Destiny 2 News Roundup – SBMM Update and King’s Fall Stats

Bungie was back this week with their This Week At Bungie Update and they gave us an in-depth update on SBMM, plus a whole load of stats for the latest returning raid in Destiny 2, King’s Fall. Today I’ll round up all the latest news from Bungie and Destiny 2, so let’s get to it.

King’s Fall Stats

Another World First has come and gone, this time bringing back the beloved King’s Fall raid from Destiny’s past. Players from all over the world gathered their fire teams and headed in to take out Oryx one more time. It was a close race! We had many teams all vying for that top spot, but only one could be victorious. The winning team was a shocker. Not because they didn’t put their best foot forward, but because this is their third time in a row winning World First! I don’t care who you are, that’s impressive.

Clan Elysium is back with a three-peat win, with an all-star lineup of Saltagreppo, Cruz, Kyros, VileFate, Moople, and Quazz along for the ride. They dodged, dunked, and made a few hilarious missteps along the way, but you know what? They had a lot of fun and the look of sweet, sweet relief on their faces when they realized that they had finally reached the coveted finish line was a sight to behold. A huge congrats again to this team. It was definitely a wild ride from start to finish.

So now we know who won, but what about some other fun facts to herald the return of King’s Fall? Let’s break it down, shall we?

King’s Fall, round Destiny 2, fight:

Total players that entered the raid
Normal: 458,120
Challenge: 35,678

Total Guardian deaths
Normal: 33,665,322
Challenge: 2,825,235

Players that cleared Totems
Normal: 158,034
Challenge: 31,649

Players that cleared Warpriest
Normal: 63,648
Challenge: 21,727

Players that cleared Golgoroth
Normal: 52,958
Challenge: 19,341

Players that cleared Daughters
Normal: 49,873
Challenge: 19,028

Players that yeeted Oryx into oblivion
Normal: 49,245
Challenge: 17,107

Total hours spent in King’s Fall
Normal: 1,989,371
Challenge: 208,295

So basically, we have a lot of emblems to give away, and quite a few raid rings to dole out. For those that do get emails (we’re still working on that), you’re going to want to keep an eye out for the next Bungie newsletter. In that, we’ll be breaking down your own #RaidComplete report to see all those ups, downs, and sideways yeets off of cliff edges.

Crucible Matchmaking Check-in

This Season, we implemented the loose skill-based matchmaking (SBMM) that we talked about back in August. To talk about the launch of this revised feature is Principal Designer Alan Blaine, here with a “vibe check” in the PvP-verse of Season of Plunder. Take it away, Alan!

Alan Blaine: Hi folks! That sure was a great start to the Season of Plunder in ritual land. We have seen a significant uptick in the number of players and amount of time spent within Crucible. While we are excited to talk about the bright future of Crucible development, I’m sure a lot of you want to hear about the loose SBMM we added to Control.

As we talked about in an earlier TWAB, one of the functions of the Season 18 Control node is to help us tune a solid 6v6 matchmaking setting before we start implementing new matchmaking features, like team size and dynamic skill ranges.

We’re going to start with a fun tale about live game development, just to illustrate why we like to run smaller tests before rolling things out wider. If you just want the weekly stats and future details, skip down to the Stats from Week One section.

CONTROL WEEK ONE

Starting at reset on August 23, loose SBMM in Control was supposed to go live. Instead, players who joined Control found they were matched up against anyone and everyone. Skill didn’t seem to matter, and neither did platform. Normally, PC players are separated from console and Stadia players via matchmaking. So, what happened here?

During that day’s update patch for Season of Plunder, the file with the new matchmaking settings was not updated, and instead Control used our default matchmaking settings. Once we found the bug and figured out what was going on, around 7 PM on the same day we got the file updated on the servers, and everyone was using loose SBMM. Hooray! This illustrates a couple of key points:

We do our updates and weekly resets on Tuesday rather than Friday, so when something unexpected happens, no one must pull weekend heroics to fix the game.

We make changes in one location first to make sure it is stable before expanding it to other uses, especially when we are exercising systems that haven’t been changed recently.

DEFAULT MATCHMAKING

The default matchmaking that everyone played until 7 PM is never actually seen in-game if things go well. It is designed as a fallback that “just works” and finds people to play with but doesn’t do anything clever.

Default matchmaking ignores platform, skill, and almost everything else. The only thing it takes into account is latency, but only for the first 15 seconds of matchmaking. After that, it’s no-holds-barred. So, on the morning of August 23, if your Control match took longer than 15 seconds to matchmake (which it almost always does) it was basically just picking the first 11 people it found, no matter how good the connections to you were, what platform they were on, or what their skill was.

And of course, this could cause laggy players, bad hit registration, and matching with players far outside your normal skill band. This is why we sometimes spend multiple minutes on our various matchmaking thresholds, making sure latency is solid as we can get it.

STATS FROM WEEK ONE

So, as of 7 PM or so on Tuesday night, we have had loose SBMM on in Control, and we have some interesting results! Let’s dive right in:

POPULATION

In the first week of Season of Plunder, 140,000 more hours of Control had been played than in the first week of Season of the Haunted, and we had around an 11% increase in the total number of players playing Control.

MATCHMAKING TIMES

Overall, our average matchmaking times went up by an average of 5 to 10 seconds. That’s a good indication that the matching is generally working but isn’t showing our worst cases—the lower population segments (extreme low and high skill).

For the highest skill band, less than 0.1% of the population, matchmaking times average around 90 seconds during high population times, spiking to just over 200 seconds at low population times. For the lowest skill band, we see matchmaking times between 120 seconds at best and 240 seconds at worst.

Just looking at matchmaking times, the current settings look like they are getting us where we want to be. However, we are cautious about making too many assumptions right now. We have plenty of anecdotal evidence of bad connections in those high-skill games, and week-one is one of the highest population moments of the Season.

SKILL DIFFERENCES

The skill differences we see in Control matches are pretty stark. Without SBMM, only 10% of matches had 600 or less skill difference between the highest and lowest players. With SBMM on, we see that 80% of games have that separation or less. For 90% of games without SBMM, the average skill of all players within a game was between 300 and -100. Now we see a significant number of games with a high average of 600 skill, and regularly as low as -500. (With a few games each day going as low as -900 average!)

GAME OUTCOMES

Looking at game outcomes, where our work really matters, we see more interesting data:
Mercy games are down 4%. Not as much as we had hoped, but it has been shrinking a little day by day.
Games where the score threshold has been met remain steady, and games that go to time rose by 4%.

These are all within range, and we expect that the first few weeks will be more chaotic than normal. As players re-adjust their playstyles from “optimized play against a wide variety of players” to “optimized play against similarly skilled players,” the skill system will adjust to reflect those changes. We would expect mercy games to shrink slowly over the coming weeks. If time limit games rise consistently during that time, we will likely look at the score goal and lower it to compensate.

As far as score and kill differences, we see a similar set of incremental improvements. Games where the best player had 30+ kills more than the worst player went from 9% of games to 2%. Games where the best player had only 10 to 19 more kills than the worst player went from 35% of games to 55%.

QUITTING

We have seen one worrying trend in the data: the percentage of players quitting before the end of the match has risen from 8% to 12% in the last week. This is especially bad with matches designed to be balanced with 12 equally skilled players. We are still investigating to see if this is localized to a specific cohort or playstyle, or if this is a natural player reaction to a new system. This percentage may reduce over time. Stay tuned!

FUTURE PLANS

There are a few points of concern that we are addressing:
The number of poor connections during matches.
The number of players quitting matches early.

Other than those, most of the analytics have been positive, considering this a new experience and will take folks some time to get used to it.

For now, we are taking a few steps:

We are extending the time that loose SBMM will sit and wait for the best connection quality between group leaders. Hopefully, this will increase the overall connection quality when matchmaking goes past two minutes.
We have some new analytics coming online soon which should give us a better picture of the connection quality between all players within matches, not just the group leaders.
We are going to be adding quitter protection to Control in a future patch. We hope to discuss more in a future TWAB.

For the immediate future, you can look forward to Iron Banner next week with the new Eruption mode. Need a refresher on what that is? No worries, we got you covered here.

Eido

What did the ocean say to the pirate? Nothing, water can’t talk, but at least it waves. Eh? Ehhh? While we haven’t mastered communicating with water yet (but we do have those sick pirate puns always at full mast), we do have a different Expedition location to uncover for Guardians looking to make Eido proud. As we should, because she must be protected at all costs.
2
This week, we’re heading to the Cosmodrome to loot all of the booty that you could possibly want. For those that may just be jumping into Season of Plunder for the first time, Expedition is the new 3-player matchmaking activity where pirate hopefuls must band together to hunt treasure while fighting off those who may want to prove that they want it more. But we’re not going to let that happen, right? Of course not. We’ve got Misraaks and Eido on our side. What could go wrong?

Let me know in the comments what you think of the SBMM situation, plus the King’s Fall stats too.