Jump to content
Sign in to follow this  
Qbit

Ranking System Discussion

Recommended Posts

I agree with everything, except for dropping tiers. If you are at (for ease, lets call it XP) 0XP in Diamond 1 for 5+ straight losses, you're probably doing something wrong. Getting from P6 to D1 should remain easy (though I would really like to see promotion matches, similar to League), but going from D1 to P6 should be more difficult. If they introduce a buffer of ~50XP or so at the tier threshold (for deranking only) it could solve the problem. 

 

To clarify, once a player reaches 0XP at D1/P1/G1/S1/B1 (or Onyx 1500), there should be an invisible 50XP buffer. Any wins would immediately disregard this buffer, meaning if you lost 4 games at Onyx1500 (so your invisible rank was Onyx1460 and one more loss would result in you deranking to Diamond 6 with 0XP), but you win your 5th, your rank would be Onyx1500+ XP from winning... the -40XP from losing the previous 4 is NULL. 

 

Also, for the INITIAL placement of a player, the system shouldn't place a player in the highest tier. I could live with the highest placement being Diamond, but Onyx should be something that is achieved only if you absolutely PROVE you deserve it. However, after a rank reset, the NEXT seasons placements ABSOLUTELY should take into consideration your previous seasons rank (see LoL). So, if your first season you grind into Onyx-1xxx and the ranks reset, your placement matches should account for your previous season, and not waste time matching you with low-tiered players UNLESS you perform extremely poorly in your matches (indicating that you were likely boosted/carried). 

 

Yeah dropping tiers as I said has Pros and Cons. 

 

In my opinion the two best solutions would be

  • Allow players to lost their placed Division (if you get placed in Onyx, you can drop to Diamond for Losing, but if you hit Onyx by grinding from Diamond 1, you get to stay there)
  • Add Promo Matches with extremely tight skill matching

We also don't know if Seasons will be a hard reset (complete MMR/CSR wipe) or soft reset (Placements will start you at your last seasons rating and demote/promote you accordingly). It would be really helpful if @@Sal1ent or @@Deez could comment on this. I think (based on our population and current ranking system) that a Soft Reset would be very beneficial towards weening people out of misplaced Divisions.

Share this post


Link to post

Everyone is so quick to claim this system is better than 1-50.

 

All I know is in h3(outside of accounts where people intentionally tried to mess up the system) generals were better than brigs, brigs were better than colonels, etc. I could go into social and predict the scores of games based on the opponents ranks like clockwork.

 

In this game ranks are all over the place. Onyx are worse than diamond players at an alarming rate. Some platinums shit all over diamond kids. You never know what you're going to get. I can search onyx swat(the equivalent of searching 50 high in h3) and get kids who are amazing one game and kids who can hardly walk and aim at the same time the next. Ive never played a game where kids the same rank as me can be so good or so bad in comparsion to me.

 

How is a ranking system that cannot accurately rank people good?

 

The only thing this system has over 1-50 is the end game champions division. Which is cool. But it could easily be added to a 1-50 system as well.

  • Upvote (+1) 5

Share this post


Link to post

Everyone is so quick to claim this system is better than 1-50.

 

All I know is in h3(outside of accounts where people intentionally tried to mess up the system) generals were better than brigs, brigs were better than colonels, etc. I could go into social and predict the scores of games based on the opponents ranks like clockwork.

 

In this game ranks are all over the place. Onyx are worse than diamond players at an alarming rate. Some platinums shit all over diamond kids. You never know what you're going to get. I can search onyx swat(the equivalent of searching 50 high in h3) and get kids who are amazing one game and kids who can hardly walk and aim at the same time the next. Ive never played a game where kids the same rank as me can be so good or so bad in comparsion to me.

 

How is a ranking system that cannot accurately rank people good?

 

The only thing this system has over 1-50 is the end game champions division. Which is cool. But it could easily be added to a 1-50 system as well.

Because an inherent trait of an ELO system is this initial randomness. As it sorts everyone out, it will become more consistent than any other system in any previous Halo game. 

 

Starcraft and League of Legends both use their own variants of an ELO system. Give it time and stop complaining only because it is different than what you're used to. For PC gamers this is basically the norm (because it shits on everything else).

  • Upvote (+1) 4

Share this post


Link to post

Because an inherent trait of an ELO system is this initial randomness. As it sorts everyone out, it will become more consistent than any other system in any previous Halo game.

 

Starcraft and League of Legends both use their own variants of an ELO system. Give it time and stop complaining only because it is different than what you're used to. For PC gamers this is basically the norm (because it shits on everything else).

How will it sort itself out when people can't rank down?

  • Downvote (-1) 1

Share this post


Link to post

Everyone is so quick to claim this system is better than 1-50.

 

All I know is in h3(outside of accounts where people intentionally tried to mess up the system) generals were better than brigs, brigs were better than colonels, etc. I could go into social and predict the scores of games based on the opponents ranks like clockwork.

 

In this game ranks are all over the place. Onyx are worse than diamond players at an alarming rate. Some platinums shit all over diamond kids. You never know what you're going to get. I can search onyx swat(the equivalent of searching 50 high in h3) and get kids who are amazing one game and kids who can hardly walk and aim at the same time the next. Ive never played a game where kids the same rank as me can be so good or so bad in comparsion to me.

 

How is a ranking system that cannot accurately rank people good?

 

The only thing this system has over 1-50 is the end game champions division. Which is cool. But it could easily be added to a 1-50 system as well.

Why couldn't they implement something like this in Halo 5?  When a player reaches level 50 there's a division above that (champion of course)  It's a grind for sure, but I get a lot more even matches playing Halo 3 than I do for this game.  I feel like I'm playing Halo 4 again with no skill ranked system.  You never know if you're going to get a good game or not.

  • Upvote (+1) 1

Share this post


Link to post

How will it sort itself out when people can't rank down?

Because the metal represents your peak for that season

 

The actual matchmaking doesn't look at anything but the elo number hiding in the background

 

These division systems are all smoke and mirrors to make bad players feel like going from Bronze 3 to Bronze 2 or whatever is an accomplishment

 

That's why the highest ranks just give you the damn elo because anyone that good wants to know what their real rank is

 

In League it is possible to drop a division but only if you monstrously screw up, as in you screw up so badly you very likely bought your account and are simply incapable of winning any games at all at that elo for something like 20+ games

Starcraft's system works a little different on how it promotes and demotes, it moves you up a rank once it's impossible for you to match opponents lower than that rank, and moves you down when it's impossible for you to match opponents of that rank

  • Upvote (+1) 1

Share this post


Link to post

Because their MMR will continue to drop. 

how long is this supposed to take? I've seen kids in plat/diamond 1 and onyx 1500 with hundreds of games played who are still getting thrown into games out of thier league.

Share this post


Link to post

how long is this supposed to take? I've seen kids in plat/diamond 1 and onyx 1500 with hundreds of games played who are still getting thrown into games out of thier league.

 

Is this independent of the actual matchmaking system that seems to place emphasis on speed rather than anything else? The issue I face is that some of my friends played their placement matches, they could barely play the game at all. Now that they are competent (again), we steamroll through all the kids in their ELO (I'm usually the highest). Other than the system trollololololing me last night (25+ wins and ~5 losses and no rank increase), I haven't had any games where I've thought the system unfairly matched me. Last night, for example, aPZesus and I were both Onyx 1600+, and the other two players in our party were Diamond/Plat and Diamond/Diamond (depends which point of the night). We consistently got matched against parties with similar rank makeup's and steamrolled them (there were actually several good/close games as well). 

 

So, I guess the issue might not start to sort itself out until after the rank reset, because people have had a chance to play and learn the game... so the good players won't be placed as low, and the bad players wont get as lucky and place as high. I suppose this is why League forces you to play until you reach level 30 (XP based) before you are even allowed to play ranked in the first place. Makes sure you've had enough time to learn the game before it allows you to subject yourself to the more serious side of the game. Counter-Strike has implemented a similar function, albeit a much lower requirement. Maybe 343i could consider something similar? Arena playlists aren't unlocked until you reach Spartan Level 20 or something -- would force familiarity with the game. 

  • Upvote (+1) 1

Share this post


Link to post

The OVERWHELMING majority of knowledgeable Halo players are for 100% w/l ranking. There is a reason why you are an overall negatively rated poster.

Argumentum ad populum, another classic logical fallacy. Just because the majority thinks something, it doesn't mean they're right. The Monty Hall Problem was a simple logical puzzle posed in 1975 and a correct solution was submitted by 1 person. Over 10,000 people, including 1000 people with PhD's, wrote to the magazine which published the solution saying that the person was wrong. People negging me on this forum means absolutely nothing, and can actually speak more about them than me.

 

This is a terrible philosophy. There are so many factors behind determining a person's skill, and if a Platinum team beats a Diamond team, it shouldn't be a strict 1-1 system. Platinum deserves a major boost in CSR and Diamond deserves to be punished.

A zero-sum 1-1 W/L system is game-theoretically the best and fairest system. For starters, from the point of view of a ranking system, terms like "Platinum" and "Diamond" players are meaningless, all that matters are the underlying numbers. Secondly, in a game with a healthy population, players/teams with a large disparity in ranking should not be being matched together in the first place, that's the point of a raking system; evenly matching games. You should be gradually facing better and better players until you've earned the right to play against those even better players. Furthermore, a players current rank is not always representative of their true skill. A players rank reflects their skill only once their rank has stabilized and starts oscillating slightly about a certain value.

 

For that reason, having a system with a variable rank gain/loss based on the relative rank differences between the players is both unnecessary and undesirable. It's undesirable because it overly punishes players who are paired up with players that have a lower rank, but have higher real skill than them. The higher rank players stand to gain less and lose more and the lower ranked players stand to gain more and lose less, which makes it always advantageous to be lower ranked when playing another person. Remember Halo 3? You always wanted to be paired up with higher ranked players because it greatly accelerated your rank-up, and you never wanted to play against players of lower rank because if you lost, the results could be devastating. In zero-sum, everything naturally and fairly sorts itself out without having shitty situations like that.

 

Here's another example. Let's say I'm the best player in a room of 100 players. I want to get the highest rank I can while playing the least amount of games possible. In that case, I might as well wait until the other players have played each other a lot and inflated their ranks. Once they've done that, I can jump in and boost up really fast because I can take advantage of that flaw in the ranking system. In zero-sum, if I want to catch up to the highest-ranked player, I always have to put in as much effort as he/she did to get there and there's no advantage in making alts or trying to exploit the system.

 

p7pE0ui.png

 

Take the graph above as an example of a typical above-average team. Said team starts off and wins their first 4 games before registering a loss. They always play against other teams who have the same or closest W/L spread as themselves, so each time they win it gets harder. Eventually, a point is reached where they play against teams who are essentially at the exact same level of skill as them, so their expected win/loss rate at that point becomes 50:50 and their position on the ladder stabilizes. If the team later improves, the point at which their rank will oscillate about will increase. Each and every game is worth the exact same amount because you're always playing the team closest to yours on the ladder. There is no fairer system than that.

 

That particular system, it should be noted, applies to organized teams and clans. For randoms vs randoms, the same concept applies but it uses a fine-grained individual K/D/A system as outlined in OP, where each individual on each team will have a very similar rank to each other and the average individual skill level for each team will be as close as possible. Again, for RvR matchmaking, there is no fairer system than a zero-sum K/D/A one; it directly matches up the nearest 8 skilled players on the ladder in 4v4 games.

Share this post


Link to post

The trueskill system is mathematically superior to ELO.  Halo3 used the trueskill system to match players and it worked well (besides people purposely tanking accounts to game the system).  The only problem was the arbitrariness of assigning 1-50 to the hidden trueskill ratings, if they ranked the top 10k players by actual ranking (#1-10,000) it would have been much better than just giving everyone above a certain trueskill rating the same rank.  

 

To prevent the gaming of the ranking system by people tanking accounts, imo, the way to do it is use the SC2 system where you receive a rank for each team combination you play with.  

 

For example if I decide to play with A, B, and C.  We play our placement matches and all receive the same rank, say bronze.

 

Then if I play with A, B, and D.  After placement matches we all receive same rank, say diamond.  So I now have both a diamond and bronze 4v4 ranking albeit with 2 different teams that are unaffected by one another.

 

To do this effectively they need a 2v2, 3v3, and 4v4 team playlist and a 4v4 loner playlist (party size = 1).

  • Upvote (+1) 1

Share this post


Link to post

The trueskill system is mathematically superior to ELO.  Halo3 used the trueskill system to match players and it worked well (besides people purposely tanking accounts to game the system).  The only problem was the arbitrariness of assigning 1-50 to the hidden trueskill ratings, if they ranked the top 10k players by actual ranking (#1-10,000) it would have been much better than just giving everyone above a certain trueskill rating the same rank.  

 

To prevent the gaming of the ranking system by people tanking accounts, imo, the way to do it is use the SC2 system where you receive a rank for each team combination you play with.  

 

For example if I decide to play with A, B, and C.  We play our placement matches and all receive the same rank, say bronze.

 

Then if I play with A, B, and D.  After placement matches we all receive same rank, say diamond.  So I now have both a diamond and bronze 4v4 ranking albeit with 2 different teams that are unaffected by one another.

 

To do this effectively they need a 2v2, 3v3, and 4v4 team playlist and a 4v4 loner playlist (party size = 1).

+1 was for the bolded part.

 

TrueSkill has a multitude of problems and is far from ideal and therefore shouldn't be used in any future Halo games because better alternatives are available, such as Zero-Sum W/L and K/D/A. Zero-Sum doesn't facilitate boosting or encourage alts, it always matches the closest ranked available players/teams on the ladder and the rank/gain loss at stake is always the same for each match. It's simple, completely fair and can't be effectively exploited.

 

As for your SC2 system example, clans achieve the same thing as that but are much more flexible. If you want to play with players A, B and C, you can make a clan consisting of those players and the clan ranking will change with each game win/loss. However, if you want to add a player, D, to the clan, you can do so easily without having to endure a new team combination rank, and that player can play with any combination of other players in the clan to rank it up. That also simplifies matchmaking as you can play under the same clan name in different lists like 2v2, 4v4 etc.

Share this post


Link to post

explain them.

I've already touched on it, see paragraph 3 of post #85. To add to that, using "hidden variables" like sigma and mu is not a good way to go about things because every game permanently affects those variables which get harder and harder to change and they change the actual rate at which you can rank up and down, which ends up not basing things purely on W/L and makes things murky. There are instances where players might win 6 games in a row and gain 1 rank, then lose 1 game and rank straight down again, ad nauseam.

Share this post


Link to post

Argumentum ad populum, another classic logical fallacy. Just because the majority thinks something, it doesn't mean they're right. The Monty Hall Problem was a simple logical puzzle posed in 1975 and a correct solution was submitted by 1 person. Over 10,000 people, including 1000 people with PhD's, wrote to the magazine which published the solution saying that the person was wrong. People negging me on this forum means absolutely nothing, and can actually speak more about them than me.

 

A zero-sum 1-1 W/L system is game-theoretically the best and fairest system. For starters, from the point of view of a ranking system, terms like "Platinum" and "Diamond" players are meaningless, all that matters are the underlying numbers. Secondly, in a game with a healthy population, players/teams with a large disparity in ranking should not be being matched together in the first place, that's the point of a raking system; evenly matching games. You should be gradually facing better and better players until you've earned the right to play against those even better players. Furthermore, a players current rank is not always representative of their

Yawn

 

Just because you think something is right does not make it so.

 

You cannot rank people individually in a team game. Not unless professional talent assessors are going to watch and analzye every single game played by every single player.

 

You just can't do it based on stats alone.

 

Which is why pure w/l works the best.

  • Upvote (+1) 1

Share this post


Link to post

Everyone is so quick to claim this system is better than 1-50.

 

All I know is in h3(outside of accounts where people intentionally tried to mess up the system) generals were better than brigs, brigs were better than colonels, etc. I could go into social and predict the scores of games based on the opponents ranks like clockwork.

 

In this game ranks are all over the place. Onyx are worse than diamond players at an alarming rate. Some platinums shit all over diamond kids. You never know what you're going to get. I can search onyx swat(the equivalent of searching 50 high in h3) and get kids who are amazing one game and kids who can hardly walk and aim at the same time the next. Ive never played a game where kids the same rank as me can be so good or so bad in comparsion to me.

 

How is a ranking system that cannot accurately rank people good?

 

The only thing this system has over 1-50 is the end game champions division. Which is cool. But it could easily be added to a 1-50 system as well.

This.
  • Upvote (+1) 1

Share this post


Link to post

 

 

 

And yet my experience with the system thus far is a joke. With the 1-50 systems I feel like a majority of my games at 50 were closer and consistent than they are with this current one. In this I'll sometimes play champions, mostly play Onyx decent amount of diamonds and sometimes plats (why and how?!) Now that's just going by CSR. The matches themselves fluctuate WAY too much, sometimes I feel like I'm playing against FB -- you can smell the sweat through your monitor when it's a team against randoms. Then the very next game, I'm playing against a bunch of people with cerebral palsy and dementia getting running riots and what not, how is this at all a good ranking system? Might as well be playing social because you never know what you're in for.

 

Lets not forget the amazing system that gives Spartanman 148378283848 an Onyx in TA and has a 1.7 KDA and consistantly goes 5-16. I'm not even exaggerating in the slightest.

  • Upvote (+1) 1

Share this post


Link to post

Yawn

 

Just because you think something is right does not make it so.

I don't just think it, I argue it. I actually explain myself and list sound points and examples in favor of my position, you do not.

You cannot rank people individually in a team game. Not unless professional talent assessors are going to watch and analzye every single game played by every single player. You just can't do it based on stats alone.

Do you ever notice while writing your posts that they contain zero substance? "You cannot rank people individually in a team game" "You just can't do it based on stats alone." You hardly ever provide any sound arguments for why anything you say is true.

Share this post


Link to post

anyone who supports Halo 3s ranking system can GTFO IMO. That ranking system purposefully made it more difficult to rank up the more games you played. 

 

any system that introduces that as a variable is instantly invalid IMO, because it then makes it unrealistically difficult for someone to rank up, no matter what skill level, if they get stuck there for a while. Meaning youre actively punishing improvement. 

  • Upvote (+1) 1

Share this post


Link to post

anyone who supports Halo 3s ranking system can GTFO IMO. That ranking system purposefully made it more difficult to rank up the more games you played. 

 

any system that introduces that as a variable is instantly invalid IMO, because it then makes it unrealistically difficult for someone to rank up, no matter what skill level, if they get stuck there for a while. Meaning youre actively punishing improvement. 

If you really improved you would climb out of that hole and rank up. Its as simple as that

Share this post


Link to post

If you really improved you would climb out of that hole and rank up. Its as simple as that

 

A ranking system shouldnt actively make it more difficult to rank up simply because you play the game a lot. 

Share this post


Link to post

A ranking system shouldnt actively make it more difficult to rank up simply because you play the game a lot. 

It got more difficult to rank up because you lost a lot, not just playing.

 

Its better than these systems which basically rewards the people who play a ton.

Share this post


Link to post

The rank floor honestly makes no sense to me.  People should be ranked by how much they are winning, and if you are constantly losing you should drop down.

  • Upvote (+1) 1

Share this post


Link to post

It got more difficult to rank up because you lost a lot, not just playing.

 

Its better than these systems which basically rewards the people who play a ton.

and that is a system that doesnt make sense. 

 

If you lose a lot you should simply rank down, not just dig a hole that not only deranks you, but makes ranking back up MORE difficult. At that point you end up with people who can theoretically be playing at a skill level above their current rank, simply because they reached a skill ceiling, lost a bunch, then improved, and now cant rank back up because the system wont allow them too. 

  • Upvote (+1) 1

Share this post


Link to post

and that is a system that doesnt make sense. 

 

If you lose a lot you should simply rank down, not just dig a hole that not only deranks you, but makes ranking back up MORE difficult. At that point you end up with people who can theoretically be playing at a skill level above their current rank, simply because they reached a skill ceiling, lost a bunch, then improved, and now cant rank back up because the system wont allow them too. 

Stop making sense.

 

I find it funny that this is really a debate. If a zero-sum W/L system worked, or if a individual stat based system worked, they would have been implemented. However, we've instead created incredibly complex algorithms that take into account a multitude of different variables to produce a rank consistently and fairly... but that's not good enough for the Halo kids with their (hopefully) passing grades in pre-calc. 

  • Upvote (+1) 1
  • Downvote (-1) 2

Share this post


Link to post

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
Sign in to follow this  

×

Important Information

By using this site, you agree to our Terms of Use.