quite clear. Glad we're on the same page though.

Quote

Movement up and down the tiers was described by its author. The way the matchmaker tries to create matches within each tier, and what it does when it (usually) fails due to lower player population, was also well described. I wrote a tutorial on it at the time.

Yup. But it's very clear looking at the system, and from Russ's comments about it, that PSR tiers are user-side only. It's simply a fixed number, and that number is represented by the bar. Russ discussed it in a town hall, when pointing out how close matches where in average rating vs. average rating (also a clear indication that the matchmaker(takes 2 inputs, outputs list to the dedicated server, nothing more) itself was completely unchanged.

Quote

While PSR pulls in players from Tiers, it was never explained how the 24 players are allocated to each team. I am suggesting that perhaps, beyond the 5 visible tiers, the 'individual player skill points' are still being used the tiers may be being used to allocate players between the two teams.

You seem to agree with this.

Yes. PSR = Pilot Skill Rating. This is still a fixed range. The only things that've changed between Elo and PSR is how scores change after a match, and that you've only got one rating instead of 4.

Quote

I already know this, and nothing I've said suggests otherwise.

Well, except from the notion that it weights things if you've won too many matches, re: quote above. See also: The overall thought of the thread, that there's something other than just bad luck to lose 13 consecutive matches. What would it be?

Quote

IF individual skill tier points at a more granular level are being used to allocate players across the two teams, then that would explain how losing streaks could occur. Everyone in a match might be Tier 1 and 2, but some will be at a very high granular score. In addition, maxbar may be a hard numerical limit, or it could actually be open-ended for all we know.

I said this. It doesn't even matter; because you know players can cover 3 tiers of rating range, you know players in a match will cover 1/3 of the total MWO player skill range. That's huge. That makes matches basically random.

Quote

If the matchmaker creates teams with a notional skill point seed number in mind, then you will get exactly what I stated - some players with very high individual skill points, counter-balanced by having low Tier 1 players allocated to their team. On the other side you might get 6 'a little higher than the seedpoint" players counterbalanced by 6 'a little lower than the seedbpoint" players.

Here's where you go wrong. You're thinking it's adding players to maintain an average. it's not doing that. It just takes all the players at the seed rating or as close to it as possible, calculates an average from that, then builds another team in the same manner at the same target average. The result is similar, but there are a couple important differences despite it seeming like it's about the same thing.

If it adds a higher rated player, it doesn't care if the next player is lower rated or higher rated as well; it's not trying to maintain an average. It's just grabbing the closest rated players, so the resulting team average can be really anything in there depending on what players were available. Then the second team is built the same way, seeded at the first teams' average. Less logic, really; it's simple. I've long thought it would be better to just grab the 24 closest players, then distribute them intelligently, but that doesn't happen (I know you're not suggesting it does; I'm just saying it should).

The downside to making one team then the other is that the first team is much more likely to be close together in ranking than the second team. It's a lot easier to just throw them together, though, I guess. With a higher population, this wouldn't be an issue.

Quote

This is in effect an attempt to create two teams of average skill. It fails because 3 high + 9 low isn't really the same as 6 a bit high +6 a bit low - or the many other possible combinations - when it comes to skill.

It's a bit worse than this, as when populations are lower it's possible that the second team is forced to build away from the first. Consider, a mid T1 based game:

The first team takes ALL the T1 players, then the second team has only lower ranked players to add.

Quote

That could be shorthand for the process I described above. But I did not suggest that the matchmaker sets out to do anything other than create two teams within Tiers and 'escape valves' ... and then allocates those players to two teams.

Fair enough.

Quote

I am saying that the effect of the way the matchmaker is that players get balanced out and in effect 'expected to carry'.

They aren't balanced out, though. In an ideal world with lots more players, it would just happen for sure. But we don't have enough. Like I said, I worked out the math. There's nowhere near enough players for that to happen reliably.

Quote

He described how it worked under Elo. In fact, I am suggesting that the matchmaking code is doing what it always did, and the PSR system is just an extra gate before players are selected and then allocated to teams.

Possible, but unlikely and pointless, because the with ONLY PSR (well, it's the same MM, but I'm sure we're on the same page here) with only one rating value and nothing else is still forced to spread valves in every single match. No mater of other "goals" the matchmaker has would have any effect at all. It's just making two teams from a VERY limited pool of players. Again, at average concurrent populations as per steam, you're looking at 50 players in total, with max release valves to build a team from. At that point, it doesn't even matter. It's wildly random.

Quote

But that's exactly what they did when they decided PSR Tiers would be an additional gating factor. And then they apparently added tonnage matching too, after Elo, IIRC.

Yes, so ridiculous that I can't believe you would even suggest that I might think that.

Tiers aren't a gating factor, not really.

As you suggest above and is fairly obvious, if they're just using your PSR (specific score, not tier), then you're looking at each tier being 1/5 of the total max possible score. Matches start within one tier, expand +/- a tier range per minute twice.

Note I said tier range. This works, whereas the "taken at face value using only tier" description given doesn't work, because if it allows +/- 2 TIERS, then a T3 match could see players from T1 to T5, but they also said you'd never face anyone more than 2 tiers away from you.

So, lets assume PSR is a number from 1 to 1000. 1-200 = T5, 201-400 = T4, 401-600 = T3, 601-800 = T2, and 801-1000 = T1.

Match seeds at 700 (middle of T2). Initially the MM, working exactly as it did before because it's unchanged, takes players from a tier range (200 points), so it's taking from 600-800 rated players. After a minute, another 200 is added to the spread (+/- 100), so 500-900 rated players. After a second minute, full three tier spread, 400-1000. The maximum gap between players is three tiers worth of rating, and it's granular too. You could never see players from T5 as a T1 player.

This fullfills both the description, with the assumption that specific tier numbers are just user facing fluff and granular skill ratings are used, AND allows the MM to be used exactly as it was before (as, specific numbers aside, that's how the Elo release valves worked too).



So, having gone in a big circle, and largely agreeing on the works, even Elo didn't cause players to have winning or losing streaks (*unless they played in the group queue a lot and ended up with an undeservedly high rating, as scores moved as a group not based on individual performance).

You COULD and CAN have big streaks, but the MM doesn't do that intentionally. It's simply an artifact of:

1) Low player counts - 50 players to choose from doesn't allow the MM to make good choices, particularly with weight class restrictions and such. Thus, ratings can be hugely far apart within a team, or between teams - not helped by making one team then another team.

2) A terribly designed PSR system that's mostly XP bar causing the players even within higher tiers to have wildly different actual skill levels. I'm not saying it's better or worse than Elo, though, because we can't really know thanks to point 1 above.

So, the same problems as Elo, because it's the same system aside from scores changing differently.

[LordKnightFandragon]

KMDD? What is this Demolition Man? Murder Death Kill!!! Murder Death Kill....I heard that like 1 time when my dad turned that movie on youtube and KMDD, thats literally all that I can figure that means...some Mechwarrior version of it.

[Wintersdark]

Zolaz, on 08 October 2016 - 01:37 PM, said:

When I get about 5 or 6 losses in a row, I take a break for an hour. Getting something to eat or doing something constructive while a different set of potatoes log on and are put on my team.

As an aside: This actually works for a reason.

As I said previously, there's basically 50 people available to match with you at any given time. If you requeue again and again immediately, you'll see lots of the same players repeatedly as there's so few to choose from. However, if you get "out of the loop", that is, the same pool that's getting out of a match and requeuing at the same time, you'll get a whole new set of players to play with, and that can make a big difference.

As Appogee said, if you're playing with a particularly wide spread of players, having a bunch of potatoes in the spread can really screw with matches, so mixing up the pool the MM is using when it's building your match helps to "shuffle the deck" if you will.

In the Elo days, you could just change weight class if you had a significant delta in Elo rating between classes, but that's not an option now. Just getting different players via waiting works wonders though. A couple particularly bad players (regardless of their rating) can sink a team easier than a good player can save it =)

[Wintersdark]

Deathlike, on 08 October 2016 - 01:57 PM, said:

For the record, PSR is currently ONE rating for your character/profile. It doesn't matter what chassis you're running (whether it be Mist Lynx or Kodiak)... it's all the same.

That will almost always be problematic by design.

Yeah, but it's not the worst of the issues with PSR.

Still, it doesn't really matter; the MM can't even make matches out of people from just one tier, so even if the system wasn't so stupid it'd still fail pretty much as much.

I figure - and this is just my opinion - that PGI knows this and designed PSR as a basic XP bar because the population issue leads to the MM not being able to do any better anyways. Just keeps newer players from being seal-clubbed, nothing more.

Unless the non-steam user count actually exceeds the (handily visible) steam user count, basically every match will span three tiers of rating, and the actual skill difference even within one tier is HUGE.

That's why I always maintain the MM doesn't make people win or lose... Even if it was trying to, that it can make matches at all is somewhat of a miracle. It can't do any more.

[MW222]

Zolaz, on 08 October 2016 - 01:37 PM, said:

When I get about 5 or 6 losses in a row, I take a break for an hour. Getting something to eat or doing something constructive while a different set of potatoes log on and are put on my team.

Keep your friends close, your enemys closer and cookies in arms reach!

[Deathlike]

Wintersdark, on 08 October 2016 - 02:52 PM, said:

I figure - and this is just my opinion - that PGI knows this and designed PSR as a basic XP bar because the population issue leads to the MM not being able to do any better anyways. Just keeps newer players from being seal-clubbed, nothing more.

Outside of the true seals, people that are "good enough to not totally suck" are the demographic that royally gets screwed here. These are the seals that get clubbed anyways.

[Widowmaker1981]

LordKnightFandragon, on 08 October 2016 - 02:40 PM, said:

KMDD? What is this Demolition Man? Murder Death Kill!!! Murder Death Kill....I heard that like 1 time when my dad turned that movie on youtube and KMDD, thats literally all that I can figure that means...some Mechwarrior version of it.

It means Kill Most Damage Dealt, you get them ingame.. Means an enemy mech died and you did the most damage to it (regardless of final killing blow)

[Wintersdark]

Deathlike, on 08 October 2016 - 03:27 PM, said:

Outside of the true seals, people that are "good enough to not totally suck" are the demographic that royally gets screwed here. These are the seals that get clubbed anyways.

yup, but ts impossible to do it better without hugely long waits.

Like I said, current average populations mean there's a pool of 50 players available to build a match within a 33% of the total pilot skill rating band at any given time. As you need 24 per match, and there's still at least loose weight class matching, you basically just get the people who are queued. The matchmaker can't choose which players are more appropriate, because it has so few options.

So PSR being junk is largely irrelevant. It's not the psr system screwing players, it's population. That's not fixable.

Of course, if we had a larger pop, then the PSR system flaws would become relevant.. but that's unlikely to happen. You'd need literall a full order of magnitude more players for the ranking system to matter much at all.

And whats amusing is the math to get those 50 players is extremely generous and assuming everyone is instantly queueing for new matches the moment they get out of a match, and that you're spending 1 minute searching for a match and 9 minutes playing the match/loading/"connecting" after.

In reality, it's likely a lot less.