A Proposed Rework to Skill Based Matchmaking
NOTE: It's only lengthy because I tried to answer as many questions as possible
It NEEDS to be much more dynamic, adjusting your skill rating after every game, either raising it slightly or lowering it. But this being said there should first be some sort of grace period of a few placement matches. The skill of those should be at the 50% on the skill curve. The SBMM should factor in engagements taken(who engages first, results of engagement), accuracy at different ranges, people engaged to people killed ratio, and play style.
Who engages first is important because of you engage first that means you either tracked that person or were sharper than that person(most of the time anyway), so that should be taken into consideration. A higher rate of this would indicate a higher skill rating. A lower rate of engaging first would subtract from the skill rating because you're getting caught off guard. If you manage to escape from an engagement that means something, and should add to your skill rating. Dying in that engagement is self explanatory. Winning the engagement should also be self explanatory. Though it should be taken into account whether an attempt to use healing items and abilities was made to indicate whether the fight was "fair" and both opponents had an equal chance of winning. Winning against something with no bodyshield when you have a purple should greatly diminish how much it adds to your skill rating, but it should still add, and winning an engagement where you're at a disadvantage should greatly add to your skill rating.
Accuracy should be self explanatory as well. If you're hitting say 70% of your shots consistently that says something about your skill rating and it should be raised. But an average should be taken from the entire match to find out the realistic accuracy.
The enemy's movement should also be taken into account, such as if they climbed, strafed, crouch spammed. Hitting a still target will give you 100% accuracy but hitting a crouch strafing target at 100% accuracy isnt the same thing. So it should be a measure of someone's accuracy on differently formed targets that is later averaged. If there's not enough data from the match I'll get to that. Also it should not change the skill rating if there isn't enough data.
People engaged to people killed. If you're seeing a lot of people but not taking a lot of gunfights that means something. It means you understand positioning at the very least but aren't taking a whole lot of shots or are prioritizing other things such as loot or positioning. This should also be factored in and add to your skill rating. Not seeing a lot of people but killing everyone shouldn't much add to your skill rating because there isn't enough data from that. Seeing a lot of people and killing a lot of people should greatly add to your skill rating for self explanatory reasons. Not seeing a lot of people and not killing anyone is also self explanatory.
Play style can help adjust for these sorts of things. It should track how long a player spends in each location, what they've picked up, and how long it takes them to move from location to location. If a player takes a long time to loot in a location, these sorts of players should be matched together. If a player consistently plays ring, these players should be matched together. If a player doesn't spend much time looting and is instead focusing on initiating, these players should be matched together. This creates a more engaging environment as no one has to sacrifice their play style to suit the play style of the lobby.
I don't think total kills or total damage at the end of the game really indicate anything as I've addressed other underlying aspects that factor into total kills and total damage like accuracy and people seen in a match vs people killed. There should be a period of 3 or 4 matches before a decision in the algorithm is made to adjust the skill rating up or down to make sure it wasn't just a fluke or a good game, or just an unlucky game because it is a battle royale. This allows for collecting an amount of data and making sure that no one is hardstuck anywhere. This allows you to also troll with your friends, play with people of lower skill levels, all of that.
The skill rating of a team should also be averaged and matchmaking should be based around that average, not just shoving all 3 out into the pred lobby cause one guy is a pred.
Solo queue players should also have this apply to them and be part of the average of their team of 2 randoms, and then get a lobby based on that average. This is what I see is the ideal SBMM system.
Key points: VERY dynamic, NO getting hardstuck, taking an AVERAGE.
I posted this already but no one read it so I'll post it as its own thing.
