Grand Arena Championships - Matchmaking
CG said they re-did the matchmaking in the GAC and I am having a really hard time believing that. I am just over 8.1 million GP and I have 4 GL's at R7. In Round 1 of this current GAC my opponent was just shy of 9 million GP with all 6 GL's at R8. I won because my opponent did not set defense, so the AI set auto-defense and I was able to clear. In Round 2 my opponent has 8.3 million GP, all 6 GL's, with JML at R9 and the rest at R8. This is not improved matchmaking, it's hot garbage.
0
- 60.8K General Discussion
- 9 Developer Q&A Archive
- 2.6K New Players
- 30.5K Arena and Character Strategy
- 4K Individual Events
- 4.6K Guild Events
- 4.6K PVP and Rivalry
- 50.1K Guild Recruitment
- 731 Art and Media
- 4.1K Off Topic
Howdy, Stranger!
Replies
Matchmaking doesn't care about your GL's it cares about your skill rating. What was the skill rating of your opponents vs yours? I bet they were within 10 of each other. You might not LIKE the matchmaking but its working exactly as they told us it would.
|| Grum the Mighty swgoh.gg || OldBaldGuyGaming Youtube
Everyone's skill rating at the start of the bracket will be the exact same. It's why this system is hot garbage. You will always be matched with folks in your division who put the same effort you do into GAC now that crystals are in play.
I wish there was nuance to the skill rating, but everyone goes up or down the same amount after each round, based solely on the outcome (win/loss). Banners don't matter.
Yes we understand this.
The problem is they could have seeded the initial skill rankings 500 smarter ways than they did. Their implementation was lazy.
1) Instead of putting all of D1 Kyber at the same skill, bracket it.
10M+
9.8-10
9.6-9.8
Etc
Rather than everyone 7.6M and up combined
2) Weight the algorithm based on having high performing characters (GLs)
3) Weight the algorithm based on lifetime banners
4) Weight the algorithm similar to HotBot gear score and mod score
Literally almost anything else than just grouping up people with 3m difference in GP
Regardless, it will all be sorted out eventually.
There are problems with every method.
Any method grouping people with a 3M gap in GP is by far worse.
I excluded #1 from my post for a reason.
1) Div 1 being flat seems a bit rough, I agree. I would suspect the reasons were population size and/or the fact that past a certain point the additional is just bloat which has no impact e.g. relic'd Tuskens and such.
2) Has similar issues as attempting to explicitly match by certain toons / ships: what is valuable and how much value it has varies by roster GP range, roster composition and new toon / ship releases
3) Lifetime banners are a combination of performance and participation and as such are not a reliable measure of performance
4) Abstract values generated by subjective algorithms which ignore the importance of mod allocation and squad dynamics not to mention player GP. Should a 1m GP and a 10m GP player with the same mod score be matched?
Total roster GP was as good of a starting point as any. There will be a significant amount of churn the first few GA brackets as actual player ability asserts itself but after that things should settle down.
Total roster GP was used only in the newbie brackets.
They, inexplicably, decided not to use it for Kyber1.
Unless of course, you think that having the same 50 players or so fight each other for the next 10 seasons or so is a very good system.
K1 is the largest division
I thought they stated that 20% of the population is in each league, and then the divisions were split from there, 10% being in div 1.
So that would make k1 2%. That's why the massive gp spread was there, as the tip of the spear is small but sharp in shape. Is there some data that shows k1 to be larger than this?
They might be the loudest though.
I just suggested a couple of potential reasons.
By GP range perhaps. By population? I'm gonna need to see evidence to back that up.
And it's working pretty favorably for me.
Not true. The amount of increase or decrease to Skill Rating also involves what your opponent's Skill Rating was. From the Q&A: "We are using a modern bayesian matchmaking system, similar to those used by other modern competitive games. The Skill Ratings of both you and your opponent are taken into account in addition to the Win/Loss factor." If you look through the rankings, you'll see people with 1-3 points difference in Skill Ratings from other people. It's a small impact on the points from an individual round but eventually people will start to separate more in terms of their Skill Ratings. Unless you're at the very top of the pack, if you win more than you lose, then also you'll eventually start playing people who have dropped down in Skill Rating and vice-versa. So you won't always be matched up with people that put in the same effort that you do...that's just incorrect.
Literally everyone on my bracket in Week 3 has the exact skill rating - 3891. Everyone went 2-1 in each week, losing the same rounds (3 in week 1, 2 in week 2).
I'd love to see any proof that someone has a bracket where someone is in a bracket with anyone else that has a skill rating difference.
Mine and my opponents curves. Apparently 1-2 then 2-1 would get him to drop those 6pts
It's also a mismatch
Interesting. You're rising, opponent is falling. So, basically, your paths are crossing due to skill rating matching at this moment.
If you start at the max SR, then, yes, you will fight folks with the same results as you. Every time.
The ranks in your subsequent screenshot further prove my point. There is no "matchmaking." They are literally sectioning off groups of 8 based on standings after each week.
It will eventually. The system needs time to settle in:
Now I hate GAC and never play. My legacy championship Points are 500 000.
The others are 900-1mil. So they have double the experience then me in gac.
This system is so those people and myself play each other so it balances out. Sure I have more in the tank but they know way more teams and counters and will beat me until I eventually learn to play better in gac. Then I won’t see them much anymore because I will move up and face what they faced. Higher GP but probably don’t care players and I’ll be the little fish.
Matchmaking is based on skills ranking. We are matched on our skills ranking. It's working exactly as described
Yes. It would be more appropriate to refer to it as "Ladder Pools" or such instead of matchmaking since they aren't trying to make FAIR matches, they are allowing the ELO/Skill Rating to determine that which it currently doesn't.
All you have right now is an opponent with a similar skill rating as you in your pool.
|| Grum the Mighty swgoh.gg || OldBaldGuyGaming Youtube