Which Rank X System was better? S2 or S3

[Supa Fern]

I have seen 2 tweets about people saying we should go back to S2 system of Rank X.
What is everyone's opinion on this?

But keep in mind Both still calculate points using Glicko2. And if you want to experience the S2 point system in Splatoon 3, Play anarchy open with friends, challenge or Pro Splatfest Power.

 

[Trent]

Splatoon 0 back in 2007 nailed it imo

 

[missingno]

The problem with S3's system is that it's kind of not using Glicko-2, the clamped minimum gain/loss frequently overrides what Glicko-2 should output. If you're going to use a tried-and-true mathematical formula to power your ranking system, you need to actually use that formula and not mess with it. Your attempt at a homebrew system will never improve on it, you will only break it. Why use this formula at all if you apparently don't trust its results?

 

[RubberCF]

2 by far. in 2, it just showed you your actual glicko power, no minimum or maximum changes. That way it consistently gave you less for winning against teams you were expected to win against and more if you werent expected to win. I think the best defense for s3 system is that it feels more consistently rewarding, but id argue no one should treat x rank as some sort of reward, all it i means is what skill level of players the game will try to put you against, thats it. if you only get a couple of points for a win, thats because the win doesnt give the game any reason to think you should be put against better players.

 

[Loctos]

S2 and it's not even close (Nobody should be able to get 1st with a 1000 xp lead in a decent ranking system)

 

[zyf_]

s2 but without the region locking that's the only good part of s3 -- the numbers work out so much better when you can play against the MUCH more densely populated japanese ranked spread

 

[Supa Fern]

The problem with S3's system is that it's kind of not using Glicko-2, the clamped minimum gain/loss frequently overrides what Glicko-2 should output. If you're going to use a tried-and-true mathematical formula to power your ranking system, you need to actually use that formula and not mess with it. Your attempt at a homebrew system will never improve on it, you will only break it. Why use this formula at all if you apparently don't trust its results?

i myself have been studying it and getting familiar with it way more than a normal human basing should, which is still not wrong to say S3 uses Glicko2 for each single MATCH.
But since I don’t want to scare people posting the formula and explaining the terms, I am pointing out the difference happens at end as the players finishes the series; You perfectly said with the result combination value getting clamped.
and this modification method has not a different name like Glicko3 or Joe method

 

[Joy]

i myself have been studying it and getting familiar with it way more than a normal human basing should, which is still not wrong to say S3 uses Glicko2 for each single MATCH.
But since I don’t want to scare people posting the formula and explaining the terms, I am pointing out the difference happens at end as the players finishes the series; You perfectly said with the result combination value getting clamped.
and this modification method has not a different name like Glicko3 or Joe method

We don’t actually know if it calculates XP updates using a clamped glicko-2, we know it uses glicko-2 as a factor but not how it’s using it under the hood.

But no this wouldn’t be an evolution of glicko: clamping results with arbitrary, fixed values undermines the Bayesian base of glicko systems. It’s simply a glicko-based system with changes so people feel better about their results, a priori prediction be damned. Splitting the player base between +- 2k also breaks the way glicko works, all for the illusion of fairness. These are all changes players might agree with but they can’t, for whatever reason, understand that if matchmaking quality has suffered significantly it is on the back of these “good” changes.

S2 is a much better system for ranking players, but was also a very disliked system because it has no sense of progression.

 

[Supa Fern]

We don’t actually know if it calculates XP updates using a clamped glicko-2, we know it uses glicko-2 as a factor but not how it’s using it under the hood.

But no this wouldn’t be an evolution of glicko: clamping results with arbitrary, fixed values undermines the Bayesian base of glicko systems. It’s simply a glicko-based system with changes so people feel better about their results, a priori prediction be damned. Splitting the player base between +- 2k also breaks the way glicko works, all for the illusion of fairness. These are all changes players might agree with but they can’t, for whatever reason, understand that if matchmaking quality has suffered significantly it is on the back of these “good” changes.

S2 is a much better system for ranking players, but was also a very disliked system because it has no sense of progression.

I would argue it is not a great system for ranking since it can not do that properly when the game wont connect to all possible players around the world.
Back in S2, given how the match making works, it creates poles of players that have very heterogenous skill levels and a more stable win loss amount of points. Mostly in JP you could feel the sense of progression much better with less -100 and +0 happening compared to NA.

But although i have no certain way to prove it, i dont think we can disprove it is still glicko2



int MatchesWon = 0
int MatchesLost = 0

Int AccumulatePower = 0

While(MatchesWon<3 or MatchesLost<3){

_______ // Splatoon 2 power of Glicko2 when a match finishes
_______Accumulate Power += Played match result

_______If match == win
____________MatchesWon +=1
_______else
____________MatchesLost += 1
}

if MatchesWin == 3
_______if Accumulate Power<10
____________FinalPower += 10
_______else
____________FinalPower += Accumulate Power

if matchesLost ==3
_______if Accumulate Power>10
____________FinalPower += -10
_______else
____________FinalPower += Accumulate Power

 

[Supa Fern]

Now my arguments for S3>S2

I think we can agree given the main issue is the inflation of points, given the player that just reached pass 4000 losing around 400 points after losing as series.

The theoretical System of Splatoon 2 Ranking system is great but in practice it is bad for any player outside the main conglomerate of players.

Everything in Splatoon 2 used Glicko-2, Isolating a region only moment are JP splatfests, during those you can see JP had always higher peak power. That happened not only because they are good, but because they have much more of a bigger active player base than other regions. The JP only Splatfests show how high they can go, and when you simulate a larger player base, the bell curve stretches further to the right(the percentage of players on that specific splatfest power); Statistical behavior

The 2 following statements are True, JP has the larger amount of good players and the large amount of bad players.
Check Gem's video to reinforce my argument on squid School
Stop acting like Japanese players are unbeatable, please - YouTube

Splatoon sadly gives priority to find a match rather than find 8 players with similar skill level, and for players outside of Japan, that can become an issue since they have less players and thus climbing the X rank power becomes harder, they have less player so the points given at the end become more volatile(-100 and +1). People were not familiar with Glicko2 that much at the time but the solution wasnt to kick players, but let more in so that extreme discrepancy between values could be smaller, again all of this is a Statistical analysis.

Now for the main reason i prefer S3 over S2, Glicko2 is a system that only cares for the winning team, since it was design with a 1v1 scenario in mind, even if you played perfectly you can still lose the match and can get -100 for something that was out of your control. Glicko2 is an adaptation of a 1v1 system into a 4v4 system. So S3 made the series approach, the final sum of the matches you played will be how many points you will get with the correction that if you won 3, it must be positive. But if you lost 3 it will take points away. This is much better to keep the player motivated to play the game and it sounds more fair since the bad performance of your teammates can be mitigated.

Yes it did create the first inflation of the game with a minimum of 10 points always being awarded. But now different from S2, climbing to get higher rank X once you are on the top like the person reaching pass 4000 becomes more motivating for the player to continue and reduces the need for a lobby with just as much of a high power so they can get more points. They would be gaining 0.1 points under S2 system and would drop the game much earlier.

 

[Joy]

The criticism of glicko-2 for 4v4 is more about the speed of convergence. Being a Bayesian system, it already naturally incorporates uncertainty. It can definitely still be used for 4v4 with minor modification, it just throws away a lot of potentially useful information with teammate data. It simply makes the assumption that over a long period of time, all potential teammate issues will cancel out, leaving your influence as the sole factor.

I’m glad you prefer the S3 system but it’s absolutely not something you can generalize. S3 is designed for player feelings over anything else, and it’s clear most people are somewhat unhappy with the matchmaking and XP system.

As far as not being a great ranking because not all players connect, I mean I don’t know what you’re addressing (if anything). XP as a system is separate from divisions and matchmaking, they all work in tandem to create a ranked experience. XP’s only responsibility is to generate ordering for all participating players in the player pool.

 

[Trent]

ranks been solved

 

[Supa Fern]

Factors Glicko2 doesnt include in the calculation
-Specials used
-enemy splatted
-objective participation
-turf painted
-number of times splatted
-objective score
-Amount of time taken to KO

It only cares for the winner of the match, the biggest criticism I spent reading on twitter when players decide to expose about their teammates, and it is the same now in S3. (Challenges have to be using the Glicko2 since even with 3 gold ribbons, you can get vey low scores of +1)

"I lost X amount of points because <insert weapon> keep feeding while i got <this amount of kills/objective, etc>". Most common tantrums on SoloQ matches on twitter. Never have i heard about conversion bein an issue i would love to hear more of your theory. Since it becomes hard to measure it if those are under the hood. So it becomes as much of a theory as if S3 is still using Glicko2 or not at this point, right?
Not arguing Glicko2 can not be used on 4v4. I am saying the modifications to S3 are closer to an a ideal world in which the amount of points is gained or lost are better proportional to your effort you give to the match compared to S2. It is an attempt to include as you mentioned "a lot of potentially useful information" thrown away.

If you want to argue feelings then I can just counter it by saying that a lot of players were also some what unhappy with S2. If both games hurts people's feelings then both are bad and there is no better system among them; Feelings are a terrible metric.

In an ideal world where all the people close to my power are able to queue up with me in a Solo Match, with powers above or bellow me being the less likely the larger the gap. However what happens is that the game prioritizes quick match making over Power, meaning i will paired up more likely to the people closer to my rank around where I live. But in S2 the higher the power i get compared to the players around me, the more difficult i can gain power since the amount of players is lower the higher you go to the pyramid.
Meaning a single loss becomes destructive to the players points, they should have been paired with the region with similar power to that player, but the game hardly ever let's them.

Anyone that had a JP copy of the game could have access to JP Exclusive fests, i played almost all living in South America(the opposite side of Japan) While it took no more than a rotation to get to king/queen. JP was so long because i couldnt find matches, and the people i played with were almost all the same. There were around 1 million JP players and i couldnt find matches fast? Clearly being in a different region locks you out of playing with other people. And even with Optic and LAN internet it didnt matter.

I meant to address " not all players connect " because it explains why 2400 power in S2 for a JP vs NA. One is much harder to reach compared to the other. The Average JP player can get 2400 much easier than the average 2400 NA.
The height of a pyramid is only as tall as its base is big.

You can think in reverse too move a JP player to NA, in order for them to get to the 3000 they were getting in JP, it would take much longer

Same logic for Splatfest is applied to league in S2.

The criticism of glicko-2 for 4v4 is more about the speed of convergence. Being a Bayesian system, it already naturally incorporates uncertainty. It can definitely still be used for 4v4 with minor modification, it just throws away a lot of potentially useful information with teammate data. It simply makes the assumption that over a long period of time, all potential teammate issues will cancel out, leaving your influence as the sole factor.

I’m glad you prefer the S3 system but it’s absolutely not something you can generalize. S3 is designed for player feelings over anything else, and it’s clear most people are somewhat unhappy with the matchmaking and XP system.

As far as not being a great ranking because not all players connect, I mean I don’t know what you’re addressing (if anything). XP as a system is separate from divisions and matchmaking, they all work in tandem to create a ranked experience. XP’s only responsibility is to generate ordering for all participating players in the player pool.

 

[missingno]

The 'problem' with S2 using pure unmodified Glicko-2 is that players don't like the honest truth about how good they really are. It feels frustrating to the player when your wins are worth +2 and your losses are worth -30, but that's happening for a reason. It means you're getting matched with players you were expected to beat, so if you do lose your rating needs to be corrected accordingly.

But number go down makes you feel bad! We can't have that!

So S3 now lies to make you feel better. You went 3-2 in the set? You're gaining points no matter who you were pubstomping, even when the truth is that it should've been a net negative. And as a result, there's runaway inflation for the top players at the far end of the bell curve who rarely, if ever, lose points. Conversely, players on the low end can get stuck in a defeatist spiral even when they sometimes get a few wins that should save them. Your rank is now a lot more dependent on the luck of the draw in who you get matched with, whether you got a set you'll go 3-2 or 2-3 in.

It's too common in modern games to have systems that lie to make players feel good. It boosts engagement. Keeps you coming back to farm those dopamine points, number go up. Frustrated players are at risk of leaving, so we lie to minimize frustration.

But lying isn't good, and there are consequences for that lie. Ratings are not accurate now, and if the ratings aren't accurate, then the matchmaking isn't either. If you're not happy with the state of S3's matchmaking, you have to realize this is a big part of why.

So what do you want out of a rating system? Do you want the game to lie so you'll feel better, or do you want the truth no matter how uncomfortable it is?

 

[Joy]

So S3 now lies to make you feel better. You went 3-2 in the set? You're gaining points no matter who you were pubstomping, even when the truth is that it should've been a net negative.

To add to this excellent post, one of the most common complaints I saw before the xp update was going 3-0, 3-0, 3-0, then 2-3 and having a net loss of points. People called this broken, when in reality it’s expected if unintuitive. If you keep showing the system that it’s wrong about where you’ve been placed, it reduces the certainty and as a result increases the possible gains/losses. The latest XP update mitigates this even further by lying even more to players, but the result is the runaway point inflation.

 

[zyf_]

The 'problem' with S2 using pure unmodified Glicko-2 is that players don't like the honest truth about how good they really are. It feels frustrating to the player when your wins are worth +2 and your losses are worth -30, but that's happening for a reason. It means you're getting matched with players you were expected to beat, so if you do lose your rating needs to be corrected accordingly.

But number go down makes you feel bad! We can't have that!

So S3 now lies to make you feel better. You went 3-2 in the set? You're gaining points no matter who you were pubstomping, even when the truth is that it should've been a net negative. And as a result, there's runaway inflation for the top players at the far end of the bell curve who rarely, if ever, lose points. Conversely, players on the low end can get stuck in a defeatist spiral even when they sometimes get a few wins that should save them. Your rank is now a lot more dependent on the luck of the draw in who you get matched with, whether you got a set you'll go 3-2 or 2-3 in.

It's too common in modern games to have systems that lie to make players feel good. It boosts engagement. Keeps you coming back to farm those dopamine points, number go up. Frustrated players are at risk of leaving, so we lie to minimize frustration.

But lying isn't good, and there are consequences for that lie. Ratings are not accurate now, and if the ratings aren't accurate, then the matchmaking isn't either. If you're not happy with the state of S3's matchmaking, you have to realize this is a big part of why.

So what do you want out of a rating system? Do you want the game to lie so you'll feel better, or do you want the truth no matter how uncomfortable it is?

important meta question to ask -- is accurate rating and matchmaking really that important?
in japan, sure, maybe, because they have a good player density and they actually use xp to mean something
but in NA/EU? I feel like most people don't attach very much importance to XP at least competitively (which maybe was a shortcoming but still), and playing with "accurate" rating feels like GARBAGE
it's not like having an accurate rating gives you games against others of your skill level

 

[youre_a_squib_now]

Factors Glicko2 doesnt include in the calculation
-Specials used
-enemy splatted
-objective participation
-turf painted
-number of times splatted
-objective score
-Amount of time taken to KO

It only cares for the winner of the match, the biggest criticism I spent reading on twitter when players decide to expose about their teammates, and it is the same now in S3.

If I stay in the enemy base for most of the game just painting as much as I can, I might get #1 turf inker and enemy base inker and special user, but I'm not helping my team win nearly as much as I would be if I went back and fought with them. Should I really be awarded extra points for that?

Some of the games I personally had the most impact in were ones where I got relatively few splats, even for E-liter. I positioned very well, and kept enemy players away from fights they would rather be helping with. But there aren't badges for those. They simply can't make a badge for every possible factor that influences the outcome of a match. The only way to accurately determine how much any one player contributes to winning is to look at when that player wins and loses. Because the objective of the game should be winning.

 

[youre_a_squib_now]

it's not like having an accurate rating gives you games against others of your skill level

Maybe it would.
It seems logical to me that a more accurate measure of player skill would lead to matches with more closely skilled opponents

 

[Joy]

important meta question to ask -- is accurate rating and matchmaking really that important?
in japan, sure, maybe, because they have a good player density and they actually use xp to mean something
but in NA/EU? I feel like most people don't attach very much importance to XP at least competitively (which maybe was a shortcoming but still), and playing with "accurate" rating feels like GARBAGE
it's not like having an accurate rating gives you games against others of your skill level

It is! Sendouq uses raw OpenSkill points (well, an affine transform on it but basically raw!) as its SP rating. That’s why it does such a good job at rating players, even if players don’t like that sometimes they win and only get 0.07 SP for the win or if they see higher level players get more SP for a win than they did.

players not using XP (assuming it was raw glicko-2) for a competitive measure is fine, XP only measures how good a player is in that environment in relation to other players. But as far as for whether it’s worthwhile to have accurate XP more generally, I’d argue yes! Not only for bragging rights that soloq warriors can engage in, but also to monitor the health of the system. It’s much easier for me to monitor the health of sendouq when I know everything about the system can be described mathematically with the given inputs!

 

[vitellary]

this isn't exactly related to whether s2 or s3's xpower is better (s2's is better at the job XP is supposed to do, objectively speaking), but it is worth noting that from my experience, x-battles just feel better and more coordinated than anarchy battles do, so i think there is merit to ensuring it's a mode that people actually like, enjoy playing? i'm not sure whether that's worth the cost of completely inaccurate power scaling but i think it is worth considering that the mode being fun is still important, and i can definitely envision s2's just not feeling fun to play from a more casual perspective

on a different, more theoretical note, i wonder whether a fusion of the systems between s2 and s3 could work and be a relatively accurate way to calculate XP? i'm no expert on how XP gets calculated, but i wonder if a series could take the average XP of the opponents you faced, and give or take points whether you won or lost that series as a whole relative to that average power, with no inflation or guaranteed points. you could still end up gaining very little points but i feel like it would potentially be less common, and would keep s3's way of making it feel less volatile by being in a series format. not really sure whether that would introduce any new issues though