43 replies to this topic

[Nightmare1]

ScarecrowES, on 15 August 2015 - 06:11 PM, said:

It's actually very easy to balance skill. If you look at games like the Battlefield series... everything you do that might result in scoring is recorded and weighted, and then considered comparatively to derive an actual skill score. MWO actually tracks a LOT of actions. Look at all the things you can get scoring for in a match. Games like Battlefield know personal skill factors like what your average score per minute is... how much damage you're likely to do per minute per class,overall accuracy, whatever. It also knows how well you play with others, tracking things like assisted damage, how often you remain with your unit, how often you perform team-assistive actions, and so on.

MWO knows these things too. But it doesn't care. It cares about whether you won or lost. That's it. If MWO actually put use to all the things it tracked, weighed how important those factors are to your team's overall success, and then produced an overall score for that... THAT would be an actual skill score. If it actually adjusted your score based on the actual results of the match you're in, by weighing your performance against that of your team AND the enemy team, it'd be about as accurate as it could be.

It'd also have the benefit of helping balance match rewards to correspond with the skill system. That would let rewards better reflect the sorts of actions players engage in to win a match. We could get away from a reward system based almost entirely on damage and kills, which would be nice.

I didn't play Battlefield past the second game, so I don't know about its system. What you describe doesn't sound so bad though.

The big problem I see at the moment stems from these two factors:

1) There is a big elo/skill disparity in the players slotted for each match
2) The MM, like you said, tries to divvy up teams based on pilot win/loss records in an effort to achieve a 1:1 ratio

The second bullet is the most damaging since it is a method that forces you to experience streaks of wins and losses that you normally wouldn't. In short, it circles around elo or skill matching.

Personally, I like the idea of gear + skill as the method for balancing teams. Even though skill is the better measure of those two, there are times when good pilots just want to run troll builds and can't carry like they normally would. If I want to run an AC/20 Raven for laughs, then I should be able to do that without dinging my team too badly. With the current system, the MM is blind to the Mech value, forcing pilots who are running sub-par Mechs to carry the team, even on matches where the MM has decided that pilot's team should win. That is just a recipe for disaster.

[Koniks]

Short answer: yes, Elo is in the game. There's no problem with it. The problem is with the matchmaking algorithm.

[El Bandito]

Xetelian, on 15 August 2015 - 01:28 PM, said:

Most matches are blowouts. 90% of my pug queue matches or more end in 12 - 3 or less.

Actually having a close match is a rarity for the vast majority.

If you're someone who is having close games more often than blowouts and steamrolls you're either a liar or in a very strange elo.

PGI really should put up the paperdolls of everyone in the end game screenshots. Perhaps then people will realize that many games were actually a lot closer than they thought. I had won many matches with only a small alpha worth of health remaining in my torsi and had to let my healthier teammates to tank the shots.

Edited by El Bandito, 15 August 2015 - 07:00 PM.

[Mechwarrior Buddah]

El Bandito, on 15 August 2015 - 06:51 PM, said:

PGI really should put up the paperdolls of everyone in the end game screenshots. Perhaps then people will realize that many games were actually a lot closer than they thought. I had won many matches with only a small alpha worth of health remaining in my torsi and had to let my healthier teammates to tank the shots.

Thatd be great for when Im still at 97% lol

[El Bandito]

Mechwarrior Buddah, on 15 August 2015 - 07:08 PM, said:

Thatd be great for when Im still at 97% lol

97% in an Assault or Heavy will cause some discontents for sure.

Edited by El Bandito, 15 August 2015 - 07:11 PM.

[Mechwarrior Buddah]

El Bandito, on 15 August 2015 - 07:10 PM, said:

97% in an Assault or Heavy will cause some discontents for sure.

They get pissy now lol They get less pissy when I put up 4-6 kills and 600-1400 damage

[Mechwarrior Buddah]

I dunno; ppl down on the solo que but I get worse games in the group que than I go in the solo lol at least in the solo Im not going against 8-11 man teams

[fat4eyes]

El Bandito, on 15 August 2015 - 06:51 PM, said:

PGI really should put up the paperdolls of everyone in the end game screenshots. Perhaps then people will realize that many games were actually a lot closer than they thought. I had won many matches with only a small alpha worth of health remaining in my torsi and had to let my healthier teammates to tank the shots.

This. After you've played a while you learn to go to the second line once you're heavily damaged so you can still contribute firepower for as long as possible.

[ScarecrowES]

Nightmare1, on 15 August 2015 - 06:40 PM, said:

I didn't play Battlefield past the second game, so I don't know about its system. What you describe doesn't sound so bad though.

The big problem I see at the moment stems from these two factors:

1) There is a big elo/skill disparity in the players slotted for each match
2) The MM, like you said, tries to divvy up teams based on pilot win/loss records in an effort to achieve a 1:1 ratio

The second bullet is the most damaging since it is a method that forces you to experience streaks of wins and losses that you normally wouldn't. In short, it circles around elo or skill matching.

Personally, I like the idea of gear + skill as the method for balancing teams. Even though skill is the better measure of those two, there are times when good pilots just want to run troll builds and can't carry like they normally would. If I want to run an AC/20 Raven for laughs, then I should be able to do that without dinging my team too badly. With the current system, the MM is blind to the Mech value, forcing pilots who are running sub-par Mechs to carry the team, even on matches where the MM has decided that pilot's team should win. That is just a recipe for disaster.

Matchmaking should take both pilot skill and mech "value" into account for matchmaking purposes. Currently, MWO makes no judgement about the "quality" of the mech a player is bringing into a match, as there is no "value" assigned to the composition of a mech. All the game knows is what class you're bringing, and what Elo rating you're at. In this case, your choice of mech class is the single greatest factor weighed in determining team composition. It matters significantly more than any other factor considered for matchmaking. Once you understand how matchmaking works, you understand that it's more or less random.

Once you move over to a more stat-driven system like the Battlefield series, awesome things start to happen. You would then have a skill system that actually reflects you as a player. This is based not only on hard statistics like score per minute, kill/death ratio, accuracy, win/loss ration, etc (whose exact numbers can be compared to the distributions of other players to determine a relative rating for use in skill score)... but also on relative performance figures from match to match. You'd finally be able to have skill ratings based on a player's impact on a match, and not merely the match result.

You can prioritize whether individual performance or team play is more important to your skill system, and judge match results accordingly. For instance, your team loses a match, but you put in a heroic effort that nearly drove your team to a win. You can weigh the results from both teams and find that, even though you lost, the results were actually pretty close. Then you can weight your team's performance internally, and see that you significantly outperformed the team average. For the purposes of adjusting individual skill scores, one of the two results (match or internal) will be the baseline, and the other will be a modifier, depending on which performance (team or individual) you prioritize.

If you prioritize individual performance, your skill would go up, but not a lot. You performed well, giving you a high individual skill increase, but your team still lost, which would pull back your skill increase by a given amount. If you prioritize team play, your skill rating will go down, but only a little. You lost, but the loss was not mainly your fault. On the other hand, for the lowest rated player on your team the results would be different. If you prioritized individual performance, his skill would go down. Not only did that player fail to perform at the average level of the team, but the team also lost. He would go significantly negative. If you prioritize team performance, he'd still go negative, as the reduction in skill from the loss would be further modified by his poor individual performance. Either way, he sees probably about the same loss in skill rating.

The winning team would see similar expected results depending on priority. Players that contribute more see better results. The player on a winning team that does not contribute much to that win isn't going to get rewarded as much as the player who contributes a lot. What's great about this is that your skill change reflects your actual actions in the match, and thus over time will better reflect your actual capability in play better than simple hard statistics can. Hard statistics matter, but only in context.

[Nightmare1]

ScarecrowES, on 16 August 2015 - 10:48 AM, said:

Matchmaking should take both pilot skill and mech "value" into account for matchmaking purposes. Currently, MWO makes no judgement about the "quality" of the mech a player is bringing into a match, as there is no "value" assigned to the composition of a mech. All the game knows is what class you're bringing, and what Elo rating you're at. In this case, your choice of mech class is the single greatest factor weighed in determining team composition. It matters significantly more than any other factor considered for matchmaking. Once you understand how matchmaking works, you understand that it's more or less random.

Once you move over to a more stat-driven system like the Battlefield series, awesome things start to happen. You would then have a skill system that actually reflects you as a player. This is based not only on hard statistics like score per minute, kill/death ratio, accuracy, win/loss ration, etc (whose exact numbers can be compared to the distributions of other players to determine a relative rating for use in skill score)... but also on relative performance figures from match to match. You'd finally be able to have skill ratings based on a player's impact on a match, and not merely the match result.

You can prioritize whether individual performance or team play is more important to your skill system, and judge match results accordingly. For instance, your team loses a match, but you put in a heroic effort that nearly drove your team to a win. You can weigh the results from both teams and find that, even though you lost, the results were actually pretty close. Then you can weight your team's performance internally, and see that you significantly outperformed the team average. For the purposes of adjusting individual skill scores, one of the two results (match or internal) will be the baseline, and the other will be a modifier, depending on which performance (team or individual) you prioritize.

If you prioritize individual performance, your skill would go up, but not a lot. You performed well, giving you a high individual skill increase, but your team still lost, which would pull back your skill increase by a given amount. If you prioritize team play, your skill rating will go down, but only a little. You lost, but the loss was not mainly your fault. On the other hand, for the lowest rated player on your team the results would be different. If you prioritized individual performance, his skill would go down. Not only did that player fail to perform at the average level of the team, but the team also lost. He would go significantly negative. If you prioritize team performance, he'd still go negative, as the reduction in skill from the loss would be further modified by his poor individual performance. Either way, he sees probably about the same loss in skill rating.

The winning team would see similar expected results depending on priority. Players that contribute more see better results. The player on a winning team that does not contribute much to that win isn't going to get rewarded as much as the player who contributes a lot. What's great about this is that your skill change reflects your actual actions in the match, and thus over time will better reflect your actual capability in play better than simple hard statistics can. Hard statistics matter, but only in context.

Yeah, that's kind of what I was trying to describe using Ghost Recon: Phantoms' Athena Ranking system. It does that sort of thing fairly well.

[ScarecrowES]

Wow... I totally called the new skill system.

[Deathlike]

ScarecrowES, on 17 August 2015 - 01:44 PM, said:

Wow... I totally called the new skill system.

It's going to be bad.. for LRM users.

[ScarecrowES]

Not necessarily. Typically LRMS users can get a lot of assists, components, and targeting bonuses. It might actually help them a bit now that we're not concentrated totally on kills. We'll have to see what the score weighting is like. At least the system now exists. This opens the door to fundamentally changing how matches are scored.

[Deathlike]

ScarecrowES, on 17 August 2015 - 01:50 PM, said:

Not necessarily. Typically LRMS users can get a lot of assists, components, and targeting bonuses. It might actually help them a bit now that we're not concentrated totally on kills. We'll have to see what the score weighting is like. At least the system now exists. This opens the door to fundamentally changing how matches are scored.

I don't think you understand.

At higher levels of play, LRMs are not used.

So, when you inflate your score through LRMs (sure you'll be getting wins for a bit), but you'll reach a peak where using LRMs are a hindrance.. there will be a flood of QQ that will occur as a result.

[Chemie]

ScarecrowES, on 15 August 2015 - 03:56 PM, said:

Russ and the PGI team have gone on record and said quite emphatically that Elo just doesn't work for MWO. In fact, Elo's system was never intended to judge the skill of individual players on a team. It's not actually capable of doing that. Russ said they're dropping Elo and going to a proper skill-based ranking system sometime in the "near" future.

Until then, please try to understand that, as implemented in MWO, for players who largely play public solo queue, their Elo scores are effectively random. Further, Matchmaking only tries to make a team's cumulative score roughly equal to the other team's cumulative Elo score. Thus, it's possible to have players with very high Elo and very low Elo on the same team. As a result, it's possible to have one team consist of 4 great players and 8 terrible players, with the other team consisting of 12 average or better players. Guess which on is going to win? So yeah, effectively even team make-up by balanced Elo scores is a totally random thing.

Actually, they went on record many times saying how wonderful ELO was and the community told them it would be random for solo pugs. NOW they say it sucks, after 2 years.

[LORD ORION]

Ding Dong ELO is now officially dead in favor of a 5 tier PSR. YAY!

Edited by LORD ORION, 17 August 2015 - 02:01 PM.

[Deathlike]

Chemie, on 17 August 2015 - 01:56 PM, said:

Actually, they went on record many times saying how wonderful ELO was and the community told them it would be random for solo pugs. NOW they say it sucks, after 2 years.

Working as intended™.

[stjobe]

ScarecrowES, on 17 August 2015 - 01:44 PM, said:

Wow... I totally called the new skill system.

Almost - it doesn't seem to take which 'mech you're dropping with into account, just player skill (read: Match score mostly). Seems like a major mistake to me if that is so.

I hope they'll eventually use their new 'mech rating system (the one they're working on for the re-Quirkening) as an additional input into the ranking and matchmaking systems.

Edit: If someone missed it, the new system is described in the patch notes for tomorrow.

Edited by stjobe, 17 August 2015 - 02:08 PM.

[Mechwarrior Buddah]

Yeah I was about to say "guess not" lol

[ScarecrowES]

stjobe, on 17 August 2015 - 02:05 PM, said:

Almost - it doesn't seem to take which 'mech you're dropping with into account, just player skill (read: Match score mostly). Seems like a major mistake to me if that is so.

I hope they'll eventually use their new 'mech rating system (the one they're working on for the re-Quirkening) as an additional input into the ranking and matchmaking systems.

Edit: If someone missed it, the new system is described in the patch notes for tomorrow.

The skill system is only half of matchmaker. The game still has to match mech composition too. Right now that's just a simple weight class judgement. Eventually, once mechs are actually objectively balanced under the proposed balancing pass, most mechs within a class should be fairly close in ability. Or at least the game will know the "value" of the mech you're bringing for the sake of team balance.

With a REAL skill system and a REAL mech value system, matchmaking should be about perfect. If, by that point, we don't have to wait 10 minutes for the system to find enough players and mechs close enough to actually play together.