If there was going to be a reset someday, it should be done with recent average match scores, and not a do-over with everyone put in tier 5. Because I doubt the playerbase would survive a few weeks of total stomps.
5
Will The Matchmaker Ever Work?
Started by Ruediger Steiner, Jan 12 2018 06:45 AM
44 replies to this topic
#22
-
-
- 2,199 posts
Member
Posted 13 January 2018 - 06:29 AM
From another thread. No, it won't start working, till bias towards increasing of rating won't be removed.
Jarl's list shows, that I should be in Tier 4:
My simulation shows, that on average I should have been having -0.63 PSR (in "Small change" units) per match:
Reality:
I'm in Tier 2 and keep climbing! So, PSR is just an XP bar!
Jarl's list shows, that I should be in Tier 4:
My simulation shows, that on average I should have been having -0.63 PSR (in "Small change" units) per match:
Reality:
I'm in Tier 2 and keep climbing! So, PSR is just an XP bar!
Edited by MrMadguy, 13 January 2018 - 06:36 AM.
#23
-
-
- 1,271 posts
Member
- LocationOntario, Canada
Posted 13 January 2018 - 06:35 AM
Back when I played World of Tanks, at a time when it had probably 10-20 times as many players as MWO, and a different and probably better MM algorithm... this sort of thing happened all the time, and WoT players have complained about it for years. And years. And still are.
Lopsided games happen. They happen even when the teams are part of competitive ladders where there is some equivalence of player experience, skill, and equipment. They happen in real life, between billion-dollar teams that are top professionals.
You can't expect an MM to fix that. The only thing you can do, and have control over, is how you play a match. If your goal at the end of a match is to say "I played a good round there" then you have control. If your goal is to say "I was on a good team and we won" then you are bound for disappointment.
PS: plus note, the match you used as example was not even all that lopsided. Roughly even match actually, given the damage spread on both sides. So the MM actually did the right job there, but 'felt' wrong to you anyway. That's an example of unrealistic expectations, not of bad MM.
Lopsided games happen. They happen even when the teams are part of competitive ladders where there is some equivalence of player experience, skill, and equipment. They happen in real life, between billion-dollar teams that are top professionals.
You can't expect an MM to fix that. The only thing you can do, and have control over, is how you play a match. If your goal at the end of a match is to say "I played a good round there" then you have control. If your goal is to say "I was on a good team and we won" then you are bound for disappointment.
PS: plus note, the match you used as example was not even all that lopsided. Roughly even match actually, given the damage spread on both sides. So the MM actually did the right job there, but 'felt' wrong to you anyway. That's an example of unrealistic expectations, not of bad MM.
Edited by MadBadger, 13 January 2018 - 06:40 AM.
#24
-
-
- 2,199 posts
Member
Posted 13 January 2018 - 06:49 AM
MadBadger, on 13 January 2018 - 06:35 AM, said:
Back when I played World of Tanks, at a time when it had probably 10-20 times as many players as MWO, and a different and probably better MM algorithm... this sort of thing happened all the time, and WoT players have complained about it for years. And years. And still are.
Lopsided games happen. They happen even when the teams are part of competitive ladders where there is some equivalence of player experience, skill, and equipment. They happen in real life, between billion-dollar teams that are top professionals.
You can't expect an MM to fix that. The only thing you can do, and have control over, is how you play a match. If your goal at the end of a match is to say "I played a good round there" then you have control. If your goal is to say "I was on a good team and we won" then you are bound for disappointment.
PS: plus note, the match you used as example was not even all that lopsided. Roughly even match actually, given the damage spread on both sides. So the MM actually did the right job there, but 'felt' wrong to you anyway. That's an example of unrealistic expectations, not of bad MM.
Lopsided games happen. They happen even when the teams are part of competitive ladders where there is some equivalence of player experience, skill, and equipment. They happen in real life, between billion-dollar teams that are top professionals.
You can't expect an MM to fix that. The only thing you can do, and have control over, is how you play a match. If your goal at the end of a match is to say "I played a good round there" then you have control. If your goal is to say "I was on a good team and we won" then you are bound for disappointment.
PS: plus note, the match you used as example was not even all that lopsided. Roughly even match actually, given the damage spread on both sides. So the MM actually did the right job there, but 'felt' wrong to you anyway. That's an example of unrealistic expectations, not of bad MM.
What I want to say about so called "snowball" argument, i.e. argument, that tries to convince us, that lopsided games happen due to some RNG factors, not due to initial imbalance - is that I'm experienced enough to distinguish matches, that we've lost due to our own mistakes, from games, that were lost due to difference in skill levels between two teams. It's not that hard. Noobs are easily distinguishable from skilled players.
#25
-
-
- Legendary Founder
- 319 posts
Member
- LocationSpain
Posted 13 January 2018 - 08:26 AM
Wishmast3r, on 12 January 2018 - 06:45 AM, said:
No. Cause we don´t have enough players. So matchmaking will probably never work, unless you want to wait 30 minutes for a QP match.
I just wanted to show you what´s wrong with MWO these days. And it will get worse.
I just wanted to show you what´s wrong with MWO these days. And it will get worse.
Dude, you haven't played other F2P games similar to MWO, have you?
In the Wargaming games, there is no matchmaking watsoever other than the type of ship or tank. No skill level or tier, so it is filled with players of completely random skill, and they have a lot more players than MWO.
At least here they try, but the faulty tier system and lack of players don't provide the best results.
#26
-
-
- The Hammer
- 1,212 posts
Member
- LocationPacifica Training School, waiting for BakPhar shares to rise
Posted 13 January 2018 - 09:25 AM
Novakaine, on 12 January 2018 - 11:03 AM, said:
Your typical 12 man quick play consists of these elements.
One Ace.
Two or three excellent experienced players.
Two or three average to below average players.
Several sub par players.
And the rest ......well you've seen them.
No amount of matchmaker can fix that.
Not since I've been around.
One Ace.
Two or three excellent experienced players.
Two or three average to below average players.
Several sub par players.
And the rest ......well you've seen them.
No amount of matchmaker can fix that.
Not since I've been around.
That's a good observation as lineups tend to shake out that way statistically, but it's dice roll every time because the matchmaker's being fed inaccurate data about player ability.
As @Vxheous suggests, 2 years after PSR's intro, we see the flaw in its design. Tier 1 — let alone Tier 2! — has wide variances in performance. You have world-class players. You have very talented and consistent players. You have players like me who statistically are better than 9/10 but not at the same skill level as leaderboard guys. You have group queue gravy train beneficiaries. And you have players who apparently have logged in enough to show that Paul Inouye's scenario where "eventually everyone will bubble up to Tier 2 or Tier 1" doesn't need special conditions to occur.
If PSR and Tier assignment were dead-on accurate, I'm not sure how to solve the problem of impractical wait times even in solo queue without intentional mixing/loose Tier valves, since distribution would be weighted far in favor of Tiers 3 and 4 (assuming Tier 5 has special, entry-level conditions).
But at least the matchmaker could with some assurance match player-to-player performance instead of relying on K/D or W/L averages that can be easily skewed by standout players.
So PGI might do something like this:
1. Reset PSR and assign all players to Tier 3.
2. Assess PSR increase or decrease roughly according to this (strictly using Solo Queue as an example):
- All 24 players are assessed as one group.
- This prevents awesome players who lose from being punished.
- This prevents not-so-good players who win from gaining any PSR rating.
- This prevents awesome players who lose from being punished.
- Score factors from objectives and victory are removed from PSR assessment.
- Again, victory shouldn't matter as much. If PSR replaces ELO, follow through all the way.
- Objectives, unfortunately, don't really reflect what PSR is trying to measure.
- Find a behavior-modeling solution to DC issues to prevent manipulation/screwing over.
- Again, victory shouldn't matter as much. If PSR replaces ELO, follow through all the way.
- Players receive the following increases or decreases:
- Top player, significant increase.
- 2nd-4th players, average of 67% of top player's increase.
- 5th-8th players, average of 33% of top player's increase.
- 9th-18th players, no increase. At all, otherwise PSR doesn't work.
- 19th-21st players, small decrease.
- 22nd-24th players, significant decrease.
- Top player, significant increase.
Outliers would generally be corrected. Tier 3 and 4 may a little fluctuation, as Tier 3 players picked up in a game with honest-to-goodness Tier 1 players get hammered down to 4, then shine when surrounded by actual 4s and rise back up to 3.
But again, the matchmaker would be getting better information about where someone belongs in a Tier.
At any rate, PSR needs a revisit. Matchmaker's relying on luck right now.
#27
-
-
- Liquid Metal
- 1,137 posts
Member
Posted 16 January 2018 - 07:07 AM
East Indy, on 13 January 2018 - 09:25 AM, said:
That's a good observation as lineups tend to shake out that way statistically, but it's dice roll every time because the matchmaker's being fed inaccurate data about player ability.
As @Vxheous suggests, 2 years after PSR's intro, we see the flaw in its design. Tier 1 — let alone Tier 2! — has wide variances in performance. You have world-class players. You have very talented and consistent players. You have players like me who statistically are better than 9/10 but not at the same skill level as leaderboard guys. You have group queue gravy train beneficiaries. And you have players who apparently have logged in enough to show that Paul Inouye's scenario where "eventually everyone will bubble up to Tier 2 or Tier 1" doesn't need special conditions to occur.
If PSR and Tier assignment were dead-on accurate, I'm not sure how to solve the problem of impractical wait times even in solo queue without intentional mixing/loose Tier valves, since distribution would be weighted far in favor of Tiers 3 and 4 (assuming Tier 5 has special, entry-level conditions).
But at least the matchmaker could with some assurance match player-to-player performance instead of relying on K/D or W/L averages that can be easily skewed by standout players.
So PGI might do something like this:
Outliers would generally be corrected. Tier 3 and 4 may a little fluctuation, as Tier 3 players picked up in a game with honest-to-goodness Tier 1 players get hammered down to 4, then shine when surrounded by actual 4s and rise back up to 3.
But again, the matchmaker would be getting better information about where someone belongs in a Tier.
At any rate, PSR needs a revisit. Matchmaker's relying on luck right now.
As @Vxheous suggests, 2 years after PSR's intro, we see the flaw in its design. Tier 1 — let alone Tier 2! — has wide variances in performance. You have world-class players. You have very talented and consistent players. You have players like me who statistically are better than 9/10 but not at the same skill level as leaderboard guys. You have group queue gravy train beneficiaries. And you have players who apparently have logged in enough to show that Paul Inouye's scenario where "eventually everyone will bubble up to Tier 2 or Tier 1" doesn't need special conditions to occur.
If PSR and Tier assignment were dead-on accurate, I'm not sure how to solve the problem of impractical wait times even in solo queue without intentional mixing/loose Tier valves, since distribution would be weighted far in favor of Tiers 3 and 4 (assuming Tier 5 has special, entry-level conditions).
But at least the matchmaker could with some assurance match player-to-player performance instead of relying on K/D or W/L averages that can be easily skewed by standout players.
So PGI might do something like this:
1. Reset PSR and assign all players to Tier 3.
2. Assess PSR increase or decrease roughly according to this (strictly using Solo Queue as an example):
- All 24 players are assessed as one group.
- This prevents awesome players who lose from being punished.
- This prevents not-so-good players who win from gaining any PSR rating.
- This prevents awesome players who lose from being punished.
- Score factors from objectives and victory are removed from PSR assessment.
- Again, victory shouldn't matter as much. If PSR replaces ELO, follow through all the way.
- Objectives, unfortunately, don't really reflect what PSR is trying to measure.
- Find a behavior-modeling solution to DC issues to prevent manipulation/screwing over.
- Again, victory shouldn't matter as much. If PSR replaces ELO, follow through all the way.
- Players receive the following increases or decreases:
- Top player, significant increase.
- 2nd-4th players, average of 67% of top player's increase.
- 5th-8th players, average of 33% of top player's increase.
- 9th-18th players, no increase. At all, otherwise PSR doesn't work.
- 19th-21st players, small decrease.
- 22nd-24th players, significant decrease.
- Top player, significant increase.
Outliers would generally be corrected. Tier 3 and 4 may a little fluctuation, as Tier 3 players picked up in a game with honest-to-goodness Tier 1 players get hammered down to 4, then shine when surrounded by actual 4s and rise back up to 3.
But again, the matchmaker would be getting better information about where someone belongs in a Tier.
At any rate, PSR needs a revisit. Matchmaker's relying on luck right now.
Your grading scheme...that seems like a pretty decent idea. I think you need to back it up with a number of some kind (or in some way reduce the rapidity of one's ranking moving around). This is to prevent natural randomness in player performance (sometimes you do well, somtimes you don't) from causing rapid changes in player tier rating, and also to prevent bad luck in matchmaking from knocking you out of a tier unjustly (what if the top .0001% of players in your tier get matched against you 3 games in a row?).
Other than that rather insightful post, I feel like this thread shouldn't exist any more. There is no convincing OP of anything.
Edited by Water Bear, 16 January 2018 - 07:09 AM.
#28
-
-
- Wrath
- 170 posts
Member
Posted 16 January 2018 - 09:25 AM
I've always thought that if we stripped away the basic assumption of encounter symmetry, the matchmaker could work much better. By this I mean: every engagement doesn't need to be 12-on-12. Hear me out.
Each time we queue-up, we are placing ourselves in a pool of players that desire a game. We assume that we will not have to wait "long" and that we will get a "balanced" match. Tonnage and weight-class have obviously not been good metrics for determining this as evidenced by the performance we have currently. Arguably, Tier/PSR also seems a poor determiner, especially at the Tier 1 where the skill disparity amongst members is quite noticeable.
Why not set up asymmetric matches; minimum 6 players, maximum 12 players on a side?
I propose the PUG (solo) matchmaker should, every 2 minutes, and perhaps using multiple simultaneous threads:
1. For the pool of requesting players, fetch the "Player Average KDR" and "Player Average MS" for each specific pilot in each specific mech they are currently piloting. This basic data already exists in the stat database that we all have access to. Call these values the "Player Metrics".
2. In the order each player entered the queue, build two teams with similar values of "Side Average KDR" and "Side Average MS"-- basically finding each side's average of the player averages. Call these values the "Side Metrics".
3. Continue to add players to the teams, in queue order, while maintaining this "Side Metric" balance, until both sides consist of 6 players minimum and 12 players maximum.
4. Make a final pass of the requesting pool, to see if adding the next player to the side with fewer players would push the "Side Metric" balance out of whack (i.e., off by more than 10% let's say). If not, add him.
5. Check total tonnage of each side. If these totals are within a certain ratio considered "normal" (i.e., maybe a 20% difference between each side), there would be no "match bonus". If this ratio is greater than "normal", the match would be considered a "Challenge Match" with a "match bonus" multiplying the lesser side's winnings (CBills and Earned XP).
6. Start the match.
In my opinion, this approach would lead to faster queues, more balanced matches, and add an element of excitement. The "lesser team" could pull off an upset and reap vast rewards. Using this method, the teams would be inherently be balanced by skill. It would only be numbers on each side and tonnage that would be variant. But with the "match bonus" feature, that variance could be rewarded.
Thoughts?
Each time we queue-up, we are placing ourselves in a pool of players that desire a game. We assume that we will not have to wait "long" and that we will get a "balanced" match. Tonnage and weight-class have obviously not been good metrics for determining this as evidenced by the performance we have currently. Arguably, Tier/PSR also seems a poor determiner, especially at the Tier 1 where the skill disparity amongst members is quite noticeable.
Why not set up asymmetric matches; minimum 6 players, maximum 12 players on a side?
I propose the PUG (solo) matchmaker should, every 2 minutes, and perhaps using multiple simultaneous threads:
1. For the pool of requesting players, fetch the "Player Average KDR" and "Player Average MS" for each specific pilot in each specific mech they are currently piloting. This basic data already exists in the stat database that we all have access to. Call these values the "Player Metrics".
2. In the order each player entered the queue, build two teams with similar values of "Side Average KDR" and "Side Average MS"-- basically finding each side's average of the player averages. Call these values the "Side Metrics".
3. Continue to add players to the teams, in queue order, while maintaining this "Side Metric" balance, until both sides consist of 6 players minimum and 12 players maximum.
4. Make a final pass of the requesting pool, to see if adding the next player to the side with fewer players would push the "Side Metric" balance out of whack (i.e., off by more than 10% let's say). If not, add him.
5. Check total tonnage of each side. If these totals are within a certain ratio considered "normal" (i.e., maybe a 20% difference between each side), there would be no "match bonus". If this ratio is greater than "normal", the match would be considered a "Challenge Match" with a "match bonus" multiplying the lesser side's winnings (CBills and Earned XP).
6. Start the match.
In my opinion, this approach would lead to faster queues, more balanced matches, and add an element of excitement. The "lesser team" could pull off an upset and reap vast rewards. Using this method, the teams would be inherently be balanced by skill. It would only be numbers on each side and tonnage that would be variant. But with the "match bonus" feature, that variance could be rewarded.
Thoughts?
#29
-
-
- Liquid Metal
- 1,137 posts
Member
Posted 16 January 2018 - 09:53 AM
Vyx, on 16 January 2018 - 09:25 AM, said:
Asymmetry is good
Problem is the snowball effect people keep bringing up, which I can only subscribe to. My girlfriend, who has watched me play the game for about 2 hours in her entire life, remarked the other night that: "When 3 more people are dead on one team, that team seems to lose most of the time."
#30
-
-
- The DeathRain
- 2,726 posts
Member
- LocationUSA
Posted 16 January 2018 - 10:51 AM
Well friend I don't think MWO has a matchmaker in any queues anymore the populace is to low so tier4 players now play tier3,tier2,and tier1 players every battle it makes finding a balanced match harder now but it lessens the wait time for a battle to start.
Edited by Ibrandul Mike, 16 January 2018 - 04:41 PM.
Quote Cleanup
#31
Posted 16 January 2018 - 12:08 PM
Look, the matchmaker can't spoon-feed wins to you. Nor can it stop players from deciding to play wonky and/or joke builds.
We get it, you are incapable of performing a proper AAR on a match, so you blame the matchmaker. Got it. I'm telling you that it isn't the matchmaker's fault. It just plain, straight-up IS NOT the matchmaker's fault.
And your whines about player population? Look, at off-peak hours this game likely still has 1000+ players online, how many is enough for you to accept that player performance is about the players and not the matchmaker? 2000? 10,000? How many? Because the logical failures in your argument are so glaring as to lead me to believe that we could have 50,000 and it wouldn't satisfy you.
We get it, you are incapable of performing a proper AAR on a match, so you blame the matchmaker. Got it. I'm telling you that it isn't the matchmaker's fault. It just plain, straight-up IS NOT the matchmaker's fault.
And your whines about player population? Look, at off-peak hours this game likely still has 1000+ players online, how many is enough for you to accept that player performance is about the players and not the matchmaker? 2000? 10,000? How many? Because the logical failures in your argument are so glaring as to lead me to believe that we could have 50,000 and it wouldn't satisfy you.
Edited by Ibrandul Mike, 16 January 2018 - 04:42 PM.
Quote cleanup
#32
-
-
- The DeathRain
- 2,726 posts
Member
- LocationUSA
Posted 16 January 2018 - 12:13 PM
Escef, on 16 January 2018 - 12:08 PM, said:
Look, the matchmaker can't spoon-feed wins to you. Nor can it stop players from deciding to play wonky and/or joke builds.
We get it, you are incapable of performing a proper AAR on a match, so you blame the matchmaker. Got it. I'm telling you that it isn't the matchmaker's fault. It just plain, straight-up IS NOT the matchmaker's fault.
And your whines about player population? Look, at off-peak hours this game likely still has 1000+ players online, how many is enough for you to accept that player performance is about the players and not the matchmaker? 2000? 10,000? How many? Because the logical failures in your argument are so glaring as to lead me to believe that we could have 50,000 and it wouldn't satisfy you.
The thing is there (IS NO MATCHMAKER) to blame anything on and at this point in MWO a matchmaker would only make matters worse and longer wait times without a few more players like 100,000 plus more.
#33
Posted 16 January 2018 - 12:20 PM
KingCobra, on 16 January 2018 - 12:13 PM, said:
The thing is there (IS NO MATCHMAKER) to blame anything on and at this point in MWO a matchmaker would only make matters worse and longer wait times without a few more players like 100,000 plus more.
If that is true than why is it in tier 2 I literally NEVER see the new player TK accounts? I have seen no such animal since I was in tier 3.
#34
-
-
- The DeathRain
- 2,726 posts
Member
- LocationUSA
Posted 16 January 2018 - 01:12 PM
Escef, on 16 January 2018 - 12:20 PM, said:
If that is true than why is it in tier 2 I literally NEVER see the new player TK accounts? I have seen no such animal since I was in tier 3.
I don't know I play a tier 1 to tier 4 player all the time no mater if its QP or FP so I just assume they removed the MM 6 months ago because of low population and wait times for both game modes.
Edited by KingCobra, 16 January 2018 - 01:12 PM.
#35
Posted 16 January 2018 - 01:23 PM
KingCobra, on 16 January 2018 - 01:12 PM, said:
I don't know I play a tier 1 to tier 4 player all the time no mater if its QP or FP so I just assume they removed the MM 6 months ago because of low population and wait times for both game modes.
So, you don't see tier 5 players in your matches, so you assume there's no matchmaker? If there's no matchmaker, than why are you seeing no tier 5 players? Are you just assuming they don't exist?
Edited by Ibrandul Mike, 16 January 2018 - 04:44 PM.
CoC violation removed
#36
-
-
- The DeathRain
- 2,726 posts
Member
- LocationUSA
Posted 16 January 2018 - 01:35 PM
Escef, on 16 January 2018 - 01:23 PM, said:
So, you don't see tier 5 players in your matches, so you assume there's no matchmaker? If there's no matchmaker, than why are you seeing no tier 5 players? Are you just assuming they don't exist?
I assume there there playing with everyone else tier5 players I know this because I did test it out this week as a new player .
Edited by Ibrandul Mike, 16 January 2018 - 04:44 PM.
Quote Cleanup
#37
Posted 16 January 2018 - 01:36 PM
KingCobra, on 16 January 2018 - 01:35 PM, said:
I assume there there playing with everyone else tier5 I know this because I did test it out this week as a new player .
If the tier 5 players are only being grouped with other tier 5s, than how can there NOT be a matchmaker? You aren't making any damned sense, you're logic is not consistent with itself.
#38
-
-
- The DeathRain
- 2,726 posts
Member
- LocationUSA
Posted 16 January 2018 - 01:40 PM
Escef, on 16 January 2018 - 01:36 PM, said:
If the tier 5 players are only being grouped with other tier 5s, than how can there NOT be a matchmaker? You aren't making any damned sense, you're logic is not consistent with itself.
That is my point I was not just with Tier5 players even on a new account in fact you can test it out yourself I was only tier5 for 1 hour playing with others up to tier 1 then tier4 playing against other players up to tier1 so it kind of proves my point friend.
#40
-
-
- The DeathRain
- 2,726 posts
Member
- LocationUSA
Posted 16 January 2018 - 02:02 PM
Escef, on 16 January 2018 - 01:45 PM, said:
Your point is that since you saw evidence of matchmaking that there is no matchmaker?
That's it friend I found NO evidence of a matchmaker being used in gameplay even on a brand new account.
Edited by KingCobra, 16 January 2018 - 02:02 PM.
1 user(s) are reading this topic
0 members, 1 guests, 0 anonymous users
