QUESTION TO DEVS - Determining Rank

(K1X455) #21

@Xenithos What I’m saying is that the three sources of information can be used retroactively to curtail player behaviour. It’s actually enforced in some way, but not specifically against match throwers and leavers.

I myself have been punished for being AFK; and quite possibly been reported as match throwing. It was because I’ve been waiting in excess of 30 minutes (I can’t actually remember how long I was waiting during NA’s peak hours) and I decided to stand up and visit the loo. I can’t remember exactly if I stood up after a accepted the match and had some sort of glitch then the match suddenly started without my knowledge.

There are occasions as well when the match started and I only knew it had gone when I heard someone screaming for medic.

(Xenithos) #22

Thanks for explaining your thoughts to me K1X.
I just realized what your name stood for by the way. Good name.

#23

I doubt the devs will tell us the exact parameters, it might lead to abuse.

(bgyoshi) #24

@Kirays said:
I doubt the devs will tell us the exact parameters, it might lead to abuse.

I have a hard time believing that any further abuse to gain ranks could happen. You’d have to guarantee matching against other players to ensure matches can be thrown for rank grinding and that’d be pretty tough.

(Smooth) #25

@bgyoshi said:
So what are the actual metric considerations around this? Is there a specific ratio of these things that go into determining a rank move?

And what does “individual match performance” mean? Is that some combination of values on the end game screen like accuracy, headshot %, revives, etc? Or is it more like your exp per minute?

Hey,

So I won’t be able to go too detailed with all of this, as we don’t want to fully expose how everything as it could lead a ‘gaming’ of the system, when what we simply want are players doing their best to win as a team.

Regarding the different ‘metrics’ I do want to quickly dive into a couple of factors which have a large bearing on the rating changes:

  • The difference in team strengths at the start of each match
  • The size of the parties on each team (bigger parties are treated as higher skilled)

For example, if a Party of 5 x Gold Recruits came up against a Party 5 x Silver Officers, the system would expect the Gold Recruits to have a ~75% chance of winning the match, so from the match the Gold Recruits would gain about half as much from a win as they would lose if they lost. The opposite is also true where the Silver Officers stand to gain much more than they would lose from a loss.

If that same Party of 5 x Gold Recruits came up against 5 solo Gold Officers, the system would ‘boost’ the Party of 5 (due to the inherent advantage of being a pre-made party) and in this case consider them have a 50/50 chance of winning/losing. So the potential gain/loss of Rank would be the same for both sides.

When it comes to the breakdown of change after that, it can vary significantly based on the different values of the factors involved but here’s a rough idea:

Individual game won/lost
This makes up the ‘base’ of the change as explained above, let’s say this is +20% of a Rank, to give the rest of this a point of reference.

Recent Win/loss ratio
When using that +20% figure, if you have a 100% win ratio this can increase the rate of change by up to around 1.5x so as high as +30% of a Rank, however typically this will add very little as most players quickly get closer to a 50/50 WLR. The opposite is also true where a 100% loss streak could take a -20% rank to a -30% rank.

Individual match performance
While I won’t dive into what this actually uses (since it would definitely cause players to alter how they play during a game, rather than just trying to win) this can increase the rate of change by up to 1.35x, so taking that +20% up to +27%. Again, the opposite is also true and it can take a -20% to a -27%.

Previous season rank
This is only ever factored in to what rating you have for your first placement match of every season, beyond that it’s no longer used in calculations.

With the way this all works, the combination a perfect winning streak and an great individual match performance has the ability to give you up to around ~1.9x the ‘base’ change, taking that +20% to +38%.

Since we’re always striving to be a team-based game and to put team-play first, it’s the ‘base’ change that has the largest impact of all the factors, and that’s decided mostly by the difference in skill of the two teams but also is heavily impacted by the size of the different parties involved.

I hope this helps explain things some :slight_smile:

(bgyoshi) #26

So then it appears relatively evenly considered when it comes to a rank move, between your current/ongoing record and your individual performance.

If I’m understanding you right you’re saying:

! A. Some undisclosed base amount of ‘K value’ for a win/loss that moves your rank as the individual game won or lost
! B. Party size alters the K value for each team depending on team size and matching at the start of the game
! C. The altered K value is shifted per person based on their recent W/L record over X amount of past games
! D. The K value is further altered with a relatively equal weight based on the individual’s in game performance as well.
!
! If that’s true then it seems to be that a player can perform strongly in a losing game to mitigate their inevitable losses. So for example, with easy numbers
!
! K value is 8, we’ll say. +8 points on a win, -8 points on a loss
!
! A Gold 5 man party is paired against a 5 man solo (all Silver) so…
!
! Gold 5 man: Win +4 instead of +8, Loss -12 instead of -8
! Silver 5 man solo: Win +12, Loss -4
!
! A player on the Silver 5 man solo team with a recent 100% loss streak would see K value on a loss go from -4 to -6, or with a 100% win streak K value would go +12 to +10 instead. But the game tries to keep players in 50/50 situations so for most people, K value wouldn’t change.
!
! On top of that, if that player has a strong in game performance on their losing streak, that -6 becomes -4 again, and on a win streak that +10 goes back to +12.

So there’s incentive to perform strongly no matter what, since you’ll mitigate your losses and boost your wins.

Here’s what I suggest…

! Under the same scenarios, if the player performs strongly on their losing streak, they’d see that -6 become a +2. If they performed weakly on their win streak, that +10 becomes +4. The actual numbers themselves aren’t important, what’s important is that the player GAINS rank even if they lose… albeit, it’s a very small gain.
!
! Now if for example it were a 5 man solo filled with Golds, then that -6 would more than likely start as a -8 instead, and their strong performance would make it a -2.
!
! The game would weight the player performance more than the current win/loss ratio in such a way that a strictly unfair match would still award a strong player with a net gain in rank, otherwise losses would simply be mitigated as usual.
!
! This would be strong against abuse (for example, people intentionally quitting to alter the K value and boost their teammates) since 5 man premades get more weight than solos. And in mostly evenly matched games (the majority?) the K value disparity would be too much to overcome by performance alone. But when you’re that high gold on a team of silvers against 4 golds and a cobalt, and 2 people on your team leave, sticking it out would mean you don’t dunk down to silver again, and you’d more or less stay where you are.

Thanks for the insight and thank you for the response!

(ThePigVomit) #27

Thank you for the feedback.

(K1X455) #28

@Smooth said:
Since we’re always striving to be a team-based game and to put team-play first, it’s the ‘base’ change that has the largest impact of all the factors, and that’s decided mostly by the difference in skill of the two teams but also is heavily impacted by the size of the different parties involved.

I hope this helps explain things some :slight_smile:

It’s a real insightful explanation, however, I still find it totally flawed because a very critical factor is not put in consideration. Your explanations confirm one of the reasons why players resort to smurfing accounts to boost their buddies and sometimes, fine-tune their hacks. To illustrate how flawed it is, move all match servers to Wellington NZ, Honolulu HW, and Moscow RU, and you’ll get a clear picture of what I mean without much deliberation… and that’s just the first.

hxLbGcw.png749x677

If I took a screenshot for every exception that crops up based on your explanation, you’ll be appalled with impression it will make on a new player.

Also, in the period of gameplay where one player vanishes for any reason (AFK, technical network issues, etc), how is that factored in the result? Is there some sort of metric which measures performance during a 5 v 5 and 5-x vs 5-y time of play? In over 1100 ranked matches, I have won only 2x shorthanded (4v5, 1 recent, match thrower left by the name of Ymmib). I guess real time computation is tough, eh?

(bgyoshi) #29

@K1X455 No name and shame dawg c’mon you know better

I’m willing to bet the base K Value does shift in situations where players leave, but I can’t imagine it moves by much.

(K1X455) #30

@bgyoshi This thread reminded me to take another picture when I had the opportunity. Wanna see it?