Relegated in GAC but winning?!?!
I’m looking for anyone to explain to me why I’m being relegated for winning in GAC? In this last GAC I lost the first round (opponent was 1M GP higher than me) but then won the next 2 rounds (albeit marginally, but still won) and then I got relegated. Am I being punished for using strategy instead of being able to just power through all my opponents?
0
Replies
- Grand Admiral Thrawn
If seems that if you keep winning 50% as intended, you’ll eventually be squished down a bracket…and maybe more if they keep squishing.
I don’t understand how this is justifiable, but some others have said it’s standard for this type of ranking system.
Either way, it sucks.
#ditchthesquish
I don't get what the points you cited are supposed to mean? Same situation, started last week at 2,811 points; win, loss, win; and finished at 2,810 points. How does going 2-1 against opponents 1.5 - 3 million GP higher mean I lose points?
You're not penalized but rewarded with playing against the best! You're absolutely amazing if you made it this far without any GL so embrace your fame and let those strong enemies be the motivation to get even better!
- Grand Admiral Thrawn
CG randomly removes or gives accounts points at the end of the 3 rounds. That's the squish. We don't know how it works. Why some ppl lose 100 points and others gain 100 points. It just happens and it robs some of us of the division we're in despite having a 50% win rate or slightly higher than 50% win rate.
The GP is irrelevant.
In most Rating systems, the points you win or lose also depends on the rating of the opponen you had in each of the battles. So probably it has to do with that.
Each division is based on a percentage of the player base. They have changed to this approach because the system can remain the same as more players join the game and the devs do not need to spend time on readjusting divisions as the game grows - which allows them to focus on developing new game modes, characters/ships, etc. If you recall from the old system, it grew very top heavy as our rosters grew, so bad to a point they needed to go back in and add more divisions. So each league (i.e. Kyber, Chromium, etc.) has divisions (i.e. Kyber 1, 2, 3, 4, and 5). Kyber 1 is the top 10%, Kyber 2 is the next 25%, Kyber 3 is the next 30%, Kyber 4 is the next 25%, and Kyber 5 is the final 10%. For the examples, I will be talking about Kyber, but it applies to all leagues.
The struggle with the percentage threshold is it is relative to your peers, so if you are close to that threshold, you can win a match and not be technically in the top 10%. Note, this is why you can technically go 2-1 a week and get demoted (because the people right under you went 3-0). They have clearly defined thresholds (like Kyber 1 is 3600 skill rating), so you have clear expectations to aim for; however, because of inflation, you will have more than 10% of the Kyber league player base end up above the Kyber 1 skill rating threshold (and likely less than 10% of the player base in Kyber 5). Same thing for each division: some will be slightly larger, and some will be slightly smaller.
For rewards if you have a skill rating above the threshold, you get rewards from that league (even if you are not technically in the correct percentage - which is great, you can get rewards you don't earn). However, if divisions remain with the slight imbalance, then the equilibrium will slowly get more and more worse - and eventually need the devs to go back and readjust leagues. To avoid this, your skill rating is slightly changed at the end of each season to match where your percentage is - this is called the "squish". Think of this like in school when your professor grades on a curve. Your spot in line doesn't change - the person who was right in front of you is still right in front of you, same with the person right behind you. If you are in the top of a division, you will normally see your skill rating go down and if you are at a bottom of a division your score will typically go up. This is why is might seem "random" because some increase and some decrease in skill rating - but it is just making sure that the person in the 11% percentile has a skill rating which matches the 11% percentile.
my alt was toward top of the division and squish raised it up. 2x. main was middle of division, dropped. Main was lower end of division, also dropped. That's the opposite of what you said. Where is the source of your "information"?
P.S. 70 points is 2 wins so not "slight" changes. And I've seen ppl be squished over 100 points before.
P.P.S. Where did CG say they will create more content modes bc of GAC?
I said "Same thing for each division: some will be slightly larger, and some will be slightly smaller." and included normally, generally, and typically in multiple other locations because it completely depends on the population of your division. Your experience perfectly demonstrates this - two of your accounts where in over-inflated divisions which were adjusted down and your other account was in the opposite, an under-inflated division and was thus increased. This shows perfectly why the squish needs to happen, because each division will shift slightly and need to reach back to the proper equilibrium.
It is very important to differentiate two things - (1) the actual definition of each division and (2) where you think you are in reference to that division. Technically, Kyber 1 is only the top 10% of players in the Kyber league - that has NOTHING to do with the values listed as the thresholds (which I will call "goal posts") listed in game. Where you think you are with regards to that percentage is based on your skill rating and the "goal posts" which are listed next to each division: for example, Kyber 1 is 3600 skill rating. Now if you have 3601 skill rating, it does NOT mean that you are in the top 10% of players in the Kyber league.
The squish adjusts your skill rating to better match the percentage where you are. The squish happens because they want to keep the "goal posts" the same. If they didn't do the squish, then the skill rating for Kyber 1 would have to be increased (or decreased depending on if there are over or under 10% of players in that league above or below that threshold) to a higher value, for example 3670. This means over time that the thresholds would expand as mentioned here (https://forums.galaxy-of-heroes.starwars.ea.com/discussion/252923/grand-arena-seasonal-skill-squish#latest) which explains (very briefly) why the squish is required.
So in the example you gave (70 skill rating change ~ 2 wins), that is how far you where away from the technical definition of the next division (i.e. Kyber 1 being top 10% of players). So if they got rid of the squish, then the "goal posts" would have to change to then reflect that 70 skill rating difference. In the case of the squish, the overall "goal posts" stay the same and CG doesn't have to go through and redistribute the leagues and divisions at some point in the future. So in that case where you drop 70 skill rating, where you think you are in reference to the true percentage for that division was off - so you think you are being screwed over, but you are still the exact same skill rating away from the technical cutoff of top 10% (or whatever value based on the division).
They never said that they are specifically making new game modes but have mentioned that this is a long-time fix which avoids the problems of the old system. So if they don't need to keep adjusting with GAC, they have more time to work on everything else, which I said "the devs do not need to spend time on readjusting divisions as the game grows - which allows them to focus on developing new game modes, characters/ships, etc."
You seem confused. Kyber 1 cutoff isn't 3600 points. It was originally ~3540ish and has been raised to 3610 after the first of the "new" GAC events. They also did the squish. So CG is both moving the goalposts AND squishing accounts, which goes against everything you're asserting. Also CG took my 4.5 mil alt and raised it up thanks to the squish far higher than it deserved or needed to go. it's now in K5. where it's being beaten to a pulp bc i don't have 22 teams to be able to do offense and defense there. So I have 3 entire weeks of gameplay of being beaten the **** out of. It sounds like way more fun than it is, believe me.
The biggest issue for me, really, is the arbitrary nature of this squish. My alt, to be extremely blunt, is a terrible account. I made it as a fun account and it does not chase metas. It has terrible mods. It has 1 GL. All decisions I made and I'm not hiding behind or regret at all. But it does not deserve to be in K5. CG artificially raised that account, who barely won more than it lost (and usually against opponents who didn't set defense), without any cause. They gave me enough points to promote me to a higher league before the new season even started ffs
But what stands out to me the most here is your continued assertion that CG doesn't need to keep readjusting GAC. They have to constantly adjust GAC bc of their laser-focus on retaining these really dumb divisions at their "proper" percentages. That means both moving the goalposts for divisions AND squishing accounts (or as I call them, stripping wins from players and giving free wins to others thereby removing player agency). Crumb has stated that the squish doesn't impact player progress. That's actually impossible. The squish removes skill rating points, or gives skill rating points. I have seen in both of my accounts it relegate me out of a division and promote me out of a league before the first match even began.
If it worked as you claim it does- the goalposts don't move and the squish is the only thing involved- I would still disagree with your argument that this is a great system but you would at least have an argument that the system is working as intended. But the system as CG designed it is ALWAYS going to have bloat in several divisions of each league by the end of the event, which they will ALWAYS have to therefore alter in some form or just succumb and make higher leagues. (I also would argue that they made Kyber and in particular K1 far too large initially and that's on them)
The system is not a great one. It's move goalposts. Do squish. With CG's system there's no chance for accounts to "settle". They're going to constantly be artificially moved so CG can make their division sizes fit. Which means CG has to change things around frequently. This new format is far more hands on than their old format.
The idea of moving tier thresholds AND adjusted points is ludicrously complex. This is a closed system, the number of players entering and exiting is similar, and relative strength by GP has a defined maximum.(We can assume the attainment of lvl85 has a relatively static floor too) Therefore the divisions and tiers under a normal curve have a similar profile. Chrome will have 50% of the players, Aurodium 15%, and Kyber 10% (curve flattened to match dev explanations of divisions). From the initial seeding, your position should only change relative to the record of those around you. Win enough and you advance, advance enough and you get promoted. If I continually finish 2-1, doing so against other players who typically finish 2-1, I will plateau in the 83rd percentile overall. That's median to Aurodium so perhaps A3 is where I belong.
But, what about those players who are 3-0 or 2-1 in A1 and need to move up to K5, and those in K5 who are 0-3 or 1-2 and need to drop down to A1, but whose overall percentiles(i.e. skill rating) don't fit the current curves for Aurodium and Kyber within the overall population? This is what's causing the "squish", the need to arbitrarily change a player's rating to change his divisional assignment or rank. And this is a massive time suck because it must be done MANUALLY.
Solution, lose the squish, abandon the divisional curves now that everyone has had a couple seasons to be re-seeded and tie continued ranking and advancement to win-lose only. This will create greater incentive to be active because taking weeks or seasons off WILL drop you. Wins mean promotion no matter what. It will also ensure that the results are fairly managed, transparent, and verifiable!!
For example, within a division, in tiers 1, 2, 4 & 5, 3-0 or 2-1 means advancement; in 3 only 3-0 means advancement. In 1 & 5, only 0-3 means demotion; in 2, 3, & 4 1-2 means demotion. This will bunch everyone into the middle ranks, push the winners to the top, and weed out the laggards, while not exceedingly punishing those who momentarily falter or have RL impose. Good, active players can quickly recover and advance. Those who are building their rosters will advance over those who are stagnant or building slower.
Specifically speaking on the squish not being random: I went through and collected data from 3 guilds on swgoh.gg and looked at how much their skill rating changed from the "squish" and plotted the data here. It is not a large sample size, but from the trends are clear. As you can see, it is definitely not random and follows a trend per division. As I mentioned earlier, those are the top of each league (such as Kyber 1 and 2) are seeing a decrease in skill rating and those in the bottom of a league (such as Kyber 5) are seeing an increase in skill rating.
Nice work.
So basically we're all pushed to the middle of our current league.
Awesome job! Yes, exactly what I was describing, the curving of everyone towards a median within a division. Is your x-axis skill rating before "squish"?
Yep.