February squish - we're under 2k in kyber 1! We... win?
The squish formula hasn't changed since at least last November, and kyber 1 is steadily shrinking. We're down under 2000 this month (1998 players in kyber1 in week 2 of the Feb 2023 season). When does it stop?
Figures:
- Squish compiled across leagues and divisions from November 2022 to February 23 (note: just did ~10-15 players in Jan Feb to confirm they fell on the same lines as previous months; Nov and Dec were the biggest data pulls. I have 278 data points represented here.)
- Raw player numbers in kyber1 over time. As the months went on we started collecting information every week rather than just once a month to see the shape. You can see that players do move back into kyber1 through banner drift, but not enough to make up for the squish each month. This demonstrates that the squish formula is flawed for this division.
-
Figures:
- Squish compiled across leagues and divisions from November 2022 to February 23 (note: just did ~10-15 players in Jan Feb to confirm they fell on the same lines as previous months; Nov and Dec were the biggest data pulls. I have 278 data points represented here.)
- Raw player numbers in kyber1 over time. As the months went on we started collecting information every week rather than just once a month to see the shape. You can see that players do move back into kyber1 through banner drift, but not enough to make up for the squish each month. This demonstrates that the squish formula is flawed for this division.
-
4
- 59.7K General Discussion
- -35 Developer Q&A Archive
- 2.5K New Players
- 30.3K Arena and Character Strategy
- 3.1K Individual Events
- 5.5K Guild Events
- 4.6K PVP and Rivalry
- 50.3K Guild Recruitment
- 744 Art and Media
- 4.2K Off Topic
Howdy, Stranger!
Replies
Even if you win more than you lose (i.e go 5-4) you will still drop. This has affected both your ability to climb into or maintain in K1. The implications on the daily crystal payouts to players and the savings it generates for CG are obvious.
They applied a dynamic squish to a system with static boundaries.
That doesn't mean that CG really understands the issue. Clearly, they don't understand Kyber 1 counters, as we can see with immediate nerfs to so many releases lately. So I don't believe that they necessarily know about the issue or that it is WAI for them. Maybe. but they just may not gronk the situation.
If meat brought it to them, then they've already made their decision most likely, whether they understand it or not.
In the end kyber 1 will be made of 1…
Did he say something else? Are they looking into it?
Not really. He was asked if he saw this thread, and he replied with "I did and made a note of it."
When pressed if it had been previously mentioned:
Basically, CG is fully aware of the squish, what it's doing, and how the community feels about it. Unclear if anything will change.
Thank you very much! I appreciate it.
Then let's hope they will do something 🤞🏻
yes, they are looking into it. They've contacted us at 50ShardsOfKAM. They've specifically invited us to comment on mathematical techniques that would be better at accomplishing what they intend. Fortunately we have a couple professors and one person with experience in mathematical topology, so we're not a dumb group, but
1) It's hard for us to know (and by "us" I mean "people at 50SoK who aren't me since I'm not the math expert) exactly what would be better since we're not entirely sure of their goals,
and
2) Even if Tatiana or someone else comes up with a better mathematical transform, there's no guarantee that CG will think it's as good as we think it is.
So communication is happening. They're thinking at their end. Tatiana and other math experts are thinking at their end. What will happen? Who knows! But it's interesting for me to have a front row seat during mathematics debates I'm a couple semesters shy of understanding.
Oh wow, that is good news! I very much appreciate the insight, thank you so much!
I think it's really nice to see that they are communicating and seem the be willing to change something.
The only thing that confuses me a little is that one would think that they have people for those equations.
When I think about my own job and how I determine my own priorities, it’s often on a “how much of an emergency is this?” basis. I have too many things that need to be done, and not enough time to do them in.
We’ve spent a lot of time thinking about squish because it affects us and our guildmates. We’ve all talked with people who think they must have gotten worse at the game because their division has dropped, or who has tried to figure out why their kyber 2 fights are so much harder than they used to be. It matters in our hobby, so we spend our time digesting it, calculating it, explaining it.
CG is managing the experiences of hundreds of thousands of players. When 10k players a month were dropping out of kyber, they responded. My guess is the slow drip of a few hundred players out of kyber 1 each month probably just didn’t hit their radar, especially if they were visualizing on the league subdivided by division. I had to play with visualization tools for a while before figuring out how to clearly demonstrate the problem… even though it was my direct goal.
But it occupies our attention, because it is us and our guildmates. And because Kyber 1 is a status, and when we lose that status, it kinda hurts our feelings ;-)
The whole goal of the feedback section of the forum is to bring problems to CG’s attention that they might not otherwise realize ARE problems that matter. It looks like we’ve done that! Now we just have to hope we were convincing enough to jump it up the to do list.
They likely dont want the upper skill rating range to grow infinitely, but because top 1 player keep winning more than loosing. (If you check the ladder there is currently the 266 points difference between 1st and 11th place) Then his rating only grows, making higher difference between him and K1 lowest number bracket.
And if i am correct and they dont want the highest skill rating to be above some number so, they just keep removing that amount of points from all ppl forcing them to lower divisions because more and more space in K1 is just empty.
In the end if Top 1 will keep increasing his rating it has no solution except completely change the squish not by skill rating but by player numbers, but then the highest skill rating will just keep growing indefinitely as long as any player will always win more matches than he loose.
As a thought experiment- what if there was no squish at all? We can speculate that the ladder would get a bit top heavy over time (mitigated of course by new players) but if that genuinely ever presented any problem, there’s always the option to create a legendary league with a new cut off (& so on ad infinitum). On current evidence, however, this doesn’t look like a huge problem.
Any algorithm will have unintended consequences- doubly so if the underlying thinking about desired outcomes is a bit muddy. Personally I trust the player base to regulate the ladder through competition more than I trust CG’s maths - and I love the fact they seem to be spending a bit more time going after cheaters so why not let them concentrate on that and we can have a cleaner competition?
You didn't buy the battlepass, you are demoted to Aurudium
Not holding my breath, of course, but rebellions are built on hope.